I still think that people put too much weight on a simple stat such as corsi when the game of hockey is so complex. In Baseball for example you can pretty much tell the worth of a player just by looking at his stats. Hockey has just too many variables.
Thank you, I had a feeling that's what it would show.
And I bet GF60-GA60 differential between seasons was a better predictor between the seasons too. Closer to real results.
For fun, I repeated same exercise with JUST Jets players, from 2011 to present.
Two things:
1) Same relationship in how the gap between CF and GF shrinks.
Binning players by how much they differed in Corsi% and Goal% and then looking at how much they differed in year two:
CF differed GF by >10 points: average 15.9, next year 1.3
CF differed GF by 5 to 10 points: average 7.2, next year 0.1
CF differed GF by 0 to 5 points: average 2.7, next year -0.5
CF differed GF by 0 to -5 points: average -2.1, next year 1.3
CF differed GF by -5 to -10 points: average -7.1, next year 0.5
CF differed GF by <-10 points: average -24.1, next year -1.8
2) Corsi predicted next year performance better.
Corsi in year one had over twice the relationship (R^2) as Goals in year one for Goals in year two for a player. (I used % form so CF / (CF + CA) and GF / (GF + GA) instead of CF - CA since 24-22 is closer to even than 4-2).