Featurettes: Financial Markets vs Hockey Analytics, Jets Play Style, Heinola Offhand Side, Is Chibrikov Real, etc
Discover the Best Reader Questions, Article Flashbacks, Hockey Analytics Basics, Powerlifting Principles, and Financial Wisdom
Another day where you post questions and I try to answer, plus some other stuff.
We'll start with the reader questions, which you can submit via Twitter @TheFiveHohl, my personal BlueSky, our Facebook page, Reddit, and our Substack chat group.
If you are a Twitter person, come check out BlueSky. It’s growing for hockey and it’s content quality is much better than Twitter right now, in my opinion.
QUICK LIFE UPDATE
Ever eaten so much food that it’s hard to keep your eyes open?
Yeah, that’s exactly how I feel right now. Woof.
I had to push this post back a day. At first, I had almost no questions for the week—actually, I only had one. But after pressing for more on Friday, a bunch rolled in.
So, in the words of Mario: here we go!
READER QUESTIONS
Twitter Questions
Oh boy! How long you got? There are quite a few parallels between hockey and financial market mechanics that create shared heuristics. I’ll provide two core examples.
I know you’re familiar with some of the financial market side of things, but I’ll go over it briefly for our readers.
FUNDAMENTALS, RANDOM WALK, AND RANDOMNESS
The Efficient Market Hypothesis (EMH) states that all publicly known information is reflected in current prices. Basically, most people cannot beat the market over the long run—unless they consistently know something others don’t. Over- and under-performance in the short term is mostly noise or variance: think of a dog randomly bouncing around while the market is the person holding the leash.
Hockey has a similar feature. The “true” expected goal skill of players and teams is like the market or the person holding the leash. Team performance with PDO (a metric combining shooting and save percentage) above or below their “true” expected goal skill is the random walk or the dog bouncing around.
In hockey, we estimate true expected goal skill by breaking the game into its fundamentals:
Shot volume (Corsi): Create more chances than your opponent.
Shot quality (xGoal per shot): Make your chances as good as possible and theirs as bad as possible.
Finishing, setting, and goaltending: Capitalize on the chances you get.
There’s still noise in the measures we use to estimate these fundamental skills. Shot volume (Corsi) has the least noise, which is why it tends to predict future performance better than expected goals (xG) or actual goals on their own.
However, you can’t perfectly estimate a player or team’s true level by combining measures either. Regular readers will recall the weighted shots model I use, which takes Corsi but weights goals more heavily than non-goal shots, with goals slightly regressed toward xGoals.
This model predicts team performance better than Corsi alone, both in back-testing and out-of-sample scenarios.
RISK FACTORS, RISK AND REWARD
The EMH doesn’t mean you can’t achieve greater returns than the market over the long run. There are underlying risk factors that historically increase ROI—provided you don’t overpay in fees. These factors increase risk, but investors expect to be compensated for taking that risk.
The idea that you can increase the risk of the worst-case scenario but expect better long-term outcomes applies to hockey, too. So does the human bias toward risk aversion.
In hockey, we often hear about chip-out and dump-in plays being the “safe” choice. The logic is that getting the puck out (and likely into the opponent’s hands) is preferable to a turnover at that spot. But this ignores all possible outcomes, their probabilities, and their expected values.
For example:
You can still turn over the puck in the defensive zone trying to chip it out—Logan Stanley does this a lot.
Even successful chip-out plays often lead to the next non-neutral zone possession being in your defensive zone anyway.
The numbers suggest teams could take more “big mistake” risks to improve long-term performance.
A drafting example: I once compared defenders who were drafted at similar spots but differed significantly in their scoring. Group 1 was “big guys who don’t score much,” while Group 2 was “smaller or medium-sized guys who score a lot.”
Group 1 was seen as the safer bet—guys who could, at worst, become depth or third-pairing players. This turned out partially true, they were depth and 3rd pairing guys more often.
But, the lack of ceiling made them more likely to bust.
I don’t fully agree with the premise of the question. Most teams, good and bad, focus too much on playstyles instead of simply acquiring good players and strategizing around them.
That said, I’d encourage more playstyle variation among lines and players. For example, the Adam Lowry line playing dump-and-chase works because of their role and how effective they are at it. But it’s a travesty that someone like Cole Perfetti—prone to chip/dump plays—hasn’t performed as well as he could in a more skilled environment.
Rude!
(FYI: Thibaud works in Europe and is awesome. My tracking spreadsheets are actually based somewhat on his because when I worked at HockeyData I was using internal software and databases).
Yes. My old site Hockey-Graphs showed this, by Dom Galamini who now works for the Buffalo Sabres.
Domenic found that, on average, pairings with the same handedness performed worse than those with opposite handedness.
There are two caveats to this:
It may understate the issue, as coaches likely try to use defenders who are best suited to playing on their offhanded side.
