02-11-2022, 03:59 PM
(02-11-2022, 03:49 PM)JaggedJimmyJay Wrote: I think statements like this are just incorrect. I mean no offense to you, KillerGoose, or even to the institution of predictive sports modeling. There is simply insufficient information to make this claim, and models based upon season-long body of work data that make these claims are operating from a fundamentally goofy premise in the first place. The Rams might indeed have the "better roster" and be the "better team", and that still does not imply that they have a 50%+ chance to win the game.
We need more deterministic models.
I've heard the insufficient data argument before, but I don't necessarily agree. Is it as accurate as baseball? Of course not, but it is still there. There are 1000+ plays for each team during a season and a wide variety of metrics to choose from to get your overall WP prediction, and this data is aggregated and used over the course of seasons. This is what Vegas does, and their lines certainly do have a correlation to victories. Could we improve the models? Certainly, and I would expect them to only improve as time goes on and more study is put into them. Hell, something as simple as Pythag projections are pretty damn accurate, even in football. The Pythag projection had Cincinnati pegged for 11 wins around week 4-5. It was a limited dataset, so you have to be skeptical but once it hit week 8 and still had Cincinnati pegged for 11 wins, you start feeling good about it. Then, you have SRS. SRS is a pretty good indicator of who the better team is and is a decent predictor of victories in H2H scenarios.
Vegas likely still has the best model for straight up W-L record, I haven't checked them against any other models but by virtue of the Rams being favored, they have a 50+% WP. Using the NFLFastR model that I have, a spread of -4.5 is about a 77% chance of victory. How accurate that particular number is, I am not sure but I feel very confident in saying that the Rams have a > 50% chance of victory.
