Corsi 6 - Relative Shot Metrics
- tmlblueandwhite
- Dec 24, 2022
- 7 min read
Updated: Dec 27, 2022
INTRODUCTION TO RELATIVE SHOT METRICS
CF% and FF% are useful in comparing teams and players across the league and even game by game. Many fans get an idea from watching the game as to which forward line or defensive pairing had the best performance of the night.
Comparing players and line combinations on a team are aided by adjusting the CF% and FF% values to make them Relative. CF% Relative and FF% Relative allow us to see how a player stacks up against his teammates.
Relative values tell us how the team performs when the player is on the ice. If the value is positive, the team performed better (more shot attempts/higher possession numbers) with that player on the ice than when he was off of it. If the value is negative, the team performed better when the player was off of the ice as opposed to on it.
Relative Percentages do not mean that certain players are good and others are not good. There are many factors that affect these numbers such as Quality of Competition and Zone Starts and so really what we are looking for is strength of performance.
We can use this information to determine where the strengths and weaknesses of the team are located. If the team’s checking line consistently has better Relative numbers than the offensively gifted second line, perhaps the usage and deployment of the line needs to be revisited. Further, large disparities between a team’s lines may indicate a team with less forward depth or heavily front loaded lines.
The distribution of a team’s Relative possession numbers is heavily dependent upon not only depth of talent but also the Usage and Deployment of the players.
PROBLEMS WITH CORSIREL
Corsi Rel is a flawed statistic. That doesn’t mean we should reject it outright, but it does mean that we have to be careful when using it.
Times CorsiRel can fail:
· A poor territorial team
· A team with lots of injuries
· An excellent team
· Quality of competition
I’m certainly not smart enough to synthesize Zone Start, Quality of Teammates and Competition, as well as Corsi, into one statistic that would comprehensively define territorial play. I’m hoping for that day soon. Until that day comes, we will be stuck with Corsi Rel – it’s far from perfect, but in some ways it’s still the best we’ve got.
IMPACT OF CHANGING TEAMS ON POSSESSION
Certainly a player’s Corsi or Corsi For % can be a reflection of their team and their teammates. However, the point of CorsiRel is to differentiate between how the team does when the player is on and off the ice. This is a form of without and with you or WOWY as it is commonly known.
I’m going to look at how year to year correlations change for players who are on the same team both years compared to players on different teams for both years. If the correlation or the ability to predict how a player does in the future is substantially decreased by changing teams, then that metric is not isolating individual player talent.
For the most part CorsiRel and THoR (all events) are primarily player based metrics rather than team based metrics. It’s reasonable to expect that there will be about a 10% drop in correlation for players that switch teams for CorsiRel, while for THoR that number is less than one-half of one percent.
We can say that about 90% of the correlation in CorsiRel does not depend upon a player’s team.
ADJUSTING FOR TEAM TALENT
Corsi Rel is a stat that, in theory at least, is meant to address the fact that a good player on a poor team is still likely to post a bad CF%. We don’t want to punish superstars who are surrounded by replacement level players in the same way that we don’t want to reward hangers on playing on Cup winners (*cough* Dave Bolland *cough*). For defencemen in particular, Corsi Rel is often a better way to measure their impact, given that they have much less control over play in general and are driven heavily (at least in terms of raw results) by the talent up front that they’re paired with.
The problem with Corsi Rel, however, is that it’s too blunt of an instrument – it assumes that each player can only affect his team’s results by a set amount, regardless of the talent of that team. A good player on a bad team is assumed to be a good player on any team he plays on, which we know is unlikely to be true in practice. A player with a +1% Corsi Rel on a 42% team is unlikely to make a 56% team into a 57% squad, but pure Corsi Rel assumes that this would be the case. So while we know that there’s value in the information that Corsi Rel contains, the question is how to maximize that value.
One way we can do this is to create a function that modifies Corsi Rel based on quality of a player’s teammates:
Adjusted Corsi Rel = f(TMCF%) + Corsi Rel
The point is that raw Rel stats can give us strange results for players on extreme clubs, and that by making simple adjustments such as the one’s outlined here we should be able to make better estimates for those players.
RELATIVE SHOT METRICS
All relative shot metrics whether it be WOWY, relative to team (Rel Team), or relative to teammate (Rel TM) are essentially trying to answer the same question: how well did any given player perform relative to that player’s teammates?
