Smart Football has moved!

Please check out the new site, smartfootball.com. All future updates will be made there.

Sunday, November 09, 2008

Texas Tech's Offense and the Hot Hand Theory

After Texas Tech's drubbing of Oklahoma St, and the much-quoted fact that they scored a touchdown on seven straight possessions, I heard yet another commentator say that their offense was "streaky." And you hear this about other offenses too, and you hear it constantly in other sports, particularly about shooters in basketball and hitters in baseball. As I've written about previously, I think the idea of "hot streaks" are overblown.


Try flipping a coin fifty times. If you chart out the results, I would wager that it does not look as even as you might expect. Just because it's an equal chance of heads or tails doesn't mean you neatly get heads-heads followed by tails-tails. Instead you get seemingly bizarre - seemingly streaky - patterns of, say, fourteen heads followed by a few back and forth then sixteen tails. The probabilities aren't all that different.

So it is with most offenses. There's an imaginary equilibrium of how much we'd expect a particular offense to score against a particular defense. This is the average score if, say, Alabama played LSU a thousand times. But there's variance; each game is different. And once you look at it like that, you see how silly it can be to get too wrapped up into comparing a couple of drives back to back.

The answer with a team like Texas Tech is that they have a hell of an offense, and we can just expect them to score a lot. How they get those points, in what order, all in the first half, all in the second, is largely a function of variance, or in other words, luck.

I am reminded of all this because the game that seemed a shining example of this was Texas Tech's 31 point comeback in their bowl game against Minnesota a couple of years back. Tech was down 31 in the second half, and, after a barrage of passes from then-sophomore quarterback Graham Harrell, Tech won, and Glen Mason lost his job. As I stated:

As most of you know, Texas Tech came back from 31 down with 7 minutes to go in the third quarter to beat Minnesota. What was amazing to me, as I watched the game, was that despite the short time frame, the entire thing happened almost sleepily. The "comeback" appeared like some odd mixture of luck and manifest destiny. Minnesota did not really lose the game like most teams who give up huge comebacks do. Indeed, Minnesota should be a team designed to control second half leads: they have an impressive running game and a methodical passing game to complement it. Minnesota did not turn the ball over in the second half, and got a number of first downs. Tech did not get particularly good field position, either. The most frantic moment of the entire game was Tech's 90+ yard drive to kick a 52-year field goal, and even that still seemed surprisingly serene. . . .

There actually is an entire field of study dedicated to this idea regarding sports, investing, and other facets of life and it is called the "hot hand fallacy." (See also here, and here.) Surely we've all experienced and witnessed the "hot streak" or the "cold streak" in basketball where a shooter has a poor half and then literally can't miss in the second. We see the swing in momentum, the crowd cheering or silenced, the shooter's swagger, his confidence, his teammates feeding him the ball, and his confidence to shoot it from anywhere on the court with a hand in his face.

Except that is an illusion. At least according to researchers Gillovich, Vallone, and Tversky: If you're a 40% field goal shooter for the season, you're pretty much a 40% shooter all the time, even if in one game you shot 20-22 and another 1-15. It evens out over time. The difference is just chance.

This same logic applies to football, and to no offense in football more than Texas Tech's. Clearly, over the last several years Tech's offense has been one of the most productive in football. It's been well documented that Leach's offense often sputters for a quarter or two before exploding to score points at an almost ridiculous pace. So maybe the comeback wasn't such an aberration. 44 points is not so abnormal for them--what's the difference if they had scored those touchdowns on every other drive over the course of the entire game, rather than scoring them all in the second half?

I did note an exception to this, though. Not all football teams or quarterbacks act like coins; sometimes they can get rattled, and the probabilities can change on the basis of perceived adversity. The "human coin" would be someone like Michael Jordan. He's shot millions of free-throws, and he was not going to be rattled. If he missed five free-throws in a row, it wasn't because he was rattled, it was because that's how the coin flip turned out (though it was a stacked coin, with 90% heads and 10% tails).

But with young players, they might let it get to them. I noted this with Harrell in that game: he was but a sophomore then, but he had a full-season under his belt. Had he not, I do not think he would have had the confidence to keep the probabilities the same. Flash forward to now. Last second drives against Texas, falling behind early against Oklahoma State. Not an issue. Harrell's just out there coolly flipping his coins. I will end with what I said about the end of that comeback game, which has renewed relevance now.

The upshot of all this is simply that, particularly from an offensive standpoint, you practice to remove emotion and to remove the hot hand effect. You want to be Michael Jordan looking at the game winning free throw like it is just the 156th free throw after a routine practice. I think what made Leach come to tears after the game is that everyone on the team - coaches, player, fans - went about their business as usual. Tech didn't come back by launching hail marys, running trick plays, grabbing turnovers, or even really getting lucky breaks. Everyone bought into the system and the program, did their job, played smart football, and performed.

I think what brought Leach to tears was the realization that, for young kids in a hyperbolic football world, sometimes it's brave and valiant simply to do your job.


