The problem with analytics based decision making is that it is limited to exactly one metric to guide decision making: “expected” points. In plain English we could say “how many points would I expect to score, on average, if I made this decision 1000000 times”. This value is going to be primarily based on data from other teams, with other players, and other coaches, in other places.
Well, first of all, we don’t get to score 3.7 or 4.1 points. We get 7, 3, or 0 as others have already point out. And we don’t get to retry the same decision a billion times and approach the statistical average outcome. We get one chance.
It also completely ignores very important factors like:
- The psychological effects of a goal line stand on both teams.
- The pressure applied to the opponent when the lead is narrowed. They may be forced to throw the ball where they otherwise would run, which increases the likelihood of a sack or turnover and may also preserve clock in the case of incompletions.
- Each team’s particular strength or, ahem, weakness near the goal line.
How did we lose to OSU when we had more passing yards, more rushing yards, greater TOP, more total plays, etc? Because we didn’t lose the turnover battle 1-0. We lost 5-1. A turnover on downs is absolutely back breaking, especially when it takes points off the board.
Take the fucking points on the road against a top 5 team with a QB making his first real start who is clearly overwhelmed. Let the team build even a little bit of confidence and momentum. No one gives a fuck about what the binder says.