Reflecting on Six Years of Tracking Data
If you’re unaware, I spend all day of almost every day during the hockey season watching hockey games around the world and tracking data of the player I’m analyzing in that game. The focus of my data collection is to clearly identify the events the player is involved in, their impact on the game, and the results of their impact one way or the other. Whether it’s defensive, passing, transition, playmaking or shooting data, the focus is simply “what is happening on the ice, what is the player doing, and what is the outcome”. One of the things I’m most often asked by fans, scouts, and other hockey people is if there are any factors in my tracked data that can be reflected upon and carried forwards with regards to draft strategy. I have been at this project in some way since 2019, and the amount of data I gather has increased significantly with refinements to my strategy and thought process. After a season, I usually have a small list of things I’d like to add to the tracking sheet for the following season, and hope that that list gets shorter every year. For 2025, the only metric I’ll be adding is tracking the total number of defensive zone cycles established by a player’s opponent while they’re on the ice. I track how many cycles a player “breaks” in some way, but have no context for just how often the player is hemmed in their own zone, which could be key context to understand. Outside of that, I’m happy with where things are, especially with my 2024 draft data, and I feel somewhat confident that I have enough players in the database with enough games tracked that we can start looking at some that correlate to greater success on the ice.
First off, some caveats… I track players from all over the world, and generally draw the line at top division junior leagues in the major European hockey nations (Sweden, Finland, Russia, and Switzerland) and the two premier Junior A leagues in Canada (BCHL/AJHL). There has been one exception in Scott Morrow at the US Prep School level who had a very near perfect dataset. Some youngsters are playing pro hockey for 8 minutes a night, others are playing 20 minutes a night in the WHL, so there is context that may colour things in some way. That said, good results are good results, and hockey is still the same game no matter where it’s played. The rink size may change, but otherwise the fundamentals are the same. My data is also all tracked at 5v5 only. Most of the game is played at 5v5, and good 5v5 players can generally be analyzed to project as useful in other game states. I also try to track games that are played against strong competition in their leagues, especially at the junior level, and quadruply so in Russia. Lastly, some of these players have small sample sizes in their individual datasets for one reason or another, but every player’s data has been dumped into a pool of individual performances looking at various correlates between tracked metrics and shot attempt differentials, dangerous shot attempt differentials, and shot attempt rates at both ends of the ice (offensive and defensive). The general focus is to isolate factors that drive good results in a hockey game, and cutting through whatever biases there might be analyzing of the game in the context of the NHL Draft. We’ll also be avoiding correlating between raw shot totals to shot attempt differentials unless they’re worth noting in general as they are obviously major drivers of themselves in some way and we’re more interested in the other tracked data’s effect on those differentials. We will also be keeping this relatively simple, calculating Pearson Product-Moment Correlations (r-value). So with that out of the way, let’s take a look and see what pops out!
Takeaway #1 - Defensive Results Are Noise… Except One Thing…
I gave a quick presentation about this topic a few years back at the Ottawa Hockey Analytics Conference during the pandemic, and a few years later, the same trend holds. The impact of defensive metrics on overall shot attempt differentials is basically noise with very little correlation, but there is one area that is a decent correlate, and that’s allowing control of the puck across bluelines if you’re a defenseman. It doesn’t matter what blueline we’re looking at; the worse you are at preventing zone transitions, the worse you’re going to be in your own end. This may seem obvious, but I see defenders struggle in this area and have concern waved away because of physical play in their own end or how they clear the front of the net, when in reality, the battle is always made more difficult when getting into that situation in the first place. I see players with a knack for being aggressive and using their feet and stick to challenge carriers and pucks in the neutral zone and they often go unheralded or labeled concerning because of a perceived lack of physical tools. This trend is a big reason why I’m so forthright about looking more closely at highly mobile, smaller, skilled defenders, as they often take a more aggressive approach off the puck to challenge play higher up the ice, and the results often speak for themselves.
When isolating specifically for shot attempts against, there are some interesting takeaways that feed into a similar story. Uncontrolled offensive zone exits and transition involvement at the offensive blueline going either direction are correlate on a similar level when it comes to suppressing defensive shot attempts. In fact, defensive zone entries, even those that are uncontrolled correlate to increased rates of defensive zone shot rates albeit at a very weak level. It appears that preventing defensive zone transition control, especially at the offensive blueline should be the focus in order to get strong defensive results.
