Monday, June 09, 2003

Sports Stats

Jane Galt's post on surveys touched on an important issue about an important matter-- opinion surveys often drive news coverage and public policy. I want to discuss another area where statistics are often misused but without the serious consequences-- sports.

Sports commentators love to toss around numbers. Maybe they are just in touch with their inner geek. Or maybe they just need something to talk about to fill time.

Some numbers are laughable to anyone who has had even one basic statistics class.

"The team is 13-1 with X rushes for 100 yards or more."

They imply causation. If the team just gets X his 100 yard game, then they should be on easy street. What they leave out are the other factors which produce the number. Like the fact that if you fall way behind early, you have to drop the running game while playing catch-up. So the lack of 100 yards didn't produce the defeat, it was the weak pass defense or an error-prone QB that really lost the game. The same thing happens in victory. Get up by ten and you run the ball alot and will rack up more yards. The running helped, but it was not the key to victory.

Other times they toss out stats based on absurd sample sizes-- "Pittsburgh is 3-1 in their last four home games played in the rain." Swap the Bengals for New England and maybe the Steelers are only 2-2. So that stat is tells you next to nothing

Other stats have more plausibility but that is not the same as validity. (Just looking around, the idea that the Earth is basically flat seems plausible enough).

I like the idea that you evaluate a football coach's tactical skills by his record in games decided by 3 or less points. Joe Gibbs always did well on this measure. And it is plausible. When the game is close, coaching decisions are more important than they are when Houston plays the Raiders. Maybe a great tactician gets a field goal where the average coach doesn't. Or a TD where the other guy's play calling ends up with field goal.

The hallmark of Chuck Noll's Super Bowl Years (he won 4 in 6 years, something no coach has equaled since the merger), was the Steelers outstanding record against losing teams(59-1 between 1972 and 1979.) The Steelers beat the teams they were supposed to beat, which some argue was a measure of Noll's ability to prepare his team and keep them focused when they might have taken a team lightly.

Maybe. But so much coaching genius seems transitory and non-transferable. Noll stopped looking like a genius when Bradshaw, Harris, Greene, et. al. retired. George Seifert was almost unbeatable in San Francisco. In Carolina it was a much different matter. Who would have though Steve Yound made that much difference.

Baseball, of course, is the perfect sport for stats geeks. It started out with a boatload of stats and has been adding more every year. But even in baseball not every thing is reducible to quantification (especially fielding). Also, when you have a bunch of stats you run into the problem of weighting all the various stats in order to make comparisons.

And that is the goal: to get everything reduced to a single number that represents a player's "value".

But in trying to do so we often overlook the dialectical nature of the competition. I've mentioned before the observation of Capt. Wayne P. Hughs (a Naval officer) that

"Yet one set of Dupuy's data [on land combat throughout history] shows that in modern battle a greater percentage of casualties has sometimes been inflicted by other than the most capable weapons: infantry small arms exceeded artillery in producing casualties after the range and lethality of artillery rose dramatically. Often the second-best weapons performs better because the enemy, at great cost in offensive effectiveness, takes extraordinary measures to survive the best weapon"

An opposing manager said that Roberto Clemente could win a game without making a put-out, getting an assist, or getting a hit. His defensive skills (especially his arm) were such that baserunners became more conservative and teams lost runs because of it. But that would not show up in conventional fielding statistics.

In football the best cornerbacks don't lead the league in interceptions. They are so good that teams don't even try to throw against them. Alvin Harper had a lot of big plays for Dallas, not because he was almost as good a receiver as Michael Irvin but because Irvin was so good that teams focused on him and let Harper have his opportunities.

(Anyone who knows me knows how much writing that last paragraph hurt.)

We like numbers because they seem concrete and give us something to talk about. The more complex dynamics of the game resist analysis and are frequently ignored. But that doesn't mean they don't exist.

No comments: