Discussion about this post

User's avatar
John's avatar

Maybe I misinterpreted the post, but are you suggesting analytical models on 4th down decision making don't take field position into account? It's a pretty major input for the most popular ones I'm familiar with (example: https://www.espn.com/nfl/story/_/id/39379626/nfl-analytics-models-fourth-graphics-method-decisions-punt-field-goal-go-it), since they're all based on maximizing a team's win probability. But again, maybe I misunderstood the post.

From a fan's perspective, I tend to bristle when I hear coaches talk about "trusting their instincts" because that seems like 1) an incredibly arbitrary way to make decisions, and 2) a way to deflect blame even more than deferring to "analytics". It feels like coaches are saying "I'm the expert here and I make decisions based off a unique intuition that is only known to me. And since there's no way to objectively evaluate that intuition, my decisions can't be criticized".

You're right that there are certainly nuances (strength of opponent, one's own team's particular strengths and weaknesses) that most statistical models don't capture. However, I'd argue that people are actually quite poor at evaluating risk to begin with, and are really bad at weighing the kinds of tradeoffs at play when considering what to do on 4th down. I can understand deviating from the model if you're really only considering one obvious data point (opting to kick a longer FG if you have a good kicker, for instance). But if a coach is trying to weigh multiple competing data points that are fuzzy in nature ("my team seems tired, but we have 'momentum', but the other team's run D is good on short yardage, but their offense hasn't moved the ball this half"), he's probably better off just trusting a model that is based off thousands of actual historical outcomes.

Expand full comment
1 more comment...

No posts