Read on Twitter
Sooooo this conversion made me realize that there is a shortcoming of the bot that made it too conservative for the GB-TB case (with thanks also to @fbgchase for getting this conversation started) 1/ https://twitter.com/PFF_Eric/status/1354222707186036736
The bot has a probability that a 4th down attempt scores a touchdown.
On these plays, the bot gives the scoring team 7 points and then the other team the ball.
Most of the time that's fine. In this case, that's stupid, because GB would obviously go for two 2/
On these plays, the bot gives the scoring team 7 points and then the other team the ball.
Most of the time that's fine. In this case, that's stupid, because GB would obviously go for two 2/
Luckily, it's trivial to compute expected win probability in the going for 2 case because a 2pt calculator is already on the site. SO, to follow is my final answer for the Packers decision 3/
THE PACKERS REALLY REALLY SHOULD HAVE GONE FOR THAT FOURTH DOWN
(sorry for getting this wrong -- sort of, the recommendation was still correct -- and this is all in the spirit of being transparent and having everything be open source. Dealing with edge cases is hard)
+3.5% "go advantage" is what PFF had per @PFF_Eric, compared to +3.8% for the updated bot, so alignment feels good here.
This is an unquestionable "go" situation and was very damaging to the Packers to not go
This is an unquestionable "go" situation and was very damaging to the Packers to not go
