So to clarify my thoughts from another thread:
Eye test: Toews>Strome by a lot, it's not close
Numbers: Toews is just a bit better than Strome
Evolving-Hockey.com | Standard Skater Stats
How do you use hockey analytics, when it's very contradictory from the eye test?
Do you throw out the numbers and ignore them?
I'm kind of confused. Analytics are supposed to be an objective measure of the game, so when it's contradictory like this, which side do you lean on?
I tend to like things quantified, so if hockey cannot ever be completely quantified I'm a bit confused on how to use this stuff.
None of the stats in the table this link is loading for me are particularly advanced. They’re just individual event counts at 5v5.
G/60 is literally just the number of goals they scored every hour of ice time. Its really no more ‘advanced’ than goals-per-game, its just normalized to account for the fact that players play different TOI.
A/60 is just assists per 60 minutes.
A1/60 is just primary assists per 60 minutes.
A2/ is just secondary assists per 60 minutes.
They’re split because A2s are less repeatable and therefore less predictive than A1s.
P/60 is just points total per 60 minutes.
P1/60 is just points minus A2s per 60 minutes.
None of these are advanced, these are the stats hockey has been following for decades, with a little bit of grade-school math to account for differences in ice-time.
iSF/60 is 5v5 shots on goal per hour.
iFF/60 is 5v5 unblocked shots that were saved or went wide per hour.
iCF/60 is 5v5 shots that were blocked, saved or went wide
ixGF/60 is 5v5 scoring chances, basically (there’s some more math in there as far as shot distance, shot angle, shot type, etc).
The more important stats, because they’re more predictive of wins and future goal differential, are on-ice counts. That is, not just a single players output, but their impact on the team success as a whole when they’re on the ice. ‘Play driving’ for lack of a better term.
With these counts, you can measure how they impact the team, and the impact they have on the opposition.
x iG/60 is much less meaningful if that player is on the ice for x+2 GA/60.
There are also contextual stats that are used in advanced stats but not traditional stats. Zone deployment. Quality of competition. Quality of teammates. On-ice sh%. Etc.
There is no Ur-stat that sums up everything in one number. There are just better, more granular, more predictive stats.