The general idea with RAPM is that it takes into account how a player is deployed (ice time), the game situation (score, venue) and comes with an expected value for the stat. Rather than using CF% directly, CJ opted to use Corsi Impact, which is generated from Relative Adjusted Plus Minus (RAPM) Corsi for and against. This way we filter out called up players who only played a little bit and therefore have an extraordinarily high or low Corsi value. The Approach & The Confounding FactorsĪs CJ put together a linear optimization algorithm to create an optimized roster with WAR, I asked him to do so for Corsi for all NHL players who played at least 500 minutes in the NHL in 2019-20. Play through the 2019-20 season with this Corsi-optimized roster. Replace the 2019-20 Devils roster with a NHL roster optimized for Corsi based on the 2019-20 season. I did indicate in last week’s post that I would be using 2019-20 data instead of 2018-19 to be consistent with the WAR-optimized team. This comment came in response to the WAR-optimized Team post, where they steamrolled most of the NHL in the regular season and went 16-1 in the Stanley Cup Playoffs.

How about utilizing the highest ranked Corsi players from the 2018-19 season on a team together to see how they perform when compared to a WAR optimized team? As suggested by MedicSBK from a few weeks back:

This week’s experiment is another exercise in optimizing a roster for a given stat. In general, he crushed it in the other leagues before having a tougher time in the best league in the world. Last week’s experiment saw Jack Hughes play his 2019-20 season in several other leagues before playing 2020-21 in the NHL and the results of how he did. Welcome to the fourth experiment in the Let’s Experiment in Eastside Hockey Manager (EHM) series.

Eastside hockey manager tips