When are Goals Scored?

When you are chosen by the public (NHL) hockey analytics gods to do research for them, you learn early that shots are the oil of the research. You might want to avoid it, but at some point, it is omnipresent in hockey research. Playing with public datasets is fun, but it often comes with drawbacks. NHL data has a lot of limitations for analysis (this is why we all need to be a bit careful). In the field of public NHL analytics, defense is defined as aggregated sum shot quality against a team/player/group of players (known as Expected Goals Against — xGA) while offense is usually the aggregated sum shot quality generated by a team/player/group of players (known as Expected Goals For — xGF). It is a bit problematic, yes, but I promised you I would talk about Shift Age earlier, so let me get back to my initial train of thought. If I made an extremely long (probably too long) discourse to introduce the concept of shots in hockey research, it is to justify why I will include shots and goals in the analysis.

At first, when I decided I would dig into Shift Age, I first wondered about how old are the shifts when goals happen.

In this figure, we can understand that goals are scored when players are still fresh. It makes sense considering that goals are generally scored by quick plays, quick transitions, and don’t require complex buildups. What’s more interesting is to see that the distributions covered by the Scoring and Conceding team are overlapping, suggesting that goals scored because of a Shift Age advantage are rare. In other words, this reflects that coaches do a good job at preventing getting scored on because the other opposing team got fresh players against an already tired group. So, even if fatigue (accrued player shift age) does occur, it does not explain a lot of goals against.

It gets much more captivating when we compare by game strength. Even if goals are still scored early in players shifts, goals on PP take more time to prepare. Teams generally take their time pass the puck to open up passing lanes to find a trigger player to get the optimal shot. I don’t want to put too much stock on Short-handed Goals, but it is still insightful to see that they are scored early in the shift. Ideally, the SH shifts are shorts and a player with a long SH shift is more likely to want to want to rush to the bench instead of scoring a goal. It is also important to note that scoring a goal is almost never the main objective when a team is shorthanded, so the sample size of goals is really small compared to the other strengths, so the results are not really significant.

This comparative violin plot summarizes all the findings we made earlier and gives us an even more comprehensive perspective on the relationship between Goals Scored/Conceded, Shift Age, and Game Strength. Again, it is striking how, when goals do occur, they tend not to occur at extreme shift ages. Did you notice too how the lengths of the violins are different? The EV violin stops around 90s, PP near 120s, and SH closer to 60s. This teaches us something compelling about hockey: the point within a shift at which goals happen varies depending on game context. When goals are scored, Power Play shifts tend to be much older, Short-handed shifts are very young, and Even-Strength shifts fall somewhere in the middle.

What About the Shot Attempts?

The first layer of the analysis is done. We looked at goals. Now, let’s dig into the analytics-doubters favourite hockey statistic : Corsi. It is also called shot attempts. Not a big surprise that the Shift Age distribution for Goals is extremely similar to the Shot Attempts’. Goals are not happening in those “fatigue moments”. This graph reflects that Goals and Shot Attempts behave similarly in this context. Finally, being a fresh player is not only important off the ice. Teams shoot and score more when they have fresh players on the ice. I don’t want to overwhelm you with charts, so I won’t show it here, but the distributions for Goals and Shot Attempts Conceded are almost identical to the ones you see up here. So, the conclusions are similar.

When we have Shot Attempts and Goals, and we divide one by the other, we get a really fun metric: Shot Conversion. The research is finally becoming fun! Shot conversion at EV and PP is quite stable until it reaches the 60s bin, where we see a significant rise. It is important to note that 60s is longer than the average NHL shift, so seeing a notable change in shot conversion at this key point raises questions. I wonder if this could be due to a goalie losing focus or defensive breakdowns after sustained pressure (both ideas would require additional research and are not in the scope of this blog), or a pattern where the offensive team is getting better looks late in shifts. Short-handed results, once again, don’t tell much.

Looking at the % change of shot conversion from the previous bin, we can build more confidence that shots conversion is higher at the end of longer shifts. There is maybe a new variable to include in your expected goals model!

The Fatigue Factor, is it real?

For the last chunk of this article, let’s end with a bang! We want to analyze how often goals are scored (or conceded) by players who are on long shifts, and comparing the scoring team vs. the conceding team.

For that, I set thresholds every 10s starting at 20. For each threshold, I measured:

• P(scoring shift ≥ T) = probability a goal was scored by a team whose skaters were on a long shift (≥ T seconds)

• P(conceding shift ≥ T) = probability a goal was conceded by a team whose skaters were on a long shift

Then, I defined the Fatigue Factor using this formula :

This fatigue factor is a number between -1 and +1 that will tell us what is the cause of the goal at a T threshold. In layman terms, we can calculating which team (scoring or conceding) had the LONGER shifts at the time of goal?

Positive values ( > 0 ) suggest scoring team more often has long shifts, so a “Pressure Goal”. In opposition, negative values ( < 0 ) suggest the conceding team had the longer shift and the goal was a “Fatigue/Breakdown Goal”.

Here, the EV line close to 0 tells us that at even strength, most goals are scored when both teams have a very similar shift age. EV goals often come from transitions, broken plays, rushes, puck luck, or turnovers, so it wouldn’t be accurate to attribute them to fatigue or sustained pressure. On the Power Play, the pattern is much clearer: PP goals are driven by offensive pressure. As shifts get longer, the pressure effect naturally shrinks because long shifts are rare, but it never reverses into defensive fatigue. Short-handed results are noisy, which makes sense: SH shifts are short and SH goals are rare. SH data are volatile due to small samples.

Can we learn something new when we add the periods to the variables we are already looking at? The answer is no. PP pressure is the only important signal, and it’s only in Period 1. Other than that, points and lines gravitate around 0 like flies around outdated food.

Coaches do a good job

To conclude, I finish writing the blog with more appreciation for hockey coaches. They are able to mitigate the impact of fatigue. In other words, fatigue does not create many goals.I looked for a link between fatigue and goals, and the data simply didn’t support it.

If you need to remember 5 things from my quick research:

1. Most NHL goals are scored early in a player’s shift

2. Shift Age difference rarely happens

3. After ~60s, conversion rate rises sharply at EV and PP. But, not a lot of shifts last that long.

4. Fatigue does not cause goals

5. PP shifts are longest; SH, the shortest; EV in the middle

There is plenty of room to expand the analysis. There would be additional value to look at players cumulated fatigue in the game as well as goalie fatigue. It would also be interesting to have much more consideration for the rare nature of goals in hockey. I feel that my analysis did not show enough consideration for that aspect. I also think the reader would have enjoyed more cool facts like a distribution of shifts lengths, team analysis, and other fun stats.

If you are still here, I hope you enjoyed this post. I tried to put some light on a concept that does not get much attention while giving a mix of hockey and statistical insights. This is meant to be a blog post for the nerdiest hockey fans. I started publishing my blogs to Linkedin, so I have colleagues and friends (who don’t watch hockey, but still want to support me), so I did my best to add context and explanations.

You can support me and my work, by giving me a follow on X.com and Bluesky, and by subscribing to the blog. Feel free to reach out if you want an introduction to analytics in hockey, I already helped multiple people for free. I also immensely appreciate feedback. You can also just chat with me, just to chat. I recently had the chance to talk with the creator of TheWinColumn and he was one of the best person I ever talked to. I think there are more really nice people out there, and the next “best person I ever talked” to might be you.

Subscribe now

I will always be grateful to have thousands of people making the choice to listen to my hockey takes on a daily basis. I have a couple projects in the books. Enjoy the time leading to Christmas and happy holidays!

Instead, this is for everyone who ever asked me, “So… what do you do exactly in hockey?” and saw me freeze, trying to find the right words to describe my passion coherently in a language that’s understandable. Trust me, it’s extremely easy for me to lose people in analytics jargon like expected goals, prospect models, transitions, puck battles, distributions, or false positives. If you are still reading this, this is for you: my brother, my sister, my mom, my dad, my friend, my crush, or dear reader, who might want to see that side of me I used to hide because I thought it wasn’t cool enough.

What’s Hockey?

Let’s start at the beginning. Hockey is a sport played on ice with sticks and a rubber puck. Two teams of six players, one goalie, two defensemen, and three forwards, try to score more goals than the other in sixty minutes of pure chaos and fun.

The top league is the NHL, the National Hockey League. It has 32 teams fighting to win the sacred Stanley Cup. The season runs for 82 games, followed by the Stanley Cup Playoffs. The league takes a lot of pride in its parity. Any team can win on any given night. The team with the best player in the world doesn’t always win. It requires teamwork and a little bit of luck.

Luck is one of the elements that makes hockey so unique and fun. In addition to unpredictability, the sport mixes speed, skill, and physicality. Put all those ingredients in a blender, and you get the greatest sport on the planet. Let’s argue later.

Hockey is also a sport defined by transactions: drafts, trades, free-agent signings, and waivers. The best players, though, usually sign long-term contracts with trade protections, so stars don’t move around like they do in other major sports leagues such as the NBA or NFL. Hockey fans still find plenty of fun debating team-building and player evaluations. And when you add math to those two concepts, there’s a lot of fun to be had, at least for me.

The Genesis: Mini-Hockey and Fantasy Guides

The odds that I’d become a big hockey fan were almost zero growing up. I didn’t come from a hockey household. My parents, from Europe and Africa, didn’t hate hockey, they just didn’t care about it. I used to do everything my older brother did, but my love for hockey came out of nowhere.

In primary school, I discovered mini-hockey: little plastic sticks, a plastic ball, and scraped knees on the snow. That was the first time I was exposed to hockey. I wanted to be a goalie. I loved the logos and the gear.

I identified with P.K. Subban because I felt he looked like me. He doesn’t, really, but he is a person of color like me. I also identified with Max Pacioretty because we shared a first name. There was also a player with the same first name as my brother, so the Montreal Canadiens were the perfect team for me. I fell in love with the Canadiens during their miracle playoff run led by Jaroslav Halak in 2009–10.

Then I discovered Fantasy Hockey, a game that involves selecting players and accumulating points based on how they perform in real games. I was obsessed. I used my birthday money to buy fantasy guides, memorized every player’s stats, salary, and jersey number. I just liked looking at the data.

After that came hockey cards, more stats and more obsession, and the NHL video games, especially GM mode, where I could make trades, sign players, and build rosters. I was optimizing lineups before I even knew what data science was. I could spend days on that game. I was also obsessed with building the best hockey roster with my cards, based purely on stats.

All of this happened while I was an awful hockey player who could barely skate. But that didn’t matter. I didn’t need to be on the ice to fall in love with the game. I knew my place was off the ice.

High School: Hockey Blogs and Curiosity

When I got internet access, my world exploded. I started reading every hockey article and rumor site I could find. Eventually, I joined hockey blogs, writing about trade rumors and the Montreal Canadiens. I even spoke on a youth radio station and visited RDS, Québec’s TV sports broadcaster, dreaming of becoming a sports journalist. I used to write my articles on sheets of paper. That’s how deep it ran.

But something felt missing. Everyone could write their opinions about hockey. I wanted to back mine with graphics and stats to make my thoughts more accurate.

That’s when I discovered hockey analytics. I didn’t understand everything, in retrospect probably nothing, but I was fascinated by graphs and charts from people like Sean Tierney, Domenic Galamini Jr., Emmanuel Perry, and Sam Ventura, pioneers in the field who later went on to work in pro sports.

