3 ms·
Burke's piece is excellent, in as a far as it goes -- however, expected value is not the only thing one should concern themselves with. Higher moments matter.
by JackFr 13y ago
Burke's piece is excellent, in as a far as it goes -- however, expected value is not the only thing one should concern themselves with. Higher moments matter. In particular, in a given situation, when ahead a coach might choose to trade negative expected value for a reduced in variance, and when trailing do trade negative expected value for increased variance. (It looks like his calculator may account for this in some way through win probabilities, but I didn't see an explanation on the site.)
- twoodfin 13y agoHe mentions at the end of the piece that his EP-based analysis only applies in "typical" close games in normal weather when the clock is not yet a major factor. And that an analysis based on Win Probability could correct for most of these factors. I'm skeptical of that last point. There are many, many variables that underlie football outcomes. Unlike baseball, where the most critical two players by far in any situation (the pitcher and the batter) are known in advance by both teams, a football play has the potential to be affected dramatically by any combination of players on the field. I am sympathetic to the idea that more teams should play aggressively on offense more often (and indeed I believe there are signs of this in today's NFL), but by no means should every team be expected to act according to the punt/FG/go-for-it chart in most situations. I think you're absolutely right to emphasize variance tradeoffs. When coaches try to reduce variance they are often criticized as "risk-averse", but that's often a winning strategy, even in games that are close enough score-wise to be considered "typical".