In this third perspective article, I want to dig into a somewhat unconventional way to do player analysis. I think this is a very interesting discussion to have, and hopefully you’re able to follow my thought process.
The 4 Pillars
Often you define and evaluate players through the 4 pillars: Technical, Tactical, Physical and Psychological/Mental.
This is a perfectly valid approach… But I do think some key aspects of the game are unclear with this type of analysis.
Defining players through Actions
The alternative is to define players through actions – Evaluating the player’s ability to execute plays.
Every single action can be described as a three-step process: Perception, Decision Making and Execution. I prefer to look at players through this lens. How well does a player perceive the play? Is he making smart decisions? And can he execute plays?
I think there are two types of perception: Natural Perception and Trained Perception.
Natural Perception is something you’re born with. Some players are gifted with a great sense of the ice. They appear to see things faster and better than everyone else.
Luckily, there are plenty of ways you can train your perception as well. Scanning the ice requires effort and focus, so if you’re strong technically and physically then you can focus more on perception and decision making. Being able to look up when possessing the puck (split vision) is an integral perception skill.
If you have a great tactical understanding, then you know what to look for and where to look. This allows you to zoom in on the most important things, and thus improving your perception.
Finally, you can improve your perception by communicating. When players are communicating, they don’t have to rely solely on what they see, but can listen as well.
This is typically what we refer to as Hockey IQ – How good are the on-ice decisions?
I think it’s important to understand that decisions can be either conscious or unconscious. Sometimes you’re playing on instinct and sometimes you’re consciously following a strategy.
Experience (learning from your mistakes) should hopefully help you make better decisions. Being able to anticipate plays is also incredibly helpful in your decision making.
Good decision making basically boils down to a Risk vs. Reward assessment. Does the potential reward outweigh the risk of the play? For very skilled players it’s probably worth playing a high risk / high reward style of game, whereas for others it’s better to play a low risk / low reward game. It all depends on the player’s ability to execute tough plays.
My final point on decision making is about offense vs. defense. I think it’s important to view offensive intelligence and defensive intelligence as two completely separate skills. Players can be great at one or the other or both, but I don’t think they’re connected. Instead of hockey IQ, we should probably talk about offensive IQ and defensive IQ.
The final step is execution. This is all the things we see. How fast is the player? How good is his shot? His passing? His hitting? And so on…
The decision making should of course match the player’s ability to execute plays. Sometimes we see players try to execute plays that are way to difficult for their skill level… But we also see the opposite. Players that play with way too low risk compared to their skill level. Probably because they are afraid to make mistakes.
Psychological and Physical Aspect
I don’t want to go into details… But I just want to mention that psychology and physiology obviously impact plays. A player with high confidence is more likely to play with high risk and more likely to successfully execute difficult plays. A low confidence player will likely choose low risk plays.
Likewise, at the end of a long shift when the player is tired, it will impact the decisions and execution.
I generally think perception is less affected by psychological and physical aspects than decision making and execution.
Time or Quality the Limiting Factor?
Up until this point we’ve only discussed how well a player performs in each step… But how quickly the player performs each step is equally important.
It’s not just how well you scan the ice, but also how quickly. It’s not just how good your decisions are, but also how quickly you make them. It’s not just how good your execution is, but also how quick it is. Getting a shot off quickly before the goaltender has time to set up, is often more important than raw power and accuracy.
Time in this regard is an interesting concept. The importance of time is difficult to quantify, but I like to use a Quarterback as the analogy. If the defense is Blitzing, then time is limited and executing quickly is of the utmost importance. If you’re too slow, you’ll get sacked. It’s the same in hockey. If you’re too slow you’ll get pressured or hit before you can make a play. In that case it’s almost irrelevant how good your execution is. You’re limited by time.
If time is the limiting factor, then you can work on two things:
- Executing quicker.
- Buying more time.
Going back to the quarterback analogy. He will either have to release the ball quicker or buy more time (usually by moving around in the pocket). A hockey player can likewise buy more time by moving his feet.
It’s important to know if a player needs to make better decisions or needs to make quicker decisions. It’s two different things you need to practice. Playing on instincts is faster than making conscious decisions.
When you move from one level to the next (e.g., Juniors to Pro), you will have less time – Pressure comes on you quicker. Sometimes we see players completely dominate at one level and be almost invisible at the next level. My argument is that this is often an indication that time has become the limiting factor. The player doesn’t even know what play to execute before he has lost the puck.
Some of the concepts in this article are oversimplified, but I still hope you understand my points.
In this last section of the article, I want to talk about statistical interpretation. Is it at all possible to statistically quantify the concepts above?
Shot Statistics Data
Almost all public data is shot statistics data. I recommend reading this chapter if you want to know more about shot statistics.
The problem with shot statistics data, is that it doesn’t give us much intel on perception, decision making or execution. We do get some information on shot selection (decisions) and shot execution (scoring ability)… But other than that, it’s difficult to use in this context.
Let’s instead imagine a world, where we had access to a much more detailed dataset – Passing, dump ins, dump outs, puck recoveries etc.
First of all, this would directly give us much greater intel on the player type – What decisions does the player make? Is he a “playmaker”, a “shooter”, a “puck carrier” or something else?
We would also know much more about the execution and risks the player takes.
Last year I submitted a Big Data Cup project on a dataset like this. You can read it here.
Because the data was as detailed as it was, I could determine when, where and how a possession started as well as when, where and how the possession ended. This allowed me to transform the event data into possession data.
This means that I didn’t only get information about what (decision making) and how (execution) the player did. I also got information about how long the player possessed the puck, so we even got some intel on the quickness of plays.
If this kind of data was available to the public, we could do a much deeper player analysis.
Let’s try to think even further ahead. When we talk about expected goals, we talk about the expected value of a specific shot… But what if we could estimate the value of a situation instead – Take a snapshot of a situation and estimate the chance of that specific situation turning into a goal.
Then we could estimate the expected value at the start of a possession and at the end of a possession. In other words, we would know if the puck carrier added or subtracted value through his possession. Theoretically, this kind of analysis would give us a much better and more holistic player evaluation.
This is obviously way above my paygrade, but with tracking data it should be possible to build a model that estimates expected value at all times based on player locations and puck location. This is what I would work on if I had access to tracking data and the capabilities to build a model like that.
That’s it for this perspective article. It got a bit theoretical towards the end, but I hope you enjoyed the read, nonetheless.