I started using analytics in my writing even if I barely knew what I was doing. I also created a Montreal Canadiens fan group on Facebook that grew to thousands of members.

High school wasn’t easy for me. I wasn’t living the happiest years of my life. Hockey became my refuge. In a sense, when I felt like I did not belong, writing about hockey gave me purpose.

College: Memes, Graphic Design, and Coding

In college, I got strategic. I wanted to run my own hockey blog with a built-in audience, so I bought, using money from selling my Facebook group, a dormant page called Montreal Canadiens Memes. I immediately caused chaos by announcing it would become my blog platform.

Fans revolted and started unliking the page in large numbers. So I backtracked, downloaded Photoshop, and started actually making memes to save my investment.

Surprisingly, it worked. The Instagram account grew from 100 to over 15,000 followers. I learned social media growth strategies, brand partnerships, and graphic design. Eventually, I got bored of memes and started making hockey posters instead. It became an obsession: I wanted to become a graphic designer.

Later, I discovered I could generate graphics automatically using code, perfect for things like pre-game rosters and standings. I wanted to know how. So I downloaded the most popular programming language: Python.

And then everything changed. I realized I could analyze hockey data with code. That was it. I was hooked.

Even though I was studying accounting, not math, I spent my nights learning programming, scraping data, and building my first hockey models. My grades suffered, but I didn’t care. During the pandemic, creating hockey graphs became therapy, a way to stay grounded when everything else was uncertain.

What About Now?

Fast forward to today. I finally got to work in professional hockey for the first time. I’ve collaborated with incredibly smart people, built models I once only dreamed of, and learned more than I thought possible. I also discovered how much I still have to learn.

I’m still coding almost every day. Still writing when inspiration strikes. Still growing. Still that same kid who fell in love with the game through mini sticks, fantasy guides, a miracle playoff run, player cards, and a video game.

I’ve started mentoring university students who want to get into hockey analytics, helping them navigate the same messy, nonlinear path I took.

And even though I think about hockey all the time, I’ve tried to find balance, like learning Mandarin. I can say bǐsài 比赛 for “game” and bīngqiú 冰球 for “hockey.” One day I’ll talk about hockey analytics in Mandarin too. I promise.

Right now, I’m working on a few big solo projects. Not for a job in the NHL, just to answer my own questions about the game. Because at the end of the day, that’s what it’s all about for me: curiosity, creativity, and love.

Final Thoughts

If I had to explain my passion for hockey analytics to my crush, I’d still tell her it’s a long story. I didn’t fall in love by playing hockey. I fell in love by drawing it, writing it, counting it, isolating it, projecting it, and living it. I followed my heart and let it guide me, and so far, it’s never failed.

This article is a collab between and ALEF.

Noah Dobson for 16th, 17th overall picks and Emil Heineman

With the Noah Dobson trade, Kent Hughes made a bold move and also managed to partake in an annual tradition, trading away first-round draft picks (twice the fun this year!). Let’s look at what this means for the Habs.

Review of the 2024-2025 defensemen corps

First, let’s look at the minutes and age distribution, taken from HockeyViz.com. Martin St. Louis split the heavy minutes between 4 main players, three of them being left-handed. The third pair was filled with David Savard and either Struble or Xhekaj. In terms of age, both Matheson and Carrier don’t quite fit within the age window of the main stars.

As for quality, it’s possible to have an idea of the caliber of the players thanks to a data dump from Dom Luszczyszyn of The Athletic. I tried to replicate a visualization used by Micah McCurdy (HockeyViz), who identifies D1 to D6 based on his synthetic goals (sG) calculations. I made a similar visualization using the Net Ratings published by Dom Luszczyszyn. The chart below presents the quality of the Habs D since the new administration took over (Gorton, Hughes, St. Louis). Since some values were stacked on each other, I added some jitters when it was time to write the name. I hope it makes it a bit clearer. As we can see, the Habs struck gold when they drafted Huston, seemingly finding a bona fide #1 D, a young one at that, in the late 2nd round. Besides Huston, Struble, (in limited minutes) Guhle, and Carrier played at what could be described as NHL level. Matheson went from back-to-back top-3 quality seasons to slightly below NHL level. Maybe he could bounce back in a limited role and with a full summer to heal injuries.

What are the Habs getting in Noah Dobson

A lot has already been written about Dobson, and a lot has been said about how his game stepped back this season. He did go from a career high of 70 points to 39 points this season. Still, per hockeyviz and Dom Luszczyszyn, he seemed to still play at a D1 quality. He is also big and right-handed, two things that are lacking in MTL right now. For the 2024-2025 season, he contributed similarly on offense and on defense. While it would make sense to match him with Guhle on a top pair, he could cross-shift with Hutson if needed. He also provides a luxury back-up option on the power play if Hutson gets injured. Dobson could still be improving, as 90% of players peak between the ages of 23 and 291 and his net rating is on a general upward trajectory.

For people who still assess defenseman with points, don’t expect Dobson to reach 70 points again. With Huston holding the keys of the power-play, it is difficult to imagine that Dobson will get the same opportunities on the man advantage he used to have with the Islanders. Dobson is extremely good when moving the puck from the defensive zone to the offensive zone. His skating and his elite passing skill allow him to be a threat when carrying the puck. We can’t call him a transition god because although he excels at carrying the puck, he struggles a ton at defending off the rush. It is one of the few areas where he needs to put up some work. When defending, he would rather use stick-checks and small pressures than big hits and box-outs like most defensemen with his frame. Most models have him as a player with good isolated defensive impact, his defense (coachable skill) will probably be subjected to adjustment under his new coach.

Dobson arrives in Montreal with a fat contract. It is now common for first pair offensive defenseman to sign at an AAV as high a 9mil. We saw and we will see more of this (Hutson 👀). During a contract, the AAV stays the same but, the impact on the salary cap (what is really important – also known as Cap%) changes. So when you assess a contract, it is important to take context into consideration and given the context, 9.5M$ might look insanely high, but in a reality where everyone know that the salary cap will not stay flat (like it once was), it is a much more reasonable deal.

Conclusion

The addition of Dobson looks like a slam dunk for the Habs organization. He fits in terms of position and profile for a long time. Hockey is a strong-link game2 (team success is mainly influenced by the performances of the team’s best players), and players like Dobson are difficult to find in the draft or free agency. The Canadiens took a calculated risk in their rebuild by trading away 2 first fround picks who might have not aligned with their contention window. Losing Heineman can feel dissapointing although he graded out as an 11th forward per net ratings (maybe due to bad luck with injuries). Despite trading him away, he’s the type of player a successful organization should not have difficulty replacing.

OT - What is a draft pick worth

To better compare the deal, we could also wonder how valuable the 16th and 17th picks are. By cross-referencing the Net Ratings of players to their overall position and applying a curve fitting method (y = a*e^-bx+c), we can see that on average, these picks are not necessarily that valuable, although the variance is really high (it is more difficult to make confident predictions because the outcomes are all over the place). Their value seems to be maxing out at close to 20 ratings, but when making such trade, it’s fair for an organization to make the safe bet that the player they get in return might be better than prospects available with those picks. We can remember when the Canadiens traded 2 picks for Andrew Shaw and one of those two selections turned to be elite goal scorer Alex DeBrincat. On the other end, the Oilers once traded a lambda 2nd round picks for Brett Kulak… and that late 2nd round pick turned out to be one of the most entertaining young players in hockey : Lane Hutson. Betting against the average is risk that can be worth taking.

Zachary Bolduc for Logan Mailloux

Logan Mailloux has all the making of a top D-man. He’s big, with a big shot, and fairly physical. If you ignore the controversy, maybe he would have more people rooting for him. He’s a two time AHL all-star, but has not broken through the NHL. Having watched him play live, he doesn’t look like the fastest skater, but I will refrain from analysing him. He doesn’t appear in the Net Rating database, and he would only have a small sample size anyway. I assume that the Canadiens know well what they had in him and yet, they have tried to trade him for over a year. They didn’t think he was ready to play in the NHL last year despite the fact that the D-core was pretty weak. In return for him, the Habs have somehow been able to get Zachary Bolduc from St. Louis. From the same draft class as Mailloux, Bolduc has already played 97 NHL games and was good for 19 goals and 36 points last year. Just on the surface level, the trade seems like a no brainer. It also makes sense from a personnel perspective. With Dobson, Carrier and Reinbacher, it did not look like Mailloux would find a role in the short to medium term, but there are many holes on the offense for Bolduc to fill, even without many unrestricted free agents leaving. As you can see below, Bolduc played well this season, and he’s a true NHL caliber player since entering the league, with what one would assume is still some room to grow. In terms of Net Ratings, as you can see below, he played as a fringe first line player last year. Without going through all the Habs depth chart, this would rank third last year, in front of Slafkovsky. He also seems to play a strong two way game, with a very good defensive rating, which could fit well with Demidov, who still has to improve on that end. The fit between the two young wingers could be a little problematic as they both prefer to play on the right side. On the Blues end, they feel like they felt like they had a surplus of forwards and that Mailloux, a right handed defenseman, would bolster their defense. We are not a Blues blog, so we will stop there talking about the St. Louis.

Bolduc has found a lot of success on the 1st unit of PP with STL. He was so good in the bumper spot that the team completely changed their strategy to run the man advange through him. His shot, his willingness to go in front of the net, his ability to get lost in coverage and his quick hands in tight have allowed him to be one of the best goal scorers of the league after the Four Nations tournament. He does not look extremely fast, but he is a born goal scorer and he strive through physical play. His style will turn him into a fan favourite in no time. In terms of expectations, it is difficult to predict if he will get time on the first unit of man advantage (that generally skyrockets point production), but he is expected to see a rise in time on ice at even strength. Bolduc should get a lot of opportunities to showcase his talent on the 2nd and 3rd lines.

Before the season, on one of my multiple burner accounts, I posted my official 10 predictions about the Montreal Canadiens.

In this week’s blog, we look over them to see how good I was at predicting the future.

He will indeed finish the season with at least a point per game, thanks to another multi-point performance.

Debatable. He was again heavily tested this season. Goalies are difficult to evaluate. Goalies are generally the position that have the most impact on the game and they also have the most inconsistent workload. Fowards’ main metric is point, and the average shooting percentage in the league is around 10%. So, I will generalize that forwards can be bad more than goalies because they will mainly get judged for the 10% of shots that they can convert into goals. In opposition, a goalie is expected to save 90% of shots he faces. The margin for error for a goalie is a lot slimmer because he will be judged on 100% of shots he faces because of the repercussions on the game. With all that in mind, there was a point where Montembeault was definitely having his worst season since he was acquired by the Canadiens and people started questionning his fatigue and workload, but lately, he has been key in the late playoff push by the team. I don’t know if it’s a successful prediction. Let me know in the replies.

I think the Kirby Dach experience at the role of 2nd center is over. Sadly, it ended with an injury. Montreal is more likely than not to try to acquire someone for that role. Alex Newhook and Christian Dvorak have played really good hockey in the 2nd half of the season, but I doubt that it is enough to justify giving the 2C to either of them.

