Quoting: ricochetii
The issue is that people treat it as definitive when it is anything but.
It provides an at-a-glance aggregation of partial statistics from already derived values.
That makes it not only incomplete but skewed.
Perhaps this is what you were looking for? https://twitter.com/JFreshHockey/status/1631673879738368000/photo/1
It's not an admission, but someone may have interpreted it that way. It basically says do not use it to define or predict a player's performance accurately.
All models vary in how well they interpret or predict data, so that doesn't mean it is necessarily broken.
Using his player cards to evaluate a player is like driving using only your rear view mirror. You have no idea what's on either side, what's in front of you, or even everything that is behind you.
The people that quote his player cards to further an argument are like that one guy at a party that's had one too many and insists he's still okay to drive.
If some people use the models as gospel that has no bearing on the models themselves. That just means someone doesn't know how to use them.
You're analogies of what the models are are totally butchered. None of that even makes sense.
What they do is deepen our understanding more broadly. You might know a player on your own team really well but how to do you compare his impact to another player in the league? Especially when you can't watch every team every night. The model can reliably be more accurate about a player than any of us individual fans of our home teams. They're far more accurate than an eye test or an opinion reliably.
You say they are biased but they are literally just data points. If you change the models inputs to fit your needs then sure that would be biased but the models don't get changed on a whim. Therefore they have basically no bias. Ironically individual opinion is rife with bias.
You say you hate it when someone quotes a player card when making an argument, how do you respond then? "Trust me bro I watch him every night"?
Thanks for the link. The fact that the highlighted part had to be highlighted I think speaks to people like
@andy_dick. He seems to think that if the models get anything at all wrong they are useless. Which is the least nuanced take possible. But ya the guy who said that about JFresh was biased clearly, and misinterpreted what was said most likely just to further his own agenda, as we all are guilty of to varying degrees. Another reason why the models help so much.