* As a final note, sorry for all being all Texas Tech all-the-time recently, but (a) I've been acutely familiar with Leach (once had a long conversation with him about applying the pythagorean theorem to calculate how long a QB's throw was) and that offense for over a decade, so it's nice for me to see their success, and (b) their past two prime-time games have really been the only football I've been able to see recently. In any event, there might be a bit of a delay before my next post, because I'm working on some more detailed substantive posts - or as Orson Swindle likes to call them, my "coach porn" articles - about Florida's offense along with a couple of passing concepts in vogue right now. So stay tuned for those.

9 comments:

Anonymous said...

I see what your saying, but, if you define the game, or section of play, or segment of coin tosses where the same result happens more than the average would predict as a "hot streak" or "cold streak" then they do, indeed, exist.

Your point is a bit moot.

Chris said...

anonymous,

Not sure I follow. Is your argument that if the standard deviation of the occurrence of heads in N flips predicts, to some degree of confidence, that the deviation from average will be no more than Avg+X, and instead the actual occurrence is greater than Avg+X, then that is a hot streak?

If so, then I don't see why that is a "hot streak"--the occurrence says nothing of the cause--but second, my argument said nothing about those situations. My argument was instead that most so-called "hot" or "cold streaks" are within predictable standard deviations.

David said...

If you're going to be delivering more "coach porn" blogs take all the time you need. The info on this site is great.

Mitch said...

This post suggests that Tech's scoring 7 touchdowns on 7 drives is not attributable to having a "hot hand," but the fact that Tech's offense is just very good, and that WHEN the points come is largely a function of "luck." I think that flies in the face of the blog, really.

It's not luck that the points come when they do. This is not like shooting a basketball. The number of variables involved in the success or failure of an offensive drive is significantly larger than those involved in the success or failure of shooting a basketball.

Your argument fails to account for the fact that there is a human component on the other sideline that can greatly increase the chances of your success beyond a standard deviation.

The Air Raid is designed to react to and systemically prey upon the given alignment and coverage. The human element of the opponent is what causes the statistical variance from the seasonal norm.

In other words, we're not just flipping one coin--we're flipping at least two. If you need two flips of heads to "win" and you have a coin on the other sideline that will only flip heads, your chances of success increase.

Your argument doesn't seem to account for an opponent's consistently showing Tech a 5-man box or only playing Cover 2---in other words, that the opponent is playing stupid football.

If Smart Football creates a favorable environment for success, then so does the opponent's Stupid Football. Wouldn't you agree?

Your post raises the issue of "stacked" or "human" coins. Let's refer to them in this situation as "smart coins" or "stupid coins," and--oh, I don't know--let's say Texas Tech has a smart coin and Oklahoma State has a stupid coin.

Anonymous said...

>>My argument was instead that most so-called "hot" or "cold streaks" are within predictable standard deviations.<<

But your assumption is that you know what the "predictable standard deviation" is. Sports statistics are not like a coin toss where you know in the long run "heads" will hit 50% of the time.

In scoring, each game's performance has an ongoing affect on the "average," so a "hot" or "cold" streak CAN occur, and the players' (or teams') average can, and does, fluctuate. So if a player or team is having a game that raises or lowers their average, then it IS a hot (or cold) streak, isn't it?

Really like your blog, btw.

Beernutts said...

Mitch, I see what you're saying, and you have a good point about adjusting to what the defense gives you. But, there's another element that affects offenses. Often times a whether a drive is killed or not is determined by very mundane things...missed assignment, dropped easy pass, slipping on the grass, etc. These things are difficult to account for in an offense, and often times these things will happen.

If it happened while a team might have a "hot hand" and stops a drive, does that make them not have that same hot hand? Of course not. They're still running their offense.

If an offense is good, they will overcome these things in the long run, but in the short run, they can pop up anywhere.

Anonymous said...

The problem here (and with the abstraction of human activity into a subject of merely mathematical science) is that you have to allow for that one exception--the "young QB" exception, or whatever--that is basically an admission that human beings are not coins, do not behave like coins, and cannot be fruitfully compared with coins. What they do is not random, and one cannot predict, in advance, which offense will be a 40% scroing offense and which will not. All these measurements are done retroactively because there is literally no way to calculate what a human being, much less what a group of 85 of them, will do when in competition with one another.

So you can deny the existence of "streakiness" if you want, but you have to admit one exception (for a young QB), then another (for a fiery coach, like Steve Spurrier), then another one (for a risky offense), then another one (for an unusually brutal pass rush), until what you're left with is the same thing social scientists are ALWAYS left with--the absurd spectacle of pretending that there is such a thing as social science in the first place, and that human behavior is basically the same, in principle and as a subject of study, as the behavior of electrons. Coins do not acquire "swagger," and they do not learn. Pretending that the difference is a matter of sematics always leads you to some overdetermined theory in which the theory occupies a single page and the exceptions occupy the subsequent fifty pages, rendering the theory basically useless as a model of prediction. And anything which cannot make predictions is not science.

Anonymous said...

Sort of a long-winded way of saying, it's not that MOST football teams do not act like coins, as you claim--it's that not even a SINGLE football team acts like a coin. Thus the Hot Hand theory.

Anonymous said...

Even more succintly: Football is not a game of chance, but a game of skill. Your argument can't get off the ground for this reason. "Streaks" do exist for human, not merely statistical, reasons.