A Guhle injury saved his career. He was 7th D at the time and Guhle’s injury gave him the opportunity of a lifetime to play on the 2nd pair with Lane Hutson. That pair was one of the best pairs the Montreal Canadiens have iced this season. Combining grit, toughness and skill. Struble has earned a permanent role on the Habs D-core ever since. No, he did not get traded.

He might run both PP units eventually. Good prediction.

It’s only a question of time until it happens. Probably the next in line.

He has 19 goals now. If he ends up with 22 goals, I officially have superpowers.

This did not happen at all. He is still a really bad defender that makes way too many mistakes. And he does not have the offense that can make up for his terrible defense. He relies on the idea that “teams need teams need a tough guy” to stay in the league.

Did not happen.

It would have been such a good prediction if I did not add that 2nd half to the sentence.

I literally don’t know how he never got recalled by the Canadiens this season. He scored a lot of goals in the AHL this season and he is one of the toughest AHL players. I feel that he would have brought some element to the game that Habs 4th line was lacking. Bad prediction.

This is probably one of my worst blogs ever. I still hope you enjoyed that light read on this Sunday morning.

Have a good week and enjoy the final days of the NHL regular season!

Life is sometimes funny.

I recently got my biggest burst in followers ever without even being particularly active on Twitter — no way I call it X — or on the blog itself. The recipe of that random rise in subscribers on my blog? The Power of Friendship. In fact, my good friend Gabriel Foley has been much active than me and his blog recommends mine. I will embed his blog so you can see the work of my amazing friend.

Foley on Hockey

Follow along as Gabriel Foley covers a variety of topics centered around NHL and USA Hockey.

By Gabriel Foley

The Power of Friendship is a popular theme in hockey lately, especially on the Canadiens. Friendships have always been part of hockey. As fans, we always dreamed that our favourite players are the best friends in the entire world even if it wasn’t always the case. From the Subban-Weber friendship — materialized with The Triple-Low-5, to the pretty Mint Kotkaniemi-Mete friendship to whoever Cole Caufield hangs out with now, Montreal always has fun friendships. Players on the team seem to attribute a big part of their current success to the bond they share on and off the ice.

There is a new friendship in town

There is a new friend in the group this season. Elite goalscorer Patrik Laine was acquire with a 2nd round pick last summer for depth defenseman Jordan Harris. What was the trap? In a context where the Salary Cap is the most important asset in the hockey market, the forward would occupy 8.7 millions $ on Montreal’s cap for each of the 2 seasons left in his contract. It represents ~10% for this season. Laine came in to help a team that is popular for having great goaltending, and not so good goal scoring power. And despite a freak injury that sidelined him for the first part of the season, his came in like a rock star. 8 goals in his first 9 games with the team. It felt like everything he touched turned into goals. What makes Laine special is his ability to score goals even when the goalie is not screened and in place to face the shot. His release has a ton of power and is extremely accurate. Now, he averages a little under a goal every 2 games — which is really good , and he has diversified his game and became a more dangerous offensive threat. We have seen him more recently create space space for his teammates at a higher frequency. The results speak for themselves : Alex Newhook (his linemate) has been one of the best Hab in the last month or so.

Why I stopped paying for Hockey Analytics subscriptions

I actually stopped paying for hockey analytics subscriptions this season. The reasons are simple :

1. I didn’t use them unless I had a point to prove on socials — and I spend significantly less time on socials since I graduated and I have a job - congrats to me — or an article to write — I don’t write much .

2. I want to build what I use because I have more control on it - and i can come up with the things you will see in the next few lines and articles.

3. I will spend more time understanding the game of Hockey through reading, videos, etc. I actually got a chance to talk with hockey consultant/coach — +many other hats — Jack Han after reading his book, and I feel better writing about hockey and watching hockey after picking up some of his thoughts.

I still encourage you to support creators if it is something that you consume a lot. There is a lot of work behind the scenes and lots of great content. I will probably get back to subscriptions once I achieved my current goals. Lots of good content out there!

The Power Play looks better with Laine on it

From my viewings, I understood that Montreal’s Power-Play much more dangerous since Patrik Laine came back from his injury. Every shot he makes is extremely dangerous. And Lane Hutson can count on his new friend Patrik Laine to help him get easy assists for his Calder run. I can also agree with the idea that Patrik Laine offers him the assists he would get if his linemates had at least average finishing skills. I feel that any Laine shot can go in. He is the first hab to make me feel like that since Shea Weber. It turns out that I am not the only one to feel like that. Many Habs opponents have adjusted their penalty-kill tactics to prevent shots from the Finnish sniper. They would rather play a 4v3 with a man always on Laine than a normal 5v4. So, for this first article in months, I wanted to go on a positive note — because I was often criticized for being negative… I am just honnest to be fair — and quantify how better Habs PP is with Patrik Laine… and I was not ready for the results in fact, I was, but I thought I would be proven wrong.

Show the Numbers!

Considering the sample size is similar, I thought it was fair to compare how the Canadiens did on the PP in games where Laine was playing against how they did in games where he did not play.

Quite surprising! A 24% efficiency would rank them among the most successful man-advantages while a 17% would rank them among the worst. Now, they have an efficiency of 21% which ranks them in the middle of the NHL. Not good, not bad either. Is Laine the problem? I don’t think so.

The Weber Theory (You Heard It Here First)

I think that what might explain this surprising phenomenon is the the Weber theory. Theory that I just created that suggests that teams with lack of talent upfront will rely on one good shooter on PP — which makes the Power Play worse because when he doesn’t shoot, the rest of the team almost forgets how to play. It happened when Shea Weber was feeding Webombs for the Montreal Canadiens. A good power play is one that makes the defense move and that is unpredictable. Laine takes 40% of the shot attempts when he is on the ice on PP.

When is is not on the ice on PP, the Canadiens like to distribute the puck a little bit more. Cole Caufield leads the team with 13% of shot attempts and is followed by Alex Newhook, Nick Suzuki and Mike Matheson just over 10%.

Laine Is Worth It

Laine deserves the right to shoot a ton on the man advantage. He scored over 31% of the Montreal Canadiens PP goals this season despite missing half of the season. He built real estate on the left flank and pushed Cole Caufield outside of his former home on the PP. I think he made the Canadiens’ man advantage look entertaining by himself with a little help from Hutson.

I think Montreal should randomly rotate Caufield and Laine the left flank to create confusion for the defending team. Call me Coach Max now. But, I don’t worry too much about the 2024-25 edition of the Canadiens’ PP since it is expected to change drastically as they welcome the russian prodigy Ivan Demidov next season.

Sorry if I made mistakes : I rushed a little bit to make sure I release the article tonight. Sorry to professional data providers if my data is not 100% accurate : I will blame the data engineer — myself. I still hope you enjoyed reading this piece. The work behind it is actually crazy : building a data scraper for months, finding out that it has mistakes as I am writing the article, building temporary fixes, temporary fixes crashing, giving up on the last part of my article because of it. Writing blogs is always an adventure full of unexpected traps.

Before I leave you I have a couple of things to tell you :

1. I might reach 100 followers on my blog — I might have reached that milestone already. I just want to say thank you from the bottom of my heart. I really appreciate you.

2. I like to talk about hockey with people; My DMs are always open to disucss hockey with a fellow hockey enjoyer.

3. Let me know if you have a burning question that you want me to address in my next blog : I will be happy to find the answers to YOUR questions

4. I developped a really fun hobby lately : learning languages. I am currently learning the 2nd most spoken language in the world right now. I started over a year ago. I suggest you to learn a language if you want to : it is a massive challenge, but really rewarding. And if you speak Mandarin Chinese, feel free to reach out to me. My other favorite programming language is Python. Please laugh.

   

My Montreal Canadiens will be in the Playoffs in a couple of weeks. On that note, I wish you a great last stretch of the regular season! I wish your favorite team to qualify for the Stanley Cup Playoffs if they don’t fight for the 2nd Wildcard spot in the Eastern Conference. The snow is melting, so we should also expect our favourite public scouts to come out of hiding with their super knowledge about prospects. But as you already know, the best takes on prospects and on hockey in general will come from… my mouth. So feel free to Subscibe to my blog and follow me on Twitter if you want to look smarter than everyone in your hockey groups.

Subscribe now

See you later alligator 🐊

NOTE : I have been writing this article for a week now. Some events in hockey happened since I started and I did not want to rewrite everything to account for things happening in real time, so just be aware of that.

Estimating the Value of a Draft pick has been a goal of mine for a long time. I once built a rudimental model to estimate the money retention based on the draft pick thrown in in the trade and despite how simple the model was, it worked really well. At the time, I had the value of a first-round pick at around $3.6M. So a team would need to add pay a 1st round pick to convince a team to retain $3.6M in salary for a player in a trade.

Even if I said I would not write about the headliners of the next Draft, I will still give you an idea of what is the current temperature among the public scouts. James Hagens seem to be the unanimous top prospect for now, followed by Porter Martone. Anton Frondell is the first European player on the list. Matthew Schaefer seem to be the unanimous top defenseman in the draft class. Sources told me to not be surprised if he moves inside the top-3 because he is a special player. The greatest name of the draft Roger McQueen is #6 and our top Russian player is Ivan Ryabkin. Quoi de mieux qu’un joueur québécois en Caleb Desnoyers pour compléter le top-10 du consensus! Major shoutout to EliteProspects.com for letting me use their data. You already know, I work with the best resources in hockey to provide you with some digestible hockey research.

Don’t ask me about the strength of the Draft. I can’t tell you right now. It is too early. I can still say that a generational talent is coming in 2026. 2025 is not the year your team will get the next Connor McDavid unfortunately. It is also way too early to run NHLe models, so I would advise being careful with some of the things you see online. I will try to have the first version of a variety of NHL Draft tools ready for New Years.

Assessing the value of a draft pick is an important task. Why? Because they are traded in almost all trades.

People who are smarter than me are doing a great job at building tools to assess NHL players. People who are smarter than me have also done a great job at finding the value of a pick, I can’t lie here. They also used much more sophisticated analysis than what I will show you today. But what I will show you here is simply a simple walkthrough of data exploration mixing NHL Salary and Draft Data.

Using the salary of an NHL Player is an intuitive way of evaluating players. It is fair to say “The better a player is, the higher his salary should be” as a general rule. It is flawed for multiple reasons, but for the sake of simplicity, let’s say this is true. The NHL has a hard salary cap (unless you are the Vegas Golden Knights apparently - jokes) and it is changing based on what the team’s owners decide (I think). So using the number value of salary is not the best way to see a player’s contract in a salary cap world because its real impact is actually the % of the salary cap the contract is taking. So from now on, these eyes will not be blinded by the lights we will use Cap% as variable instead of player average annual salary.

In finance, the value of a project is the sum of all the cashflows you will get out of it. So, in my simple analysis, I say that the value of a player is the sum of all Cap% he will make in his career. Where we compare that value with the overall pick a player was selected at, it gives an interesting graph. I found that using a logarithmic regression curve was good enough to fit the relationship.

Using R^2, we might think that there is no relationship, but if we remove “busts” (players that don’t become NHLers), the relationship between the cumulated Cap% and the draft rank becomes interesting. It is fair to assume that your top picks are going to be good regular contributors in the league and they should get paid a decent amount while you should not expect much from the later picks.