For forwards? It’s a bit of a mess and there wasn’t much that factored into preventing chances defensively that are reliable enough to get into.
Some of the key points of interest that improve shot attempts against (SATA) and shot differentials (SAT%/DSAT%) among defenders.
OZExitUnc - Uncontrolled offensive zone exits | OBLT - Total offensive blueline transitions | DTS% - Defensive transition success percentage
Note: DTS% is a the opponent’s success rate on their offensive transitions against the subject. Lower is better, so as DTS% increases (bad), SATA increases and SAT% decreases (bad), leading to the correlations displayed.
Takeaway #2 - Offense > Defense
If anything is clear when looking at the data, it’s that offensive output drives overall results at a much more consistent and reliable rate than defensive suppression does. As we found above, even after adding multiple new metrics to track defensive play, nothing makes much of any difference except defensemen being aggressive on defensive rushes. There are positive correlations just as strong for forwards when it comes to attempted slot passes, completed slot passes, total completed passes, shot assists, and offensive zone turnovers generated. Transition success is a much weaker correlate and transition involvement has no correlation. For forwards, the balance of the game is so heavily weighed towards the offensive zone that the rest is a secondary concern on paper.
Switching over to defensemen, things are a little bit more mixed, but the importance of the offensive blueline holds true once again as a good correlate with all shot attempts as well as when we filter out low danger opportunities. It appears that defensive transition play affects defensive results reliably, but at the other end, involvement in the offensive zone is paramount, and individual low danger attempts from the perimeter is a poor correlate when compared to slot passing and stepping up from the offensive blueline to make plays.
There are so many offensive metrics that have such an impact on the overall balance of shot attempts in a game of hockey and so few defensive ones that over the years, I’ve simply leaned more and more into valuing players that bring offensive upside and drive strong results in that area more and more.
A selection of offensive and defensive metrics for forwards and their impact on shot metrics. Note how many offensive metrics have correlations that boost overall results while defensive metrics are significantly weaker.
iDSAT - Dangerous shot attempts taken by player | DShA - Dangerous shot assists | DPass - Slot passes
CB - Defensive zone cycles broken
Takeaway #3 - Stop Shooting From Nowheresville!
Another one that might seem obvious, but is interesting to see presented across multiple leagues and levels of play is the trend that connects volume shooters and questionable offensive results and projection. As indicated, defensemen drive better results when challenging opponents in the offensive zone rather than peppering the net from the perimeter and hoping for chaos to reign supreme in front of the net. The same is true for forwards by an even larger margin. If we focus on shot differentials from dangerous areas of the ice, how much a player shoots from the perimeter is basically a non-factor. In fact, when zeroing in on some high end prospects whose shot attempts come from the perimeter more often than not, we see some interesting names pop up: Alexander Holtz, Joakim Kemell, Jacob Perreault, Zachary Bolduc, Sasha Pastujov, and Matthew Beniers are in there. Granted, some are better off than others at current time and there is much more to their overall story (especially for Beniers), but these are largely productive, high offense players often coveted in the NHL Draft. Only Pastujov was a pick outside the first round, but many outlets had him just outside that range at the end of the season. All of these players have had some level of struggle projecting their goal scoring/production to the professional levels of hockey, but of course they have time left to continue to improve and carve out a role if they haven’t already. Trevor Connelly, Dalibor Dvorsky, and Terik Parascak are notable examples from the last couple of drafts who fall into the same category, so keeping an eye on that will be interesting in the near future.
Highlighting the importance of shot selection, weighing the subject’s shot attempts by location (iwSAT) to favor dangerous attempts correlates well to offensive results. Low danger attempts correlate to team dangerous shot attempts for as well, but much weaker than stepping deeper into the offensive zone. There is also more evidence of stronger correlations to improved shot differentials from offensive data than defensive data.
