Quote:
Originally Posted by Dominic
Favored? You could build a team that would beat the Pats 19/20 times
Quote:
Originally Posted by 1989
I would agree. Only positions the Pats would have an edge at would be QB, TE and HC (+Slater+lolkicker). Defense would be completely in favor of the wild card allstars
I think it's actually close. PFF gradeout is an interesting way to do it, but you'd have to model individual player win shares. Instead, I roughed out a model with DVOA and Sagarin predictor. Neither of those are perfect, but I think they are good enough to get a ballpark estimate, and I can also easily get the data going back several years. I'd use Vegas power rankings in place of Sagarin if I had a good source for historical data.
I'm using DVOA to deconstruct down to the offense, defense, and special teams levels. The idea here is I'll pick the best units from the WC round and put it all back together. So I modeled the Sagarin predictor based on those components, chose the best WC units (KC O, JAX D, KC ST), and then produced a predicted Sagarin predictor (!).
One convenient thing about DVOA and Sagarin is that they work really well in a stock OLS regression model. The relationships are linear, the error terms aren't correlated with the predictors, there are no interactions or quadratics, etc. The model fit is really good, so I just have a simple model specification here and will use the parameter estimates to reassemble a Sagarin predictor for the fake team.
Don't forget the 2.44 points (Sagarin) to NE for home field advantage:
I get NE favored by less than a point by Sagarin predictor. Now we can start subbing in player upgrades to our KC-JAX squad and argue how many additional points they are worth, which will obviously be > 0, but not obviously worth more than a few points. At least not obvious to me, anyway.