When we compare forwards and defensemen, we find 2 curves that are really similar, but Fs value is slightly higher in the first round.

When we add the age in the analysis, things get really interesting : it confirms the obvious a fact that the highest-paid players are generally the players who get drafted earlier.

If the previous chart was not easy to read, I made one using Round Pick number as colour variable. Again, we find out that players drafted earlier are making more money than players drafted later.

To conclude, I want you to leave the reading with 3 concepts :

1. It is really important to assess the value of picks because picks are involved in most hockey trades. If you want to make a trade (because you are a hockey team) or you want to assess a trade (because you are a media or a fan or just yourself), you NEED to find a way to evaluate for the value of a pick.

2. A Logarithmic Regression fits the relationship between a pick and its value really well and it is simple to understand. In good English, as we go further in the draft, the difference of value between a pick and the following pick keeps getting smaller.

3. This one will be a game changer in hockey : the players selected higher are generally more likely to make more money in their career, unless their name is Joe Pavelski.

I think it is also important to remember we are dealing with human beings when we talk about prospects. It is often easy to talk about them like they are just a playerId in our database, but even I, often forget that the players I am writing about are people that have their own feelings and their job is consistently monitored by the public that does not hesitate to tell them when they are bad at their job (sometimes wrongly). I will use that small break to wish a great season to all the players, to the personnel on teams, to media contributors (who work like their life depends on it to create content for the growing fanbase), and most importantly to my old friends at La Victoire de Montréal. I wish you a lot of Victoires and I hope to catch one of your games this season! This is the moment you find out my title was clickbait because I did not tell you what was the exact value of a draft pick. This article is the appetizer of a whole meal. This is the first taste of a series of articles where we explore fields that did not have much attention from the public sphere. It is also a learning journey for me.

Let’s cook 🧑‍🍳!

Good for us, the NHL did not make it too difficult to access. In about a minute or two, I was able to fetch the data in a Colab Notebook.

Before we go further, I am not a pro or an academic, so the steps I will follow might be questionable or bad or good if I am lucky. My first step is to find a video reference for a goal that I like a lot and that does not bother me to watch a multiple times. So I picked Cole Caufield’s PP goal against the Leafs.

Second step is to quickly replicate it and draw early conclusions.

Here is what I notice :

• the X/Y scale is different from what we usually have in PBP (x = [-100,100] ; y = [-42.5, 42.5]). A rough estimate would be something like (x = [0,2222.22] ; y = [0, 1000]).

• For some obscure reasons, my plot horizontally mirrored. Something to look for eventually.

• It is really easy to identify the puck (1) in the data. For the other ids, I still need to find out if the numbers mean something or if it does not matter.

  

  
  I wish I had more to tell, but I need to go sleep. Let me know if you find something interesting. See you soon for more good NHL data news!

At the end of the season, when the NHL changed a bunch of things in their API, I also fixed mine and I took the time to the regular season data in case I needed it over the summer. After spending weeks on prospects work, I did not think the day where I do something with my NHL data would arrive, but it actually arrived today. I woke up in the middle of the night over the weekend thinking about RAPM and the voices in my head told me to build that RAPM model. And this is what I did today.

A little bit of history before we go further

RAPM is the evolution of the APM stat that comes from Basketball. It was created by Jeff Sagarin and Wayne Winston about 20 years ago because they thought that raw +/- was not enough to predict how good would be a player. So the APM, was created to estimate how good a player would be when a coach plays him on court. With a lot of data, it was widely considered as the best metric for player evaluation. The RAPM uses the Bayesian method Ridge Regression (also called L2 regularization) to give even more predictive power when it comes to smaller sample sizes. It comes with many criticisms. The famous @yolo_pinyato came up with this chart that summarizes the weaknesses of famous metrics.

RAPM was meant to be predictive, but I am not sure it really is. A simple excercise to see visualize the predictiveness of RAPM is to look at how it varies for a player from a season to another.

It is a really rudimental expriment that comes with its load of “this does not tell the whole story”, but it is a really easy-to-understand way of showing the predictiveness of RAPM for different metrics. I took the RAPM data available on Evolving-Hockey for the past 7 seasons for this test. It turns out that despite the fact that APM (the ancestor of RAPM) was created to predict how good a player would be, it turns out that we can’t fully rely on it to assume how good a player will be next season. The real power of RAPM comes from RAPM C±/60 (Corsi - Shot attempts) that hold significant predictive power. Even if Corsi is being more and more ignored at the expense of more complex metrics such as expected goals, it is important to note that when it comes to predictions, you might get better results with the RAPM C±/60. RAPM does a great job at explaining what happened in a context relatively complex. I think it fair to say that it is the best metric available when it comes to player evaluation.

RAPM is a really important metric in hockey analysis. It serves as the egg and the milk for the cake metric that the WAR/GAR (Win Above Replacement / Goal Above Replacement) is.

So what?

I built my first RAPM model using as target variable C± (on-ice shot attempts differential). The results are actually interesting.

Analytics darling Auston Matthews comes on top of the fowards in 2024 at 5v5 rankings using my RAPM model. Matthews has played with really talented players this season, so there is a chance that the model has attributed him too much credit for the offense created when he was on the ice. I don’t think it happened, but I think it is a valid question to ask when we look at model outputs. HockeyViz does a great job at isolating coaching, teammates, and other external effects on players performances. I think it is important to keep in mind that all those factors impact the model and even our own interpretation of what we see when we watch hockey. Everyone can watch hockey, but it comes to dissect the information we captured during our viewing session, it is really when the fun starts. Hockey is so complex that we will probably never come to a concensus and models do a good job at keeping us away from any hope of concensus when we think we are close to achieve that impossible task. Another element we can notice is the presence of role players on the chart. We can just assume that they did really good when they were deployed. So they were just great role players last season. They are not better than Connor McDavid, don’t worry.

As for the defensemen, my model tends to agree with with the Norris voting. Quinn Hughes seem to be an excellent choice for the trophy awarded to the defenseman of the year. Again, few surprises, but as I explained earlier, it is explainable. I also think that Bouchard was probably much better offensively, but offense was credited to his teammates. It is something to keep in mind like I said.

When I was much more immature, I used charts to dunk on players, teams, and hockey people in general. Growing in the field of data science, I realized I was doing a disservice to the profession of data scientist and to the creators of models by acting so. So I am not going to post the worst players, simply because I don’t want people with bad intentions to use my work in a way that I did not meant to.

Few players who caught my attention

I was one of his biggest critics (to not say hater), but he had a strong season and really established himself as a good 3rd pair NHLer this season. He outgrew the Sherif role that was probably easy to settle into and became a player who could provide offense and defense when he was tasked to. He should deserve every penny of the contract he will get this summer.

Despite the model probably underrating his offense, we can happily see that Selke McDavid has arrived. I think that if the model correctly rates his offense, he is probably in the 100th percentile with Matthews in terms of impacts.

Connor Bedard would’ve broke the league in pieces if he came in 10 years earlier. Unfortunately, the league was ready for him and players like him can’t tear up entire defenses by themselves at the age of 18 without a good supporting cast. Good for him, the Blackhawks spent a lot of money this summer to improve the team.

The Capitals did not only get CapFriendly over the summer. They kickstarted the hockey summer with the blockbuster trade that sent their backup goalie to Los Angeles in exchage for the infamous Pierre-Luc Dubois. Dubois has changed teams mutliple times in his young career and after signing a massive contract with the Kings, he stuggled to perform up as well as you want a player you just signed to a $8M per year contract to do. In Washington, PLD wants to prove that his 2023-24 season was only fluke. I think he is overhated because most of the times, “overpaid” players are generally negative players, but unlike them, Dubois actually provides positive value even if it is not 8M$ value. He can prove the haters wrong on that completely reshaped Capitals team, and most importantly he can try to help Alex Ovechkin to catch Wayne Gretzky.

Finally, let’s talk about the biggest free agent of the summer : Jake Guentzel. The Tampa Bay Lightning made the biggest splash of the summer when they sacrificed their captain Steven Stamkos (and Mikhail Sergachev) to sign Jake Guentzel. Guentzel is a former underrated cheap Crosby-merchant (just like a lot of others) who learned so much from playing with Crosby that he outgrew his Crosby-merchant role to become an elite player himself. Now, he is one of the best offensive players in the league and he brings some defensive value. Was it worth it to throw the captain away to sign the big fish? Only time will tell, but so far, it looks like a home-run move that you must do if you if you have a chance to.

On that note, I hope you enjoyed the article. Feel free to ask me in private messages if you want the card for other players. Feel free to correct me if I made some mistakes. And enjoy the summer.

For nerds like me, it is the opportunity to show the product of weeks and months of work. This one Draft is special. I got to work on TheDraftDigest with Chace McCallum and Josh Khalfin for the past 2 years. Those 2 people are some of the smartest people I have ever talked to and I feel really lucky to had the chance to meet them and build things with them. For those who don’t know about, TheDraftDigest is a tool using Data Science method to assess hockey prospects. Unfortunately, we had to shutdown the Shiny App we created… but it was for a good reason : Chace made it to the NHL. I don’t think there is someone more deserving than him.

So I gave myself the challenge of re-building the website. I can’t say I succeeded because I don’t have a fully working website ready for people to use, but I don’t arrive empty-handed. I built a variety of tools to support prospect assessments.

League strength

This one screenshot of the strength of leagues shows an iteration of my estimation of the strength of the best 15 leagues in hockey. I think the lates To come up with those numbers, I used a method inspired by Katerina Wu (now Pittsburgh Penguins) in her work “Which League is the Best?” that she presented at the CBJHAC2018. She used transition matrix and difference in points-per-game for players changing leagues within a season or in the following the season. Then, she used a OLS model to estimate league strength. To see the league strength per year, I just iterated the process per season.

NHLe has its limitations, but I think it is a decent indicator of league strength.

If we forget about the NHL and the fact that all league seems to be the same strength and we use a different scale, we see a different story. Chace was the first person I know to notice that the KHL was losing its power in term of 2nd best league in the world while the AHL should be taken more seriously as a difficult leagues. All of the iterations of my model have shown this trend. Unfortunately, this viz does not have 2024, but I can confirm that no league is challenging the NHL yet.

Player ranking tools

When I am writing those lines, there is a big feeling that Zeev Buium should be the #3 pick. EliteProspects’ API comes with an aggregator for public draft rankings. It comes handy when we want to see the level of consensus for the expected draft rank of a player. For non-math people, a standard deviation close to 0 means that most of the people agree on a draft rank while a higher standard deviation means that a player’s rank is more subject to debates.

Who is the best defenseman of the draft? Apparently the draft is stacked with promising defensemen that should have their named called early. I arbitrarily selected 5 Ds that top the 2024 class and with this tool I could compare the distribution of draft rank. We also have an idea of some statistical informations about their ranking distribution. Comparing distribution helps us to see trends. We notice here that Levshunov seems to be leading the way while Silayev is the object of many debates in scouting rooms.