Takeaway #4 - Offensive Zone Forechecking Matters
No matter which way you slice it, forwards who create turnovers in the offensive zone in some fashion leads to better shot differentials. While the correlations on the defensive side are quite weak, when looking at offense and overall differentials, there is a decent correlation across the board. Sitting back on a forecheck is a recipe for limiting your offense, which may seem obvious, but we can hear criticism of a 2-1-2 forecheck due to the defensive holes it may create. Combining the results that favour generating turnovers in the offensive zone and defensemen being aggressive at the offensive blueline, we start to see a picture that more aggressive teams and players at all positions not only drive better offensive results, but may not affect defensive results on paper, and have a positive impact on shot differentials, even when removing perimeter shot attempts at both ends. When you think about it from a strategic point of view, this all makes a level of logical sense. Getting a breakout organized and moving up the ice is a pivotal skill to nail down for any team, but it can be a chaotic process no matter how good your players are. Once players get a head of steam going, the defensive approach becomes gap management and guiding carriers to the perimeter, which can be difficult against faster, skilled players who can see the ice well, read options and get through defensive layers in a more controlled environment. In many cases, it seems to pay off to be risky and more aggressive than it does to sit back and let the game come to you. This is another area in my work that has remained important and reflected in my rankings, but I will also admit that tracking offensive zone turnovers is an addition to my 2023 tracking and the sample isn’t as big, but some names to keep an eye on here? Teddy Stiga, Brodie Ziemer, Ilya Pautov, Hiroki Gojsic, Berkly Catton, Will Smith, Zach Benson, Konsta Helenius, Jacob Battaglia, Ollie Josephson and Adam Fantilli have all been standouts. Others have gone undrafted but are on my radar as they age such as Theo Kiss, Griffin Erdman, Isac Hedqvist, and Christopher Thibodeau who have all been mixed into the previous group.
Some key forechecking and aggression metrics for forwards and defenders. Increased offensive zone turnover generation (OZTo) for forwards is among the stronger correlates for driving good results at both ends, especially offensively. Combining with the correlations for aggressive offensive blueline defenders makes the case for increased risk for greater reward.
Takeaway #5 - Good Differentials ≠ Good Player
This one is a little more anecdotal and dips more into the traditional hockey world, but over the years you absolutely pick up on details and facets of the game in of the NHL that changes your philosophy on players outside of the NHL. There have been players who drive great results in some way in my experience, but struggle to project to higher levels, let alone the NHL. Yes, in the data there are little details and quirks that can raise questions, but some players drive great results, even in good leagues, but the confidence that that trend continues may not be there for a variety of reasons. The same can be said about the opposite case. Having a firm grasp of the NHL’s speed, physicality, intensity and seeing how the best of the best be that way at the highest levels is paramount to avoiding errors and relying on your data too much. For example, the some leading forwards among big sample players in dangerous shot attempt differentials I’ve tracked over the years? Lorenzo Canonica, Jack Devine, Logan Stankoven, Teddy Stiga, Joel Jonsson, Aydar Suniev and Lucas Raymond. Some of these players went undrafted, one in the first round, two in the second and one in the seventh. Why? Canonica is a great example. He’s objectively small but vitally lacked the elite skill and speed necessary to escape or navigate through defensive pressure at his size, even at the QMJHL level to consistently drive actual offensive results. In my view, Stankoven, Stiga and at to a lesser extent Jonsson buck that trend. Raymond was simply a very talented two-way player in the SHL and absolutely deserved the elite status he had in 2020.
On the flipside, Brad Lambert has had the second worst shot differential of anyone I’ve tracked, but I remained steadfast that there was something there based on other metrics and his video evidence, and his career in the time since has been significantly more promising. 2024 first round pick Stian Solberg was also awful when it came to shot differentials, and you can see why in the video for a variety of reasons, but does that make him a reach or bad first round pick? To me, no! I love that guy. Does it mean he’s an NHLer this year or next year? Highly unlikely, but it goes to show how important it is to look at every detail in aggregate and incorporate what you’re seeing as you gather information, how those things intertwine and what you may be able to extract from that information to make better bets on players. It’s funny to say, but being good doesn’t always mean you’re good, let alone in the NHL. The more I watch the NHL, the more I worry about raw pace, intensity, forechecking pressure, and quick thinking, and that’s become a prime area of interest for me, with the data being an informative backbone to that strategy.
