MGoBlog has moved. The new site can be found at

Friday, March 11, 2005

David Sklansky is a well-known poker author. He runs Two Plus Two publishing, which prints a free internet mag. In the latest issue he addresses a topic I've gone over before: when to go for it. If you were interested in the Romer paper I linked to before but math makes your brain hurt, Sklansky provides a good basic summary of the principles behind it.

To summarize: football coaches are too conservative, often choosing the 'safe' play over the 'risky' one, even if the risky play is statistically far superior. The article also spawned an interesting thread on their forums. Serious gamblers have a state of mind that makes football coaches' decisions incomprehensible. In their opinions, positive expectation (i.e., how 'good' something is for you) is always more important than variance (i.e., how likely it is that something very good or very bad happens instead of something in between). Now, this may not be the case in an unevenly matched situation. If you are the heavily favored team you should probably adopt the lower variance strategy, expecting that you'll increase your chance of winning by lowering the variance more than you would by taking a marginally positive expectation.

Lloydball is thus somewhat justified by this point of view... if Lloydball actually reduces variance. I submit that it's not necessarily that clear. Running and eating the clock increases variance as you're reducing the number of plays/drives in the game and thus reducing the chance that your assumed superiority will show on the scoreboard by the time the game is over. That's akin to watching Michigan play Illinois in Crisler and seeing Illinois run every shot clock under ten seconds before getting into their offense. If you are a heavy favorite and your offense is equally effective running and passing, you should be playing to lengthen the game by getting in and out of the huddle quickly, throwing frequently, and getting out of bounds. (This obviously only applies to strategy early in the game. The particular situation you find yourself in drives decision making later.) Punting, on the other hand, is always the low variance play. The opponent will likely have the ball about 30 or 40 yards back from your current position on the field. Going for it has two very different outcomes: you've got it or they do. It's high variance.

That doesn't necessarily mean that a favored team should always pick the low variance play, it just means that they should have a higher expectation threshold than an underdog. A perfect example of this was a game a few years ago between a very down Alabama squad and #1 Oklahoma. 'Bama started the game off with an onside kick which they recovered. They then proceeded to scare the pants off the Sooners before some scatback named Works pulled OU's butt out of the fire. The decision to onside kick it was brilliant for Dennis Franchione, Alabama's coach at the time, but would have been a blunder had Bob Stoops tried it.

In conclusion, Lloyd should hire Doyle Brunson.