Player cards

I think the most popular tools are the NHL and star model. They are also the ones I worked the least amount of time on. In 2 days, I built a tools that uses 80 variables to estimate the likelihood of a player becoming a NHLer or a star. To become a NHLer you need to have played at least 300 games in the NHL. My star % will be based on the odds that a Forward has 5 seasons in the 92.5th percentile or higher in PPG (so about 36 stars per season) or a Defenseman has 4 seasons in the 95th percentile or higher in PPG (for about 15 stars per season). So my datasets had the problem of being severely imbalanced. I also did not have the time to test other models and run parameter optimization (which could have helped to get better results). I also wished I took more time to run PCA on models. For the models, I am using the random forests algorithm. It is a popular classification algorithm for such tasks. “All models are bad (except for the one Chace built last year).” My model is bad. Really bad. So, don’t be surprised if it returns weird results.

Philosophizing the Draftology

My model has not been exposed to the same level of information that every single one of you got. It is really easy to say “Celebrini is the best prospect in the Draft” even if you have not watched him because throughout the year (and maybe even last year), you were exposed to some kind of consensus that Celebrini was the best player of the Draft. My model’s only variables are box score stats (points, goals, assists, +/-, PIM, all those stats relative to team and league, NHLe), handiness, age, and position. I think such models can be really useful (I am obviously biased, this is my bread and butter) when being made properly and used in the right conditions.

I think it is important to take a step back before using models and make conclusions using them. Some questions deserve to be answered before you start yapping around early conclusions.

• Who made the model?

• What algorithm was used?

• What are the inputs?

• What are the outputs?

• Strengths of the model?

• Weaknesses? Future improvements?

• Why it was built?

• How you evaluate the model?

• In what context to use and to not use it?

On my end, I think my points equivalency method is good. Unfortunately, my star cutout might be too exclusive which makes the dataset too imbalanced and prone to bad results for the model. I also think my method is too unstable : a player NHLer/Star odds switch too much from a year to another. I also started to do that, but I want to reward more my model for finding stars. When you train such a model on a such imbalanced dataset, it is easy for the model to just say that everyone is a bust and get good scores at the end. The model think it is doing great because it will perform well in terms of “I am right more often than not”. But it is not really what you want. Finally, I think my workflow and the skeleton of my project is much better than it used to be which is promising for what I want to release next year with more time and more knowledge and experience.

When I started to use model, I was a young immature teen who did not really understand the math behind. It was fun to dunk on teams and players for “being bad” or making “horrible moves” because “the model said so”. With more experience behind my belt and more maturity, I understand why some hate data analytics in sports. I hate the person I was when I was making fun of players for having red spots in their players cards. Being in position to make real recommandations to real teams really thought me to things into perspective. We deal with real human beings and our recommandation lead to real moves. So what you do, what you say and how you say it has much more magnitude than how you can make tweet in a random afternoon. Data Analytics is really important in sports and I think the scientific process of approaching problems is the best way to solve business problems. But, unless you work for a data science product, the data science is only there to enhance the infrastructure that already exists. So do not think that all teams with a lot of analytics hire will only trade for players with blue player cards and will get rid of players with red players cards.

What is coming in the next couple of years will be massive. I think we are only seeing the tip of the iceberg in terms of how data science can have an impact on the draft and I am really excited to see how it will evolve. I am disappointed to not have a website ready for you today, don’t count me out yet : I am motivated to come back next year with an even better model, a website and more tools available for the public.

Have a good draft!

The Chosen One

Last year, for his first NHL Draft, Kent Hughes ignored the player that was the most popular among the fans to select a late riser from Europe in the Slovakian winger Juraj Slafkovsky. Same story this season. Despite the loud appreciation of the fans for the Russian prospect Matvei Michkov, Hughes went off the boards by choosing the Austrian defenceman David Reinbacher. I think this article by my friend Thibaud Chatel has a lot of information on the new Hab, so I encourage you to read it. Thibaud has been tracking data from the NL (Switzerland) for a couple of times now and he might be the best person to talk about Reinbacher.

Thibaud Chatel’s Newsletter

What Montreal sees in David Reinbacher

You know, we, NLIceData are a company focused on the Swiss National League. I, Thibaud, live in Montreal. The Canadiens just drafted David Reinbacher with the 5th overall pick in the 2023 NHL Entry Draft. This is a perfect world. At NLIceData, we manually track a large number of National League games for our subscribers and the teams we are consulting for…

Read more

3 years ago · 6 likes · 3 comments · Thibaud Chatel

As I said earlier, Reinbacher is a late riser. Up until the 2021-22 season, he was on no one’s radar for the first round of the NHL Draft. But like Thibaud explain in his article, a series of fortunate circumstances lead the young Austrian to face challenges and overachieve himself overcoming those challenges. That’s how he went from a projected 3rd round pick to the #1 defenceman on his team. It is impressive considering the fact he plays in one of the top leagues in the world at only 17 years old.

Reinbacher is mobile, shoot-first defenceman who produced a lot of points for his age and his position. When I say he is a shoot-first player, trust me he is. He is not a puck-moving defenceman. When he has the puck in the offensive zone, he usually makes “the safe play” by blasting the puck forward hoping it goes in or a teammate picks up a rebound. Despite a surprising point total, he is not an offensive defenceman. He fits the mould of your Habs defenceman who can skate, shoot, do good defensively and hopefully play a role on a second unit of power play. His good performances in one of the best leagues in the world make him one of the best prospects of the 2023 NHL Draft. The Canadiens hope they can build off his solid start in a professional league to make him a top-pair defenceman in a near future.

What was once a huge strength of Montréal, the right side of the defence (with Subban, Weber, Petry), became a huge weakness after the captain Shea Weber was force to stop playing hockey due to injuries, and that Jeff Petry was traded for a lefty (Mike Matheson) … leaving David Savard as best righty (D) on the team. Before Marc Bergevin got fired, he tried to address this need by selecting Logan Mailloux in 2021. A couple of months later, Kent Hughes was trading Artruri Lehkonen for a promising RHD in Justin Barron. The handedness of Reinbacher made him a valuable prospect for the Canadiens organization.

I worked on TheDraftDigest, a project by Chace McCallum, Josh Khalfin and I where we basically built our own prospecting tools using EliteProspects.com data. Our model suggested Reinbacher was the best defenceman of the Draft and had him higher than the consensus. Relative to most public analytics model, Reinbacher makes sense at #5, but relative to the consensus, it looks like Kent Hughes outdid himself and made another reach with a top pick. Not all the stats brands were high on the Habs new protege. Private hockey intelligence company Stathletes had Reinbacher at #11 … behind another defenceman: Axel Sandin Pelikka.

What Habs do wrong?

Accelerating the rebuild with safer prospects

After a season and a half, Kent Hughes has shown his imprint on the team. In back-to-back drafts, he has traded picks for former really promising young centres (former 1st round picks) who struggled to get a top-6 role in their respective teams. Last year, the Canadiens made a big splash by trading Alex Romanov to acquire Kirby Dach from the Chicago Blackhawks. A season later, this bet is looking good for the Canadiens. Dach looks like a rejuvenated player under Martin St. Louis. His recent glow gave hope to the Canadiens who wish to apply the same magic to the recently acquired centre Alex Newhook. Newhook is a good player. He produced a lot of points everywhere he went and was projected to be a good top-6 C in the league. Last season, despite an Avalanche team riddled with injuries, the young talent could not find a regular role as a second centre of the team, the role that he lost to JT Compher. As a 3rd C on the Avs and with almost no good PP looks, he still managed to score 14 goals. Kent Hughes hopes that his former client gets blessed by the same MSL magic that touched Cole Caufield and Kirby Dach.

So here, we can see a small pattern: Bet on 20-year-olds who used to be really promising prospects. How good is this? The biggest advantage of this method is that the player’s floor will generally be high. Dach and Newhook both had NHL seasons under their belt before being traded and showed enough to not be worried to have to play in Europe in a near future. For both, the floor is “offensive 3rd liner that can help on the 2nd line,” which is a decent outcome. Since the team does not have to worry about wasting a prospect, they can focus on the ceiling of the player. This comes with a problem: betting on an older player is a safer bet because the player is less likely to “bust.” But the uncertainty also relies on the ceiling (or “upside”) of the player. Newhook and Dach have not shown star potential yet and they are extremely likely at this point to not be stars because their potential is about to plateau. This article about Thibaud Chatel explains really well this phenomenon about the certitude about a player’s potential relative to their age.

Thibaud Chatel’s Newsletter

How Sure are we about a Prospect?

One of the big project we had this summer was to work on prospect projections. Not only for the NHL, though it attracts the most attention, but for any league or country as part of our World database project that should be online this summer. Quick intro below if you missed it…

Read more

3 years ago · 4 likes · Thibaud Chatel

One thing we can conclude is that swinging for safe players is good on one hand because you are likely to get what you see. The backhanded side of this strategy is that you are not allowing space for surprising positive outcomes (stars).

The Newhook trade also raised concerns for me regarding the fact that I was pretty confident that some really good prospects that were overlooked in the first 30 picks would be available at the 31st selection. I am thinking that the possibility of finding a prospect with a more interesting profile than Newhook was real. My initial reaction to the trade was “Montreal paid a lot for him!” Giving the 31st and 37th selection as well as Gianni Fairbrother seemed a lot to me.

The summed estimated value of the 31st and 37th selections (orange) is roughly equal to the estimated value of the 23rd pick (in pink). In front of a situation, it is not out of pocket to ask if the current version of player that Alex Newhook is more valuable than the Draft picks. Personally, I believe that I would have preferred keeping the picks and selecting prospects instead of trading for him. I want to be clear when I say that I do not think that MTL lost a lot of value in this trade and therefore, I do not see the trade as a heartbreaking loss for the Canadiens. But I am not a fan of the trade although I am hoping I am proved wrong in the future.

Targeting the Wrong Needs

Kent Hughes said that if David Reinbacher was a left, there is a chance the team would have smaller interest in him. So it is clear that the team hopes to answer an immediate need with this selection. I am not here to bring back the debate about “Drafting for needs or not?” The Canadiens selected Reinbacher because they badly need a top right-handed defenceman because Shea Weber and Jeff Petry vanished after helping the team to reach the Stanley Cup Finals in 2021. From my assessment of the Canadiens, they have some more urgent needs than the RHD position. Montréal deeply lacks high-end offensive talent in the forward group. Honestly, with all the respect due to the players, if you team’s best forward is Nick Suzuki, do not expect to be a contending team anytime soon. Caufield, Suzuki, Dach are nice pieces to have around in the forward group. They are not bad players by any means. But this team needs game breakers. They need elite talent. I thought Matvei Michkov was part of the answer to that need. He has been producing points at an historic pace, and his most recent performances made him one of the top prospects in our records. He is just a pure offensive talent: extremely good with the puck on his stick and he has an extra sense to make some exciting offensive plays. I think Montreal needs that type of player as soon as possible to hope to build a winning culture. Even if they missed with the Drouin trade, they need to try again.

I also think that the team needs to go a little bit more for pure offensive defencemen. The Canadiens like their players to have a particular type. Their defencemen are big and can skate. Their forwards generally work hard. In my opinion, the Canadiens should open their doors new type of players like more Lane Hutson’s. On the power play, you can be big and skate, but you want the QB to be able to distribute the puck like he is a forward. It partially explains Drouin’s success on PP with the Canadiens … until Mike Matheson decided to show how it is done when he came back from injuries. Matheson was the best player of the Canadiens last season. His mobility, his hands and his ability to always take the right risks made him one of the best defencemen of the NHL last season. In today’s NHL, you need to have D’s who can jump on the plays like Shea Theodore. But there is a lot of hope that Lane Hutson becomes that defenceman who makes highlight plays on a nightly basis. It is also not impossible Reinbacher keeps his rocket rise and develops his offensive abilities in the upcoming seasons and eventually become the #1 D Habs scouts saw in him.