Final Takeaway - This isn’t Magic, but You Can’t Live Without It
You may note that a lot of the correlations here are not definitively strong, and we’ve been over some of the caveats in the collection process that might lead to that situation. When I started this project, I set out to really dig into what findings could be gathered from looking more closely at performance data in the draft and how it can be used to improve strategy and output. After many years of this, I’ve learned simultaneously that nothing in my work “solves the draft” or creates a foolproof approach to never miss a draft pick ever again, but I also am strongly of the belief that any scout in the NHL needs to be data-literate in 2024 with a firm grasp of performance data, how it’s gathered and presented to check what their eyes are seeing at the very least. A major source of frustration to me is the “you have to be in the rinks” line of thinking when after so many years from this in my office, I’ve been able to go around the world, watch individual players, see their performance, and go through anywhere from 4-10 players in a single day without having to get in the car, get on a plane, find parking at an arena, and wait between shifts. The efficiency of this process needs to go hand in hand with the networking, research and intelligence that the scouting world also does require. The raw numbers in the database don’t necessarily predict the future, but it helps avoid mistakes when branding and analyzing a player. I can’t tell you how many times I’ve sat back and seen notes from others where I think back to my video records and tracked data and think “If this player does this, then why are they failing at it so many times? Why can they not turn that into better results?” Similarly, if I see criticism of a player, I can easily see cases where the same criticism isn’t levied against someone else, or I can easily see if there is evidence that the criticism isn’t really a huge factor in the grand scheme of a game. You gain a firm understanding of what exactly it is that you’re seeing in a player and how it lines up with what you’re looking for.
Finding out what a player brings and doesn’t to a game in a concrete way is arguably a better question to ask than “is this player any good”, and over the years that has become the major focus of my work. A lot of the hockey operations world seems to function like a corporation without a finance department that is a focal point for business development. Sure, you may have some analysts on staff, but if the business development and executive team run roughshod regardless of what the finance team is presenting, the risk of making poor decisions significantly goes up, and for a long time many organizations didn’t even have a finance department in that example. Coming from someone who worked in the private sector in a situation like this as an analyst, I can tell you that mistakes are easy to make if you’re uninformed, and if anything, data is great for informing you of what you’re looking at. Decisions still need to be made, decisions are difficult, and decisions can still be mistakes (see some of my Team Scouching selections and you’ll definitely see some), but the whole idea is to be better in aggregate from a bird’s eye view. In my view, having this information combined with an eye that has seen the evidence behind where the information comes, and being able to do so on a global, agile and fluid crossover basis as a part of your strategy is an advantage, especially the further on that the draft goes. At the very least it puts you in a better position to understand the marketplace, who brings what, and how to ensure your organization stays oriented firmly in the right direction, even if there isn’t a magic formula that alleviates you of all risk and maximizes output from the draft.
Hockey is a crazy, nuanced, fast and bizarre sport with a drafting and development process that is just as crazy and bizarre. The NHL almost entirely drafts teenagers, and they can come from all over the world playing at wildly different levels of competition. While it is true on paper that past a certain point the draft becomes a crapshoot but in my experience, it is also increasingly difficult to ignore that there are worthwhile bets that bring strong current value that could grow to buck the trend of the NHL Draft being that crapshoot. Tracked data from performances may not be a golden ticket with plenty of context and nuance that change the story, but it is clearly helpful for sorting through what’s out there, clearly identifying factors you’re looking for, and hedging your bets as much as possible. When seeing factors that lead to greater success, we can also spot some areas that may not have been exploited as much as they could be in the draft, especially when keeping vital factors like level of competition in mind. At the end of the day, a big lesson I’ve learned is that there’s no one way to play the most effective hockey, and so much comes down to the evidence that drives the data being a major deciding factor. That said, there are important takeaways of where to focus your attention such as valuing offensive metrics and aggressive off-puck success.
I’ve said many times that the longer I do this the less I trust my data and the data of others to do the work for me, but I can’t live without it and feel nearly as confident in my analysis of a player. Knowing exactly what it is a player brings to the game and doesn’t is vital, and cutting through whatever conscious or unconscious bias you may have could be the make or break between really nailing those pivotal first few draft picks and spending piles of money developing players into players that are hopefully as effective as those taken later in the draft. In my mind, a 21st century scouting department requires a team of both data literate crossover analysts building an international database and awareness of the entire landscape, with the road warriors doing the work those analysts can’t, acting as the intelligence arm of the team gathering key information and building a more rounded picture of the players you’re examining. Too much of one or the other, and you’re missing out on key information that could be a huge benefit, and with the NHL Draft being more a more valuable tool for player acquisition than ever, any potential advantage should be a welcomed one.
If you enjoyed this, consider supporting our work using the link below to subscribe to Scouching to get access to exclusive data tools and content, and fund our work going into future drafts!