Teammate combinations with similarly high time on ice (TOI) together will, more or less, have similar on-ice numbers whether it be goals/shots/etc. As a way to evaluate lines/combinations, on-ice numbers are often fine. But what if we want to try and separate Cheechoo from Thornton? What would that look like? How might we go about that? This is where the idea of relative shot metrics came from and why they were popularized.
Relative To Team:
It’s relatively simple: how did a player’s team perform when they were on the ice vs. when they were off the ice?
We often see this calculation in a differential or percentage form (Rel %, Rel Diff, etc)
In general, we get a better idea of how an individual player performed because we’re looking, specifically, at how a player’s team performed when they were on the ice vs. how their team performed when they were off the ice.
However, this method doesn’t really address the teammate issue that is present in on-ice CF%. If a player played a large percentage of his time on a good line (or with the best teammates on his team), his respective relative to team metric will be significantly influenced by those specific teammates. And often, we’re right back to where we started. How much better was one player compared to another if they had similar on-ice and off-ice numbers? If a player plays a large percentage of their TOI with the best or worst player(s) on their team, that player’s on/off-ice numbers will be influenced in a way that makes player evaluation difficult.
Overall, the relative to team approach (on ice – off ice) is generally an improvement over using on-ice numbers for player evaluation as it gets closer to isolating a single player’s contribution compared to their respective team.
Wowy
The idea behind WOWY is simple: how did a given player perform WITH each teammate and how did they perform WITHOUT each teammate. This is often used alongside on-ice CF% as an addendum
The theory of WOWY is:
1.) if a player’s teammates generally perform better with him than apart from him that’s good,
2.) if a player’s teammates generally perform worse with him than apart from him that’s bad.
While WOWY’s are useful and can give us a deeper understanding of a player’s abilities based on how they performed with or without certain teammates, there are some issues with this approach. First of all, each player plays with more than one teammate, so splitting every player-teammate pair into separate observations can be misleading and can sometimes oversimplify analysis. Additionally, varying amounts of ice-time with or without a player can lead to somewhat subjective interpretations. Each player will have varying amounts of time apart from one another – where should the cutoff be? Which teammate is most important?
WOWYs are rather difficult to use outside of situational analysis. They can be great for looking at how lines are constructed, or how a given player impacted his teammates (who’s a drag, etc.), but the output is rather cumbersome – WOWYs are not meant to be “summed”. The fact that we need to look at all pairs of players at the same time introduces the possibility of misinterpretation, and comparison between players is often clumsy and difficult to set up.
Relative To Teammate:
The idea behind this method is a combination of the prior two I discussed above. Relative to team gives us how a player performed relative to his aggregate team, and WOWY’s give us a big table of player-pairs that can be examined and visualized. Relative to teammate (Rel TM) combines both methods into one number.
Broadly, the final number is attempting to remove what a player’s teammates did in the time they were away from that player. The theory, at least in my view, is that a player’s on-ice number is made up of what that player did plus what their “aggregate” teammates did. To find how much of a player’s on-ice CF60 can be attributed to them and how much can be attributed to their teammates, we need to find a way to measure the performance of a player’s “aggregate” teammates. This is achieved by measuring what each teammate did without that player.
To determine teammate strength, we want to remove as much of the respective player’s impact as possible – hence why we are looking at teammate performance without that player. And finally, we’re weighting the “aggregate” teammate number based on the percentage a given teammate played with a given player. A player’s on-ice CF60 will consist of a higher percentage of contributions from the teammates that played the most with a given player. To account for this, the method uses a weighted average. At the core, Rel TM revolves around the WOWY player pairs. More specifically, we need to find every teammate a given player played with during a given length of time for every player.
Problems:
All relative to teammate methods inevitably suffer from “multicollinearity” when players spend a large amount of time together.
Even during a full season, pairs of players often play so much time together (90%+) that their relative numbers will be heavily influenced by what one teammate did in a relatively small number of minutes away from that player. This problem becomes apparent with the way the weighted average is calculated (a player’s teammates who played the most with that player are given the most weight in the calculation).
Team Effects:
Team strength has an impact on the final Rel TM number.
Specifically, players on the worst teams appear better and players on the best teams appear worse relative to the league.
Comments