It may overstate the issue, since players who play on their offhand side are generally from the bottom of the roster.
To address this, Domenic created a model to predict the impact of handedness. This is important because, when choosing defensive partners, it’s not just handedness that matters—it’s also the overall skill level of the two options.
A slightly better, same-handedness option might be preferable if the skill difference outweighs the negative impact of handedness.
For Heinola, playing on his offhand side means we need to adjust performance expectations. This is especially true because he’s partnered with a player who tends to bring down the performance of his partners.
There are several reasons why the Jets are playing Stanley:
Size and physicality bias
"The devil you know vs. the devil you don't"
Seniority
Penalty killing
Sunk cost bias
The Jets want four penalty killers in their lineup and currently believe Stanley can kill penalties while Heinola cannot. That’s debatable, but it’s likely a driving factor behind their decision.
Players with term are valuable because they’re not just rentals but also potential stopgaps while you try to stay competitive.
The trade for Nino Niederreiter was fantastic for this reason. Similarly, turning a rental into a player with term—like Vlad Namestnikov—can be a big win.
Jets window is open for as long as they have Connor Hellebuyck being Connor Hellebuyck, but they need elite talent, especially at forward, more than depth.
His point production in no ways tells you that he should still be on the team when Ehlers returns. It does tell you though that he’s as likely as it gets to sticking around despite waivers being a thing.
When using small-sample models like RAPM Corsi and xGoals, we can get a slightly better picture:
Nikita Chibrikov ranks 7th out of 15 forwards in goal differential, 5th in expected goal differential, and 11th in Corsi differential.
These numbers, while extremely noisy due to small sample sizes, suggest Chibrikov's performance is unsustainable. However, they also indicate he’s not an anchor and can compete at the NHL level.
It’s best not to read too much into these stats until we get to the 10–20 game mark. Boring answer, I know.
BlueSky Questions
At HockeyData, we had a three-pronged approach.
We would offer our services to teams at a significant discount—essentially losing money—in return for higher-quality video feeds than the ones publicly available (none of the kiss cam or similar distractions).
We would then use that data for modeling and reselling to interested NHL parties.
However, the real money came from creating reports for individual players on the team. Teams would inform players that these services were available, and we also made agents aware of them.
Individual players were not particularly profitable, but since the data was already tracked, the process was highly scalable. The reports were largely automated based on trends in the data.
We would provide every player with two reports. These reports had the same datasets, visuals, and information, but the analysis would focus either on strengths or weaknesses.
Weaknesses Report
This report highlighted areas where players could improve. For example:
If a player had low goal rates but high shot rates, we would analyze their shot distribution charts. If these charts indicated poor shot quality selection, we’d discuss ways to improve individual shot quality based on their microstatistical and analytical profile.
We’d also make specific suggestions for players to share with their skills coaches.
At the NHL level, players are typically selected for doing things the right way to the best of their abilities within their skillsets. This is less true in developmental leagues, where the talent pool is more diluted. Consequently, this type of information tends to have a much higher ROI in those leagues than at the NHL level.
Strengths Report
This report highlighted the player's strengths. Using the same player as an example:
If the player had high shot volume, we would emphasize how they likely tilt the ice in their favor to generate so many opportunities.
This type of analysis is particularly valuable for players outside the NHL. Non-NHL teams lack the scouting resources of the big leagues. For instance, a Junior A player hoping to play in the NCAA may only have been scouted live three to six times by a prospective team.
A detailed report package allows players to reinforce what scouts observed. For example, a team might read the report and say:
"Oh yeah, he did very well in transition and was always looking to put the puck on net."
That’s an oversimplified version of the process, but it captures the basic idea.
Reddit Questions
None.
Substack App Questions
My thoughts on this are that it’s probably a good thing for the sport. More goals generally mean more drama. That’s not a certainty, though; the largest sport in the world is also one of the lowest-scoring.
I have a pretty good theory on where this trend is coming from:
The average expected goal production rates for teams have been increasing steadily as far back as the data allows. The median team today generates more expected goals than the top teams did in the early years of the Behind the Net era of hockey.
Power plays have been a factor, but so has shot selection. As teams have gradually adopted analytics into their strategies, they are optimizing approaches to increase scoring chances.
This evolution mirrors what happened in basketball:
A few other factors also relate to arguments analytics has been making for years:
Gone are the days of three-minute ice-time pugilists. And—while Jets fans can argue about Logan Stanley—so are the real “coke machines on skates” who simply sit in front of the net, doing little more than clearing the crease.
The game is also becoming faster and younger as coaches begin to accept that peak performance occurs earlier than was once believed.
While these changes may have happened eventually without analytics, it’s no coincidence that these were focal points of early analytical research and that teams have gradually adapted.
