WPAR
What follows is my attempt to create an aggregate hockey statistic using “well known” metrics and linear regression.
I feel combining multiple aspects of the game into one number can be insightful and extremely useful (and fun). Additionally, I am a strong proponent of replacement level as a theory in sports statistics.
To start, I’ll borrow some methods from baseball and create a “points above average” metric to measure a player’s points above (or below) average for each season.
This “above average” method will provide the framework for this entire model.
Using the average rates, we can perform the “above average” calculation for each component for each player.
Replacement level is the performance we would expect to see from a player a team could easily sign or call up to “replace” or fill a vacancy. In theory it is the lowest tier NHL player.
Unsurprisingly, the performance of league minimum players is pretty bad – we can see an increase in performance as the salary increases.
We can now add in “replacement points” to each player’s wPAA to arrive at Weighted Points Above Replacement (wPAR).
Replacement level for qualified players (estimated top 13 forwards and 7 defensemen for each team) is ~22nd percentile for forwards and ~32nd percentile for defensemen.
For all qualified players, replacement level is around 26th percentile.
There are almost twice as many forwards as defensemen so it would make sense that replacement level for defensemen would be higher than forwards.
My initial goal with this project was to create a relatively simple replacement level aggregate statistic using well-known metrics
skater defense is incredibly difficult to measure using linear regression techniques and well-known metrics. While an aspect of skater defense is captured I was, not able to measure everything. I won’t say that defense is completely absent, but as of right now the final number is rather offense-intensive – I feel viewing this metric as a “weighted point” total above replacement is appropriate.
Commentaires