In conclusion, for a nth straight year, the Canadiens bet on a player that is not the favourite among the fanbase. This time, it is different from last year. It was a big risk to select Slafkovsky (according to our model) with the 1st overall pick while we can see a much safer bet in the selection of David Reinbacher at #5. One element that I need to add at this point is that Reinbacher is not your usual defensive defenceman who plays extra-safe. Despite being in a men’s league, he had the habit to carry the puck at a surprising rate. This level of confidence is rare for a player of his age playing pro. I do not think he will be an offensive dynamo in the NHL, but we need to not exclude the possibility that his offence translates to the NHL at some point and improves. As for the Canadiens, they will eventually need to get some offensive help. We saw on the 2nd day of the Draft that betting on young talented offensive players was not necessarily a priority since they loaded the prospect pool with overagers, goaltenders and the little brother of Arber. Like Vegas is showing every year, it is not impossible to trade for stars or sign them as UFAs, but … it is definitely easier and less expensive to select them at the Draft.

Let me give you the ingredients for a perfect prospect ranking that will earn you the praise from the nobles of the hockey scouting community on twitter :

1. Do what everyone does for the top-6 (something like : Bedard, Fantilli, Michkov, Carlsson, Benson, Smith)

2. Add some random european prospect top-10 because “he is one of your favourite”

3. Drop a North-American prospect to late 20’s because “he has bad skating and everyone overrates him”

Now, you have the perfect recipe to cook a perfect prospect rankings list.

If you did not understand the irony, I declare that the introduction was made on an ironic tone. The goal is not to laugh at people who make prospect rankings and shame their work. I rather focus on how to digest those lists that are appearing on a daily basis on our feeds. I did not ask everyone to show their lists, but I am still exposed to them, and it might have an influence on my personal opinions to be in touch with some kind of consensus. This phenomenon has a name in psychology : social influence. It affects people at different levels, but it has a lot of impact in hockey conversations, especially prospect conversations, mainly because people are talking about things they do not necessarily know about and fill the gaps with elements they have heard outside their own experience/research.

Prepare yourself for prospect assessment

I think that when it comes to prospects, we need a lot of humility. Aside of the major outliers — Generational players, prospect assessment is generally making betting that a 17 years old hockey player is a hit on his 0.0001%-10% chance of turning into a NHLer. Usually, these odds come a load of external factors that are not taken into consideration in models (character of the player, injuries, physical condition, etc.). So it is fair to say that you should start that feeling that, as much as you are confident in your top-5, there is a good chance that things do not go as planned. You also have to remember why you are assessing prospects. The reason why you are doing it will have an influence on your work. Watching prospect tapes just to have an idea of who they are is a very different task than working for a NHL team and providing reports based on what you have watched. I feel that you work in hockey more smartly when your work has real repercussions.

So, the article is not about complaining about Corey Pronman’s latest article… at all, and I suggest you to ignore his articles if you do not like them. I think the best draft ranking is the one you make yourself.

The Cons of NHLe

As someone who worked on tools to assess prospects, I feel that a lot of people who support our work are overly confident in the insights our models suggest. This is cool because people support our work, but the reality is that our tools are there to support scouting and not to take over or replace scouting as a whole. And NHLe is a perfect example of the beauty of analytics in prospecting… as well as how horrible it can be.

First, let me explain NHLe so that we are all on the same page. It is a measure created a while back by Gabe Desjardins (one of the Builders™ of Analytics in hockey) to put all point productions in one same unit of measure regardless of the league the player played in. It is amazing because we can now compare, with points, any player in the world. So we are all understanding that NHLe is simply an estimation of a player’s point production in NHL points? Right. The idea to put all players point production on the same unit of measure is really innovative and it helps to contextualize point production. For example, it shows how big is the difference between scoring 0.5 point per game in the NHL and in the AHL. The introduction of NHLe really changed scouting in hockey. Few public tools emerged of it and NHLe essentially became the poster stat for analytical research in prospecting.

Now, let’s do this exercise. We have 2 amazing prospects that had amazing DY seasons. Who would you rather select with a 1st overall pick if they were available in the same Draft?

Easy, right? Prospect A is having a better production in a much more difficult league while he’s significantly better than his teammates and the other players in the league. Prospect B plays in a easier league and we understand that he scored a lot more than anyone else in the league. The gab between the standard deviation between the scoring relative to the league and the team suggest that prospect B also had really good teammates. If you fall in the super easy trap I set for you, you should pick prospect A.

Now let’s see who those players are…

Now that you have the names, you would probably pick Connor McDavid first overall as you should. The goal of this exercise was to show that solely basing your evaluation on NHLe can be dangerous.

Here is an extensive list of potential flaws of NHLe — this is a list of the flaws of my first model (for the next time you want to dunk on it) :

• Model does not account for variation of scoring rates from a year to another in a given league and what cause leagues scoring to vary.

• When the model is applied to a player’s point per game, it does not give importance to the sum of game played and points. I think it does not correctly estimate the value of an additional point in a given league.

• Even if we account for scoring variation in other leagues, the NHL (target league) is not exempt from scoring variations. 

• Model does not account for how better a player is from his peers.

• Not enough emphasis on uncertainity

• Only about point production

Here are the pros :

• Easy to understand

• Really good for complementing scouting work

• If you are curious enough, it creates new questions to answer

• Can be used in other models and is performing quite well

But do not worry, for most of these problems public scouts do not have answers for them. So make your own research. NHLe has its flaws (that we try to tackle on a daily basis), but it is still one of the best tools for prospecting publicly. Great hockey analysts like me work on additionnal tools using publicly available data to make better assessments of hockey prospects for future NHL Drafts, so keep your eye open for the next time talking about prospects will be cool again (in December for the World Junior Championship).

The Case of the Eye Test

You will happy. As much as Iike coding, my boss sent me scouting a few times this season. I watched a lot of film and I had the opportunity to attend a few games to provide insight on specific players to my team. The really fun moment was when we signed players I was sent to scout. Scouting is a real job. It is a lot of work when done properly. Most of the times, I had to do film or do scouting, I was paired with at least one other person. We all watched the same games, but we usually all came up with different reports. Elements in the game stood up to me, other things stood up to my collegues. We watched the same game. Same thing happenned when we were asked to watched specific players. My collegue came up with details that I did not necessarily noticed. We also had some disagreements on multiple topics. This is why I feel that reading one person giving his thoughts on all prospects is not my prefered way of learning about prospects. I need first to trust the skill level (in scouting) the person I am about to read. Then, I want to watch the prospects myself and I want to have a discussion and challenge ideas using data and video. Lastly, we need to stay away from the easy prospect coverage lines : "Good skating, comes from good family, great leadership, comparable to [random NHLer from a random team you've seen twice in your life]". For me, this is the optimal way of doing scouting publicly. If you work with a team, you want to assess team needs because at some point player talent will be roughly the same for everyone and you will need to make a choice on who you want to develop. This choice could be be influenced by few factors like how crowded your team is at a specific position. I think doing film and going to games is super important for prospect assessment, but we need to be really careful when reading the work other people did.

Without further ado…

I told you in the title that all prospect rankings are bad except for mine, so here is my top-5 :

1. Connor Bedard : You already know him

2. Adam Fantilli : He is good offensively

3. Matvei Michkov : Really good player that did not get as much exposure as his peers due to political situation

4. Leo Carlsson : Big and produced a lot in a league where players of his age usually struggle to get their name on the scoresheet

5. Zach Benson : Small excellent two-way player who led a team, with few 1st round picks, in scoring.

This is the only good list.

Make sure to follow me on Twitter because I will have good news to share soon.

Now you are ready for Drafting season. Do your own research. Subscribe to my Substack. Have fun.

You already know I have not posted a lot in the last couple of months. Why?

1. I am still a full-time Finances student

2. I got hired in professional hockey (it is the part where you congratulate me) — Wait a couple of weeks to find out where… or just guess it

3. A lot of the work I have been doing lately was Data Engineering related (ETL — Finding the Data, Extracting it, Transforming it for Analysis Purposes, and Loading the Data in a Database where Chace and Josh were able to make amazing visualizations)

I am still partially a Data Analyst in my heart. I am still a huge fan of story telling and data visualization and this article is a testimony of that.

So what is this article about?

After building around 20 data scrapers for about 10 different hockey leagues (because you need schedule data, play-by-play data, players stats, etc.), the data analyst in me wanted to join Chace and Josh who were having a lot of fun visualizing and modelling the data I have been scraping. I already had a lot of interest in covering the QMJHL (because I have special ties with the Gatineau Olympiques and they were really good this season), so I jumped on the opportunity to make a quick analysis of what what happened in the QMJHL playoffs.

So I built a GameScore model using the last 3 seasons of the league. I was inspired by the work from Mike Murphy. If you did not know already, the GameScore is an estimation of how good a player was in a given game using Box Score statistics. It is not meant to replace video and it has a lot of limitations, but with the right usage, you can make really good analysis and tools just like Dom is doing with theAthletic. For me, it was also a quick way to get something going with QMJHL data without having to work for an additional 3 months before I make public work again.

Let’s start diving right in the research!

Who were the best players of the Playoffs?

The QMJHL featured some really good hockey players this season… No Alexis Lafrenière this season, but we got to see some players that might have a career in the NHL in a couple of years. The table above shows the 20 best average GameScore per player during the playoffs. Players on the list are usually older players (in their DY+1, DY+2,…) who played on powerhouses. Josh Lawrence is an interesting case. Size kept him from getting drafted, but his skill is undeniable and I think he should get some attention from NHL teams, although his point production might be inflated (to some extent) by the fact that he played all season with really good players (even if he is a good player himself). Zach Dean had a really good season and he kept it going in the playoffs (until the Québec Remparts completely shut him off offensively). The St.-Louis Blues should be really excited by the progression of their new protégé (he was involved in a trade (NHL) that sent Ivan Barbashev to the Golden Knights of Vegas). He was the best player of a really talented Olympiques team. To complete the top-3, we have Justin Robidas (Carolina Hurricanes prospect). What is fascinating about him and James Malatesta (Columbus Blue Jackets prospect) is that they have been consistently good… even in the last 2 rounds of the playoffs. Against the top teams in the league, they have found ways to get on the scoresheet.

Blame the QMJHL for not having a transparent logo of the Voltigeurs’ logo

The table we see here shows the best playoff performances (GameScore wise) we have seen this year in the QMJHL playoffs. Without any surprises, most of them are top players on powerhouses feasting against the worst teams of the playoffs.

