Monday, June 01, 2009

Optimal Sports Decisions

I hesitate to say that economic analysis has shed new light on decisions in the sports realm, but I will say that increased statistical analysis has had that effect.

The analysis in baseball has been perhaps most visible. The Bill James-led sabermetrics revolution has brought on-base percentage and OPS (on-base percentage + slugging percentage) into everyday conversation. Moneyball played a large role in making this well known, but the process started prior to its 2004 release. Typical baseball strategy involved bunting and stealing bases; sabermetrics highlighted the deficiencies in assuming this to be the goal at all times. The game is in transition at the time, and I'm eager to see where it goes-- both because of my personal interest in the sport and its possibility for continual evolution. I think where things sit now are not where they will be sitting in a few years; I'm eager to see where it goes next. I think the bunt/steal situation hasn't been flushed out fully-- at least not for public consumption, anyway.

Football's been analyzed statistically as well, though not quite to the extent of baseball, and I'm not certain that many of the recommendations have been incorporated. For example, statistical analysis shows that teams punt too often. And while some teams may choose to go for it on 4th down a bit more often now than, say, 10 or 15 years ago, I do know the estimates for how often they should go for it are still quite diverged from the rates they actually try it.

Basketball is moving down that path of statistical analysis, too. The question here was mainly what to measure; influential (and accurate) statistical analysis depends, naturally, on substantive statistics. There have been some posts at Freakonomics (and possibly Marginal Revolution) concerning this development, and there was also a New York Times article on Shane Battier concering the general trend of more statistics in basketball. I'd expect in the next 10-15 years a similar evolution along the lines of what happened in baseball to happen in basketball-- and given the state of the NBA right now, it couldn't happen soon enough.

All of this was brought about by watching the first two Stanley Cup Finals games over the last two nights. Hockey, it seems to me, has evaded any sort of statistical revolution. I'm not sure that any of the four sports are immune to gains from new insight-- obviously-- and I don't think hockey is immune to statistics either. What advances have been made that I've missed, or what areas could a team improve in order to enhance their chances of winning?

Part of me thinks that teams don't pull their goalie early enough. For the uninitiated, when a team in hockey is losing by one or two goals, they remove their goalie from the ice for an extra attacker to try and make up the difference. (The 1- and 2-goal lead is just my observation; it seems that a 3+ goal margin leads to a team ceding the victory.) Clearly, there's a tradeoff here-- if you pull your goalie, you have an increased chance in scoring (since you have an extra offensive player on the ice), but you also increase your chance in giving up a goal (since, well, you don't have a goalie). Statistical analysis can give some indication of a superior ex ante strategy, albeit through the analysis of ex post outcomes. If your team scores without a goalie, then ex post, it was a wise decision; if your team didn't score, or gave up an additional goal, then it wasn't wise.

The timing of when teams decide to pull their goalies is of interest to me. It would seem to me that the odds of giving up a goal without a goalie as compared to the odds of scoring without a goalie would be independent of the time the team spends without the goalie in net. Given that situation, it would also imply that the odds of giving up a goal were greater than scoring one-- otherwise, teams, in the long run, would score more goals playing the entire game without a goalie. As it sits now, teams generally pull their goalies with 1 minute left in the game if they're losing by 1 goal, 2 minutes if losing by 2 goals-- is this optimal? It just seems to me that a lot of game end without an additional goal one way or the other-- which means that there's probably some marginal adjustment to be made that could improve the welfare of the losing team.

Optimal shift length? Optimal number of lines? I think there's enough market feedback to hone these over time. The goalie issue still piques my interest.


jb said...

I guess I take a Misesian approach to sports...throw out the Over-The-Top stats and enjoy the game. However, as this is yet another overly regulated part of the economy, it is appropos to strangle the fun out of it with OTT analysis.


Matt E. Ryan said...

On the contrary, the actual course of play itself is about as unregulated as it gets. Any team is free to pursue any strategy that best achieves the goal of winning.

I presume you mean regulation on the use of steroids, perhaps? Agreed, that is heading down the path of regulation, though much of that up until recently has been self-imposed-- testing schedules, for example, are agreed upon between the union and the owners. Labor market regulations, salary caps and the like, again, are self-imposed as well. Besides MLB's anti-trust exemption, I still can't think of a large scale regulatory role in sports-- though even if there were, that's still beside the fact of how to proceed on the field in winning games.

Matt E. Ryan said...

That being said, for me:

Enjoyment of sports = f(overkill by statistics)


Anonymous said...

Good day!
hi there, my name is Tom.

found this website and read some great discussion and feedback so decided to join

i am happy to help others and offer advice where possible :)

Keep smiling!