Thursday, January 31, 2008

Super Bowl Special: Causal Analysis in Football

by Alan

Sportswriter Allen Barra has carved out a niche for himself as a contrarian, as evidenced by the title of his 1995 book, That’s not the way it was: (Almost) everything they told you about sports is wrong. Seemingly whenever he gets the chance, Barra likes to take on conventional wisdom in sports.

Viewers of the Super Bowl football game this upcoming Sunday will likely hear the announcers cite statistics purporting to show that some early development in the game may presage a victory by one team over the other. These are the kinds of pronouncements Barra loves to challenge.

One of the lines of thinking he attacked in his book was that, “You need a strong running game to win in pro football.” As noted by Barra, fans often hear a claim of the form: When running back A runs for 100 yards in a game, his team wins some high percent of the time (see this example involving the running back Ahman Green). Yet, statistical analyses by Barra and colleagues did not appear to show that the better teams rushed the ball better than did the poorer teams. Eventually, Barra reached the following conclusion:

What we finally came to discover was that football people were confusing cause with effect in regard to running the ball…Stated as simply as possible, the good teams weren’t winning so much because Tony Dorsett (or Walter Payton or Roger Craig or whoever) was rushing for 100-plus yards – the runners were getting their 100-plus yards because the teams were winning. Teams with a sizable lead in the second half have the luxury of running far more plays on the ground than their opponents; this not only allows them to avoid sacks and interceptions that could help their opponents get back into the game, it allows them to eat up the clock by keeping the ball on the ground (pp. 173-174).

Among the statistical findings presented by Barra were that, “Most playoff teams led in most of their regular season games by halftime,” and that, “Most playoff teams get as much as two-thirds of their rushing yards in the second half when they already have a lead…” (p. 174).

Several years after the publication of Barra’s book, in 2003, I attended an informal gathering of academics and sportswriters in Scottsdale, Arizona to discuss the application of statistics and research methodology to sports decision-making. Another possible case of football spuriosity that came up was the likely correlation between throwing interceptions and losing games. Were such a correlation to be confirmed, many observers would probably interpret it as the throwing of interceptions causing a team to lose (i.e., by giving the opponent good field position and/or killing one’s own drives). Following the same logic as in the above example, it could be that a team falls behind for reasons having nothing to do with interceptions, but once behind, throws a lot of risky passes in an attempt to catch up, which get… intercepted!

1 comment:

Jon Mueller said...

Nice example and analysis. It's great to run across this new blog. I've subscribed to your feed. A great topic! Good luck with this. I look forward to reading more.

Jon Mueller