Panda Bear
Registered User
- Apr 2, 2010
- 6,572
- 5,703
I've been playing fantasy mock drafts for a few weeks in a row, and I wound up developing a pretty robust spreadsheet for ranking players.
As a test, I wound up comparing the different rankings categories to actual league standings for non-overtime win percentage and goal differential. After adjusting the rankings to those coefficients, I fiddled around with player weights and so on and so on.
PandaRankings has an r-value of 0.854 with actual league standings, which I think is impressive given how xGoals, Dangerous Fenwick and Score-Adjusted Corsi are involved in scoring. xGoals, in particular Moneypuck's Shooting Talent-Adjusted xGoals and created xGoals, feature prominently. This means that I hit an r-value of 0.854 featuring a battery of statistics that have better predictive power for future goals than actual production.
While it's not like introducing a new statistic or anything amazing like that, I'm pretty happy with how it's developed so far.