MGoBlog has moved. The new site can be found at

Thursday, October 12, 2006

Over the last few weeks, broadcasters covering Michigan games have noted a couple things with awe and reverence: Michigan's run defense, which is #1 in the nation, and Michigan's ranking in time of possession, also #1. One of these things is meaningful. The other is an effect, not a cause, and should never be spoken about again. As you've probably discerned from the title, the latter is "time of possession," the most fradulent stat available for the sports media professional and paradoxically one of the most revered.

Time of possession is revered because it's really easy to look back at past years, check the TOP charts, and see that the top performers there are also the top performers in win-loss. Because a lot of craggy old coaches swear by the thing. Because football broadcasters are specifically prohibited from saying anything that hasn't been said since the 1940s, they will approvingly cite a teams massive first-half advantage in time of possession, say something about "keeping the ball away from your opponent," and then cut to the studio. Meanwhile, my fists clench and unclench spasmically as I restrain myself from doing something rash and stat-related.

Time of possession is a fraud. It is a fraud for these reasons:

  • You cannot "keep the ball away from your opponent" any more than a basketball team can keep the ball away from their opponent. When you score or turn the ball over, they get the ball back, no exceptions. Unless you attempt an onside kick, your opponent is getting the ball back after you're done with it. They will have the exact same number of possessions you do, plus or minus one depending on end-of-half and end-of-game hijinks.
  • It describes the actions of teams after they acquire a big lead and not what they do to get said lead. One of the primary reasons Michigan is #1 in time of possession: they've jumped out to massive leads in many games and cruised home. Opponents like Michigan State and Notre Dame have spent entire halves in a spread hurry-up emphasizing quick movement of the ball. Meanwhile, Michigan leisurely pounds the ball into the line until the game is over. Result: in the second half Michigan three-and-outs can take more time than 80-yard touchdown drives by the opponent. This is hugely distorting and tells us nothing more than "Michigan has a big lead."
  • It places undue emphasis on the run game. Michigan features a crushing ground game and a crushing run defense. Result: lots of Michigan runs and very few opponent runs. This naturally helps TOP, but the reason Michigan is good isn't because they possess the ball for relatively large amounts of time but rather those crushing units. Time of possession obscures the real reasons for Michigan's success.
The reason TOP is historically significant is that before, passing was losing. The game was on the ground and even Anthony Carter was a freakish sideshow to the real meat of the game in the trenches. As passing attacks have gotten more efficient, the relative importance of running has come down.

There is still a kernel of truth in TOP: if you have a dominant run game and a dominant run defense you're going to be a good team because you'll hardly ever find yourself in third and long and your opponents will with frequency. Passing is fundamentally higher variance (more incompletions, more 15 yard gains, fewer three yard gains) than rushing and if you are very good at something low variance you will be very good -- consistently good -- indeed. But the amount of time you have the ball doesn't matter. It's what you do with it when you have it.

Basketball is slowly moving away from raw numbers in favor of rates. John Hollinger, Ken Pomeroy, Big Ten Wonk, Dean Oliver, and others have brought forth a Moneyball style statistical revolution, producing stats that encompass more of the game and reflect it more accurately. The fundamental theorem of basketball, as I understand it, is this:
If you score more points per possession than your opponent, you win.
That's obvious, right? It's a tautology. The possession is the atom of basketball. But why, then, are people so often judged by what they do per game when one game may have 60 possessions and another 80? When stat wonks changed the focus from the game to the possession, stats became more illuminating.

Football is a much more complex game, but this is the fundamental theorem of football just as it is for basketball:
If you score more points per possession than your opponent, you win.
The possession is the atom of football, as well, and they can vary even more wildly than they can in a basketball game. Michigan and Minnesota had eight meaningful possessions in their recent game; last year against Northwestern they had more than double that number. And yet the yardage totals ceded were added up in a raw state and presented plainly even though Michigan's defense had twice the opportunities to fail and bend and cede points. And no one ever notices this. In football, the play's the thing, but it's just an electron or a proton or whatever. The possession is the atom.

Wow... uh. Right, this went somewhere far away from TOP. But there you go.