In case you were wondering “how do they make that chart where one team has a 99% chance of winning” right before they lose?
www.pro-football-reference.com
In that model, it starts with the Vegas spread; for us against NCST this week, we had “credit” for an 8.5 point lead before kickoff. From there, the formula adjusts the odds based on score and how much time is left in the game, then adjusts it again based on expected points based on down and distance and position.
It uses a normal distribution, which works well over a ton of games. Maybe a distribution with faster tails would work better for individual games.
Last night’s 30-29 game was nuts