Before looking at points, NHLe and everything we need to understand the context of the league we are assessing. The QMJHL is a SUPER UNBALANCED league. Blowouts happen more often than in pro hockey. Good teams are really good and could probably compete against a couple of teams at the professional and university level. The non-top-teams are bad and really bad teams. They are usually really young and they probably have a lot of draft picks. “Tanking” is a common theme in junior hockey. When you don’t have a shot at winning the Cup, you usually sell all your assets of value to bank on draft picks and prospects to prepare the next run at the Trophy Gilles Courteau. Aside of few surprises in round 1 and 2, the outcome of the playoffs was not a huge surprise : we mostly had a really good idea of which teams were going to be in the semi-final and final. Québec had a good run last year, and they have added few players to an already solid core to win the championship.

Screenshot of the standings in the QMJHL. All teams in the top-4 ended up in the semi-finals and the #1 team won the final against the #2 team.

So in recap, what did we learn today so fair :

• Unlike the NHL where the difference of talent between the best and the worst teams of the playoffs is smaller, the QMJHL has more blowouts during the playoffs and some players can take advantage of that to inflate their point production.

• In your team, you want players that will find ways to produce against good teams (because at the end of the day, that is more important than having a 6-point night against a bad team)

Looking at the playoffs of 3 players you might be interested in

James Malatesta, the MVP of the playoffs

Aside of being a Columbus Blue Jackets prospects, James Malatesta is also one of the most talented players in the QMJHL. I remember he was a highly anticipated prospect when he got drafted in the league a couple of years ago. During the playoffs, Malatesta showed up big in big moments. Scoring 14 goals in 18 games is not impressive enough, 5 of those 14 goals ended up being game-winning-goals. This is why he got the MVP of the playoffs. As we can see in the plot above, facing top opponents did not slow him down. He managed to keep a really good level of play in the semis and the finals to help the Québec Remparts to win their 1st QMJHL Championship since… 1975-1976.

Joshua Roy, the Hab’

2 Habs prospects suited up for the 2023 QMJHL playoffs : Riley Kidney and Joshua Roy. Despite being the one who got drafted the lasted among the two, Roy is the one who is getting the most attention after a really good DY+1 campaign and a really good showing at the latest World Junior tournament. Joshua Roy was a former QMJHL 1st overall pick who started his junior career in an atypical way : lots of concerns about his attitude and his overall involvement in the play. The bad noise surrounding his name got him to fall to later rounds at the NHL Draft despite evident signs of talent. Since the Phoenix of Sherbrooke traded for him, it is a new man and all the negativity surrounding his name completely vanished. He got drafted by his the NHL team he grew up watching and he is making a case for a really good Disney story. Now that you know about the story behind Joshua Roy, we can focus on his playoffs. Like expected, the best player of the Phoenix showed no signs of mercy for his opponents in the first two rounds. He was everywhere offensively, making the right plays defensively, he looked like no one could stop him… until he faced the Halifax Mooseheads. That Mooseheads team was really difficult to play against. They gave ferocious adversity to anyone playing them. It looks like they competently nullified Joshua Roy. He was left with no answers. It was still good playoffs for him.

Tristan Luneau, the D

After a season where he broke almost all franchise records you can imagine in the history of the Gatineau Olympiques, Tristan Luneau had the task to take the Olympiques to the Gilles Courteau Trophy without his Anaheim Ducks teammate Noah Warren. Luneau was one of the best players of the most talented team we have seen in Gatineau in years and he was eager to push his team to the Cup. Unfortunately, it did not go as planned. Like expected, he looked amazing in the first 2 rounds against weaker opponents, but he could not help his team to steal a game against the Remparts. The Remparts have swept Gatineau, but almost all games ended up in overtime or with a 1 goal difference. It was a hardly fought series and Luneau did good… but not good enough. Costly turnovers and bad reads defensively, and it’s a Québec goal.

Closing thoughts

I hope you enjoyed your read. This is a ~1,500 word applications of some work I have been doing for months now.

In summary :

• Follow Chace and Josh (you can also follow me) because the best public Draft Analysis tool — TheDraftDigest — will be released soon

• I built a lot of data scrapers for hockey data because it might be useful in the future and…
  “The ability to take data – to be able to understand it, to process it, to extract value from it, to visualise it, to communicate it – is going to be a hugely important skill in the next decades.” —

  Hal Varian, Chief Economist, Google.

• All leagues have their own characteristics and understanding them is key before you make conclusions

• Difference of talent between the powerhouses and the rest of the league is a lot more significant than what hockey fans can be used to in the NHL

• What made a difference is that the best teams were able to shutdown the opposition’s best offensive players

• Good games against weak teams inflate aggregates (I know it is simple and super logic, but I think it is worth mentioning)

• With video analysis and having understanding of the injury context, I think we can dig even more into why top players such as Joshua Roy and Tristan Luneau were hand-tied against the best 2 teams in the league, and we can suggest more applicable answers.

Make sure to stay put because a lot of public content is coming from me... if I decide to take a break from learning again.

Spoiler: I did not win one of the primary prizes, but my team won the secondary prize.

A hackathon is a programming competition where organizations show us real-life problems. In teams of around 3 or 5 people, we have around 12 hours to submit a solution to the problem using programming. There are prizes involved. There are also professional (mentors) out there ready to answer questions if needed.

What did I learn?

Variety of skills matters

One team had three people doing a Masters in Project Management and only one person who knew how to do programming. Having a diversity of skills is a really big advantage in any project. Someone will take care of the Data Engineering (extracting data, transforming Data, loading Data … yes, it’s the famous ETL), someone will take care of the building the models and someone else will work on the User Interface where the potential client will navigate in the Dashboard. This is some basic way of organizing team based on skills, but it can be different based on what you have in your context. It allows everyone to be able to work simultaneously and not waste any time (since time is limited for any coding competition).

Learn Git

Lately, I learned a lot of it about sports analysis. One of the main things I learned about are the things I don’t know that I need to know. Among those things, there is Git.

Git is a distributed version control system that tracks changes in any set of computer files, usually used for coordinating work among programmers collaboratively developing source code during software development. Its goals include speed, data integrity, and support for distributed, non-linear workflows. (Wikipedia)

When you start coding, you do not realize it but you will face a few problems in a near future:

• Collaboration: You will work on projects with other people

• Version control: Managing changes in your code

Why is it important? When you start working on bigger projects than print (“Hello World!”), working on the same project as your collaborators at the same time while you are all focusing on different parts of the code or on different files from the same project. You will also find it useful to keep track of all changes made to the project. It makes the team more efficient, and it can help to have a lot of time. Learning Git on top of my to-do list.

Communication Matters a Lot

One of the biggest challenges of my time at the hackathon was the communication. On a team, it is important that everyone follow the same goal. If one part of the team is going left and the other part of the team is going right, you will be really lucky if your submission is successful. Having a leader is a really good way to organize the team and make sure that commutation is not a problem. I feel that the lack of communication within my team almost made us submit nothing for the hackathon. Communication in the presentation is also really important. One of the judges told me that we spent time presenting things that did not matter much while we left out information that could be more relevant. We only had 3 minutes to do the presentation and we should have spent more time working on it to make sure that those 180 seconds really reflect our work on the project and how useful it can be to solve the problem. When I write an article, like here, I do not expect you to have 30 minutes to read me, so I have to go straight to the point so that you get the full idea of what I wanted to share.

I Won

As I stated earlier, my team failed to win one for two main prizes. Instead, we won the secondary prize. But I think it is what makes me a winner. I think that I won because I reached my goal and I went further. I made connections with people that have the same interests as me. I also found out that with enough focus and organization, we can do a lot of work in a short period of time. I feel that if I consider any task like a Hackathon problem and I use the same mindset as I had during the event, anything will be achievable. Additionally, I already knew I liked to work with sports data, but this event opened my mind to other types of data. In my project, specifically, I had to deal with the geospatial data and I thought it was really interesting. Solving problems using Data is what I want to do in life. Finally, I found out that that competition like Hackathon is something I really liked to do and I am looking forward to participating in other programming competitions in the future. It was a great learning experience, and I had a blast.

My project was about finding a way to identify Buffer Strips in the Outaouais region, look for their quality, see if they respect city rules and look for any anomalies. Finding Buffer strips data for the Outaouais region was really difficult, so we ended up using Data for the city of Laval. Then, we mixed real information with fake Data we generated using Python scripts in order to be able to submit a prototype of an answer that makes sense. Here is the Dashboard we built and we submitted for the project. (It is in French)

Read more

Getting familiar with few terms

I think most of you are familiar with those terms, so it should go fairly quickly. I have this short glossary section because I was told to write articles like my mom (who does not know about hockey at all) would read it and it is also a way to all start on the same page.

Shots on goal

It is a shot that the goalie had to save or a goal. It does not include blocked shots (shot blocked by forward or defensman) or missed (shots that hit the post are missed shots per convention).

Corsi

Any shot attempt. Goal, miss, blocks, shots on goal. Any shot attempt.

Fenwick

Any unblocked shot attempt. Goal, miss, (no blocks), shots on goal. Any unblocked shot attempt.

Expected Goal (xG)

Probability that a shot attempt results in a goal based on a variety of characteristics.

In the public sphere (I don’t know what they are doing in the depths of the private sector unfortunately), Expected Goals are Fenwicks that were weighted for their shot quality.

How often does the team who won on paper wins in real life?

From far, I could not watch all Habs games but I am always following how good they are doing using my favourite stats websites. I noticed that the Canadiens are winning a lot of tight xG games and looking at it more, I was curious to see what was the outcome of Habs games based on their xG differential. But, it was not enough, I needed to know more. So I came up with a question that we will attempt answer together this article : How often does the team who won on paper wins in real life?

To define who was winning on paper, I am using the sum of xG (using HockeyR’s model) during the regular time (so non-OT and non-SO) of the regular season. The reason behind this choice of only using the first 60mins of regular season game is that I think there would be no OT if the game was won on xG during the regular time.

For the article, I am using the play-by-play data of last 3 seasons (2019-20, 2020-21, 2021-22) that I was able to gather using HockeyR. At the end of the article, I am using 2022-23 data because I thought people would be interested to see how the research can be applied to their team this season.

How xG relates with wins?

The team with most xG won the game 63% of the times in 2019-2020, 69% in 2020-2021, and 71% of the times last season. So, roughly, 2 games out of 3 were won by the team with the most expected goals.

In the legend of the graph, we see a NA label because in 15 occurrences, the home and away team had exactly the same amount of expected goals. It is something that should be rare since all xG values are numbers with 6 digits after the decimal point.

I thought you would be interested by how the Canadiens did in the last 3 seasons.

We can see that in the last 3 seasons, the Canadiens have lost good share of the tight xG games.

I am not satisfied, and I want to have a little bit of fun, so let’s build a simple linear model that guesses who won the game based on their xG differential after 60 minutes.

The model was actually worse than I expected. It guessed who won the game 50% of the times. I was expecting something closer to 66% since we saw earlier that teams who had the most xG won roughly 2 games out of 3.

To help with the fun, I made a variety of models to help predict who would win a game based on the sum of metric for both teams and with the metric differential. By example the corsi model uses home team corsi and away team corsi while the corsi model* uses the corsi differential between both teams. Those were really simple linear models as basic as you can build in R. The findings are actually really interesting. The Corsi model did better than all other models using only one metric. All models roughly had a similar performance (around 51%). The model who did the best (by a significant margin) was the model model using the differential of all variables.

