Allow me to stretch your mind a little bit. Do you think that ESPN's Win Probability Model factors in who is playing Quarterback for each team?
The obvious constraint to this data, is that ESPN would ascribe a 10% win probability to Green Bay regardless if Aaron Rodger was playing or not. It's playing the odds on the situation in general. You could plug in Tim Tebow into the game and ESPN's computer formula (remember it's just a robot) would still tell you that "Teams win 10% of the time in this situation".
But you intrinsically know that changing from Aaron Rodgers to Tim Tebow in that situation would lower the winning probability. Since the formula they use does not accurately reflect the individual situation, I don't think it is a great metric of whether or not the correct decision was made.
For instance, let's say I am at the Casino playing blackjack and I am dealt a 15 and dealer is showing a 5. Under normal probabilities, I have a 54% chance of busting my hand if I take a hit. HOWEVER, let's also assume that I am counting cards and the count is significantly negative (there are more low cards than high cards left in the shoe).
Since I KNOW more high cards have been dealt than low cards, I know that my chance of BUSTING on a hit is lower than the traditional 54% the book says it is. In this situation, it may actually be a higher percentage play to take a hit even though the book (in general) says it's the lower percentage play.
Does that make sense?