Some additional strategies analytics has advocated for include:
More 4F1D power play units
Increased use of controlled zone entries and exits
Handedness optimization
Roster optimization
Player archetype and chemistry optimization
HOCKEY ANALYTICS FYI: WOWYs, RelTM, and RAPM
Ryan, a gym buddy of mine, asked a question on Twitter when I was discussing Cale Makar and Nathan MacKinnon’s WOWYs.
For context: Most models early in the season suggested that Quinn Hughes should be favored over Makar for the Norris Trophy. Those in favor of Hughes highlighted how Makar has performed without MacKinnon.
WOWYs - With or Without You
WOWYs evaluate how two players perform together and apart:
For example, we can see here that Mark Scheifele performs well with Nikolaj Ehlers but poorly without him. Ehlers, on the other hand, performs about equally well in both scenarios.
However, WOWYs have two key limitations:
WOWYs only show three states: players together, Player 1 apart, and Player 2 apart.
WOWYs are agnostic to differences in other variables, such as linemates and opponents.
RelTM - Relative Teammates
RelTM is essentially an enhanced WOWY. It aggregates all a player’s WOWYs, weighting each by ice time, to assess overall impact.
For example:
If Scheifele’s Corsi% improves by six percentage points when playing with Ehlers, that’s a RelTM of +6% for Ehlers.
Now let’s assume Ehlers skates with three other linemates whose effects are +6%, +2%, and +2%, each with equal ice time. In this case, Ehlers’ RelTM would be +4%.
RelTM can apply to percentage shares, like Corsi%, or differentials, like goals. It’s a solid measure for determining whether a player makes their linemates better.
For fun, the top 11 forwards in 5v5 RelTM goal +/- per hour since the 2017-18 season (minimum 5000 minutes):
Mark Stone, Auston Matthews, Artemi Panarin, Conor Garland, Pavel Buchnevich, Nathan MacKinnon, Connor McDavid, Robert Thomas, Anthony Mantha, Chris Kreider, and Nikolaj Ehlers.
Ehlers also ranks 4th in RelTM Corsi +/- per hour among the same sample of 221 forwards.
RAPM - Regularized Adjusted Plus-Minus
And this is to go even further beyond relTM.
To simplify a complex system: RAPM uses regression techniques to analyze each shift as a unique data point. Each shift includes ten skaters and their differentials (Corsi, xGoals, or actual goals).
Adjusted Plus-Minus estimates each player’s value.
Regularized Adjusted Plus-Minus uses regression to reduce uncertainty in those estimates.
RAPM also accounts for variables like zone starts, game score, and back-to-back games, providing a more nuanced measure of performance.
WOWY evaluates how a player performs with or without a specific teammate.
RelTM determines if a player consistently makes their teammates better.
RAPM measures a player’s ability to improve teammates and hinder opponents, accounting for situational context.
For fun, the top 11 skaters in RAPM goal +/- per hour since 2017-18:
Auston Matthews, Nathan MacKinnon, Artemi Panarin, Patrice Bergeron, Connor McDavid, Matthew Tkachuk, David Pastrnak, Mark Stone, Aleksander Barkov, Pavel Buchnevich, and Nikolaj Ehlers.
Ehlers also ranks 6th in RAPM Corsi +/- per hour over the same sample of 221 forwards.
PERSONAL HEALTH & FITNESS: Creatine For General Health
This isn’t completely novel or breaking news, but more and more research is emerging to support the idea that creatine—the best supplement for strength and hypertrophy—also has cognitive health benefits.
For background, creatine is widely regarded as the best legal supplement for sports performance. Very few supplements make it to what Dr. Eric Trexler calls the “Tier 1” category. That list essentially includes just creatine, protein supplements, and caffeine.
If you want to dive deeper into creatine, here’s a three-hour podcast on the topic:
But yes, increasing research suggests that creatine supports brain function and may help prevent cognitive decline and related issues later in life.
So, it’s time to get jacked and smart.
PERSONAL FINANCE: Cheating
I’m cheating a bit here—today’s personal finance topic is actually the earlier question comparing hockey and financial markets.
CLOSING REMARKS
Thank you for joining us in this week's featurettes. We look forward to next week’s Monday Review, where we look at ongoing Jets story lines, prospects, and such.
If you've found this article or others enjoyable, please consider sharing it on your preferred social media platforms. I would really appreciate it if you follow us on Facebook and share this post to anyone you feel could benefit from our information!
Your support is greatly appreciated, as sharing helps us expand The Five Hohl and allows us to dedicate more time and resources to this project. It means a lot to me and I really appreciate your support. I’ve said “appreciate” a lot but it’s true.
Finally please subscribe. If you are financially able to, take a look at potentially upgrading to our paid tier. There you will recieve manually tracked microstatistics and analysis on said stats, the ability to comment on articles, and access to TFH’s archives.