The statistician in me loved that thread. Not only does it give a glimpse of how their model works, but it highlights the inherent flaws of any such model and a reminder not to trust predictions blindly.
For those wondering about the thread, he is not talking about what makes Adam Fox the hockey player unique, but rather what makes Fox's route to NHL unique in the sense that due to lack of data(*) and similar situations his model was unable to predict Fox's success. Out of the 6000 players in their data, nobody had a similar route to Fox... pretty interesting if ultimately useless knowledge.
(*) his model lacked the context of Fox's actions and lumped him in with the spare toys
That's a computational output. It's interesting but I've built more useful ones. What they are doing is trying to apply financial analysis to hockey, and it just doesn't work across the board. It's not a true hockey analysis if it doesn't understand the game. There are certain predictions that are safer bets than others, but I think there is a real lack of sophistication if you don't apply the context of the game itself versus not. I'm getting to a point where I can tell which teams have aces and which ones don't. It's pretty funny
.
There's something you can do to really jack up the corsi numbers. Three teams in the league do it and I'm 99% certain they all have one thing in common. There was one team that was confounded about something and to me it's obvious, but if you look at the publicly accepted take on analytics, it will tell you the diametric opposite.
Team level analytics are something else. Pierre was on Chiclets asking for an equation, and I already built it. The player he mentioned right after saying that is number 1 on my list. Most feared in the game. Bar none. Why does a grown man sporting the finest swag jackets cry on national TV about re-signing a player like Gallagher? He's a
true top line player. Shows up in the guy you win with category.