Logically, it makes sense that the team with the most shot volume (and shot quality) does not necessarily win the game. Looking at metrics like shots on goal, shot attempts, unblocked shot attempts and expected goals, I completely ignore one aspect of the game that can have a big impact on the outcome of the game : goaltending. When the opposing goalie has a good night, you can have the best offensive chances in the world, but you might still leave the game with a loss. The opposite scenario also exists. When a goalie has a “stinger”, your team can play good defense but if the goalie is a swiss cheese (allows a lot of goals on few shots), you will end up with a loss and you won’t have much power on it.

What does that mean for the Canadiens

The Canadiens still have a losing record, but they generally win the games where they have the edge in xG. They have also won a good share of games where they generated less quality than their opponents, thanks to clutch offensive goals by Cole Caufield or clutch goaltending by Samuel Montembeault.

Closing thoughts

Before writing the article, I thought that winning the game on paper was attempting more shots and having more expected goals than the opposing team. My theory had a major issue : it implies that only skaters (forwards and defensemen) have an impact on game outcome. I mentioned that goalies could have an impact on game outcome, but let’s not forget about coaches and referees. I think that by playing different combinations of players together and matching different players with opposing players, coaches have an impact on game outcome. Referees have a neutral role in the game, but it does not mean that they have a neutral impact on the game. They are the ones enforcing the rules and calling the stoppages of play. Their actions can have major repercussions on the momentum of the game and eventually the outcome of the game. Calling a penalty late in the the 3rd period of a tie can change the script of how the game will end.

I thought winning on paper was winning the xG battle, but I think there is a lot more in “winning on paper” now that i think about it.

Acknowledgements

Dan Morse, the person behind HockeyR

MoneyPuck.com (for their Deserve to win -o- meter) for inspiring this piece

It was a special year for me. Just like a lot of people, I have experience a lots of downs, but what made it special was the ups outshined the downs. I have pushed myself to new levels and the biggest win of 2022 is that I’m proud of the man I’m becoming and I feel more ready to attack the year 2023.

Let’s talk about 2023. I have a lot of projects coming for this year. I don’t think I will be able to show 100% of the things I will be working on this year, but something to keep an eye on is a project I’m working on with some super good people. Chace McCallum messaged me in November to ask me if I wanted to join him on a research project about prospects in the NHL. Later on, we have added Joshua Khalfin to our team. Chace and Josh are really smart guys who have done a lot for the Hockey “Analytics” community in the past. Sometimes, I feel the “imposter syndrome” when I think about the idea of working with them. On my Twitter account, I like to share glimpses of things we are working on… so keep your eyes open because we will release something really cool for everyone a little before the NHL Draft.

In 2022, I also got to meet Mikhail Nahabedian. Mik was my hockey research mentor for more than a year now. He has pushed me to get better like a dad would do for his son. In 2022, La Force de Montréal (WHKY) offered him a job (much deserved btw). I’m glad I had the chance to meet him. I also had the chance to make new connections with a lot of people from the community and I’m grateful for everyone I had the chance to interact with.

In 2022, I sent my resume to few QMJHL teams to help with their research and development team (or analytics department) and I got no answer back. My goal in 2023 is to have the opportunity to get a “real” internship with in hockey. In 2022, was also a year where I learned a lot and I became a lot more educated. I hope 2023 will be similar. I can’t stop learning.

Thank you all for the support in 2022 and wishing you all the best in 2023!

Max

All those linking and association reading made me think about… HOCKEY surprising! The book showed me a tutorial on how to see how we can find out how related are the products of a store based on what people have bought using the Apriori algorithm. So I did the same process using the NHL play-by-play data. I won’t lie here, I did not know where I was going with that… but I assumed it would be giving me chemistry between players.

Before we go further, we need to understand what is Apriori. Here is what Wikipedia has to say :

Apriori is an algorithm for frequent item set mining and association rule learning over relational databases. It proceeds by identifying the frequent individual items in the database and extending them to larger and larger item sets as long as those item sets appear sufficiently often in the database.

You get it? If you are not sure, you might get a better idea later when I explain how I implemented Apriori with hockey data.

Basic Methodology

1. Get the season play-by-play data of the 2021-22 NHL season

2. Filter the data to only keep the rows where the event is a “GOAL”

3. For every team, I have filtered the PBP data and applied the Apriori algorithm on the skaters who were on the ice for the team who scored the goal

4. Merge all the results in one big dataframe and look at the mess I have done

I am filtering for goals because I assume that if players have a good chemistry, they will be on the ice when the team scores goals and we can find associations there. Also I did not know what else it could mean (in all honesty). I have also filtered out goalies because they are always on the ice.

So here is what it gives me when I sort the values by the strongest link in a descending order :

Because I am a Habs blogger and fan, here is the Canadiens view :

Oh! That’s cool! I did not know what to expect, but if my “chemistry tool” tells me that the Tatar-Danault-Gallagher line had the top score, I assume it is good! Actually, what roughly happens is that for each Habs goal, the algorithm sees who were the Habs players on the ice and he gives a score based on the frequence of which they are together relative to the some of all goals Habs scored last season. In simpler terms, for every group of at least 2 players, the algorithm gets their number of occurences divided the the sum of all the goals the Canadiens have scored. Did I just repeat the same thing twice? Yes, I did and hope you understand better.

Adding few steps to the Methodology

5. Do the same procedure as before but for goals against each team and get the revulsion (I just needed to find a new word who means the opposite of chemistry) of each group of players

6. Create a few metrics like link or chemistry differential (positive_link - negative_link) and link or chemistry percentage (positive_link / sum of both links)

7. Merge everything together and look at the mess I have done

What the column values mean 

link_x : Proportion of team goals for where the group of players is on the ice
link_y : Proportion of team goals against where the group of players is on the ice
diff : link_x - link_y or +/- of the proportion of team goals for and against  
t% : Proportion of the Proportion of team goals for where the group of players is on the ice relative to all the goals scored when the group of players was on the ice. Formula : link_x/(link_x + link_y)

As I am writing this, I am feeling that the Apriori algorithm is giving me on ice GF% stats for any groups of players in the league… and it is really fun. Before I leave, I am giving you the data of Habs groups who have both a link_x and a link_y (because one thing I noticed is that some groups have only one of the links, probably due to the lack of data in the other link).

This is it for this week. I hope you enjoyed the read and finding out with me how we can use the Apriori Algorithm with hockey data. The next steps for this experiments will be to weight events and maybe try it with expected goals. This experiment made me be curious about algorithms and I would like to try to play with more algorythms with hockey data.

If you need more informations about algorithms, I strongly suggest you to read about it. Have a good week and see you next week for the next article of the newsletter.

Today, let’s just make a quick analysis of the cap situation for the Canadiens. Let’s start by looking at how teams allow their cap for each position compared to the average. The Z-Score metric shows the distance from the average.

Here we have an overview of how every team of the league is allowing more or less % of their the money they spend on players per position, but we want to focus on the canadiens, so let’s only look at the Canadiens.

Few things we need to be aware of before we go further :

• If a player is on IR or LTIR, it will automatically affect the team’s Z-Scores because his cap hit won’t count on the cap the team is allowing for their players. A good example is Joel Edmundson. He is has a cap hit of over 3M$. Since he is injured, the Canadiens’ cap hit data doesn’t care about Joel Edmundson for now, but when he will come back healthy, the team will be spending an additionnal 3.5M$ on defensemen. Proportions for D’s will get bigger (Z-Score too), and proportions at other position will get lower and their Z-scores will be affected. It will also affect the league’s means for proporptions of cap hit% spent on all positions.

• So just keep in mind that some teams are dealing with injuries and players out. Data is not perfect here.

So, let’s have a look on Montreal.

With Carey Price off the cap, their proportion of cap hit allowed for forwards is spiking at 67% and we see it is significantly higher than the rest of the league. With many D’s on ELC or on cheap contracts (because of the injured Mike Matheson and Joel Edmundson), we see that Montreal’s defensemen are not taking a big portion of Montreal’s cap hit. With Price out, Montreal is rebuiliding with a cheap duo of goalies.

If we use money allocated for each position instead of using proportion of cap hit, we would get the same story for Montreal. They are spending a lot on forwards (despite them not scoring a lot of goals), and under the average on defense and goaltending.

If you want to see the cap hit of every player on the Canadiens, consider looking for the information on CapFriendly.com

If you missed it, consider reading my previous article.

HabsBrain

Habs season survey - Results : Part2

Last week, I published the 1st part the results of the Habs Season Survey I made before the season. Why did it take that long to publish the results? I am still a university student managing mid-terms with work and life. But now we are done. Here is are the final results of the survey…

Read more

3 years ago · 2 comments · brain

Lets’s have fun

I wanted to know if teams spending more on a particular position were winning more. So, using the data from spotrack, I complied the data from last 2 season and I found out that teams spending more on forward (and allowing more % of their cap hit on forwards) were winning more than the teams who spent less money on forwards.

For fun, using the data from the last 2 season of how teams allocated their money in each position, I built a multiple linear regression model to estimate how many wins a team would get based on their cap allocation per position. The model won’t be perfect, but it is just fun to build models.

Here, the model has an R of 0.59 meaning there’s a moderate positive correlation between real wins and predicated wins of the model… and it is perfectly fine this way for the excercise.

With Montreal’s current cap situation, the model predicts 37 wins for the Canadiens.

Subscribe now

Of course, this small excercise lacks of precision. NHL teams salary cap is something that changes every day. Minor events (like recalls and injuries) can have major effects on team’s salary cap and the model didn’t really give importance to this important detail. Also, only 2 seasons of Data is not enough. More data would be more than welcome to build a stronger model.

Finally, Cap allocation itself can’t predict how many wins a team will get, but it is always fun to see different relationships between variables and how they correlates with winning.

HabsBrain

Habs season survey - Results : Part1

This is the first of many articles you will see on a weekly basis. Here, the goal was to visualize the results of a survey I made about a week before the season and that I closed after the first game of the season against Toronto. No real analysis to make here, just enjoy the viz and vibe…

Read more

3 years ago · brain

Last week, I published the 1st part the results of the Habs Season Survey I made before the season. Why did it take that long to publish the results? I am still a university student managing mid-terms with work and life. But now we are done. Here is are the final results of the survey.

I could have posted it only on Twitter, but since I care a lot about you, I am assuming that you will be happy to give me clicks on my Substack in exchange for being able to see the best quality of the Viz. We are all winners here.

I am thankful for everyone who answered the survey because without you there would be no visualization.

Subscribe now

See you next week for the next article of the weekly newsletter.

For the viz you see above, I wanted to challenge myself into doing something different from what I have been doing before. This is why I spent a good 2 days of searching for inspiration. For the points viz, there was no real tutorial on how to make this precisely… so I had to do what any programmer would do : separate the big task into multiple smaller sub-tasks and find the answer for each of them on Stack Overflow. Form a personal experience, part of getting better at programming is knowing more about what you are looking for on Stack Overflow. Programming is big and I don’t think I will know everything by heart one day, but experience has taught me they key words I need to write on Google to find the answers I need when I need them. I think this viz has about 50% of the questions of the survey. The goal is to make the viz with all the questions/answers for next week.

I have started working on some work about potential topics for the newsletter and I can tell you you will like it. Next week, I’m releasing part2 of the survey results, but after that, we might take a look about the performances of the Canadiens early this season.

Subscribe now