Jumat, 12 Januari 2007

Rock-Paper-Scissors, Edgar Allan Poe, and Play Selection

Despite the lofty title, this post focuses on the narrow topic of calling the right play in a football game. Coaches spend an enormous amount of time studying film, determining tendencies, creating gigantic scouting reports for each opponent, and then distributing them to their other coaches and to players who do not read them. Barely sleeping is a badge of honor, particularly at the highest levels, we are sure that more work equals more success.

This is surely true, but how is time best spent? And how should the entire idea of "play-calling" be thought of? Despite all the time spent on preparation, when asked most playcallers, such as Notre Dame's Charlie Weis, say that playcalling is "more art than science." If so much of it is gut feeling and chance during the game, then maybe the better strategy is to get some sleep during the week. I'm kidding, as I put a high premium on preparation, but can active playcalling do more harm than good? And what are the boundaries to knowledge and insight into what the other guy is going to do? Is it ever better to pick your call or choice randomly? Don't we already do that quite often?

Poe's Purloined Letter

In Edgar Allan Poe's the Purloined Letter, a character recounts a story of a young man who excels at game called "odds and evens," more popularly known as "matching pennies." The game is a two-strategy version of rock-paper-scissors: Each player secretly turns their coin to heads or tails and then both reveal their choices simultaneously; if the pennies match (both heads or both tails) then one player gets a dollar, if they do not match then the other player gets the dollar. As told in the story, the young man quickly sizes up his opponents, gains a psychological advantage, and amasses a fortune by outguessing his opponents.

I suppose all playcallers think themselves like the young man, but most are probably more similar to the suckers. But here's the rub: The suckers could nullify the young man's psychological advantage.

How?

By choosing randomly. If the suckers put no thought into whether they chose heads or tails, they would do better than if they tried their best to outthink him. They would break even--a fantastic result against the world's greatest matching pennies player--an unnatural genius who, according to the story, would go through lengthy Sherlock Holmsian deductions to determine if his opponent was going to choose heads or tails.

This is a breath-taking result. But it is also scary--would I be better off picking my plays entirely randomly?

Rock-Paper-Scissors and the Bend-But-Don't-Break-Defense

Playcalling, at least oversimplified, is a lot like matching pennies, or--for a more common game--rock-paper-scissors. If I choose rock and you choose scissors, I get a first down. If I choose rock and you choose rock, I maybe gain a couple yards. If I choose rock and you choose paper--whoops, I just got sacked and maybe fumbled too.

A lot of football games come down to who has the bigger rocks and scissors (more talent), but tough, highly competitive games really do come down to whether you picked paper vs. his rock or vs. his scissors. But how many supposedly great calls were just luck? Probably a lot. We try to make educated guesses, but there's something to be said for going random.

Let me backtrack for a moment. John Wooden, the best basketball coach ever, talked a great deal about focusing on his team. Norm Chow, now offensive coordinator with Tennessee, mentioned how very often he really does not know what the other team is even running right then, and it would be hubris to act like he always knew. When a playcaller says that it is more art than science, he's really just saying that he's out there making (educated) guesses, but guesses nonetheless. Wooden's insight about focusing on his team is that time is best spent byfocusing on what you can control: developing your own talents and self-scouting--to avoid situations where you do become predictible.

The message? When you're scouting you're looking for sure things. Times when you know the other team is going to blitz, or is going to run that one screen pass they like or whatnot, and the best thing you can do to win games is make sure that you don't have any of these true "tendencies" that your opponent can act on. The fact that the other team knows you run it 37.4% of the time on 3rd and 4 1/2 on your own 43 is simply not useful information because it doesn't materially narrow their decision-making. If they know you only run it 3.74% of the time, that is material.

To carry the metaphor, you help yourself the most by preventing your opponent from ever knowing that if you lose twice in a row, you always shoot rock. You may still lose three in a row, but you've given him no advantage. Again, this is powerful. Even if you are playing the world's greatest playcaller or rock-paper-scissors champion, you can still break even, and then wait for those rare times when you know they are going to blitz, or come out with scissors, and hopefully carry the day.

So what's that about the bend-but-don't-break? Imagine: You are playing rock-paper-scissors. Whoever wins gets $1, if you shoot the same no one gets anything, but if rock wins over scissors, the winner get $10. What will this do to the game? Anyone with any sense is going to try to play rock more often than anything else and rarely, if ever, play scissors. If you shoot scissors you can win $1, break even, or lose $10. If you shoot rock you can lose $1, break even, or win $10.

This is the theory behind the bend-but-don't-break defense (and to some extent the more wide-open offenses). The idea is that if you play a gambling type defense, you may win more than you lose, but when you guess wrong, you give up a TD or a big play. The bend-but-don't break will concede by giving up many short passes and runs, and hope not to give up the big play. I am not saying this is a superior strategy, and in fact may be a long-run loser, but it's important to understand the theory. The person practicing that defense recognizes that they will probably be wrong more than they are right, but they think it will be worth it in the long-run--the risk is acceptable to them.

Application

This "mixed strategy" thinking is not meant to supplant gameplanning. (Offensive Coordinator: "Sorry Coach, I'm not doing any work this week, Chris's website told me to just go out there and 'wing it.'" Head Coach: "You're fired.") Indeed, much of gameplanning should fit into your estimates of what will and won't be successful, and then you can engage in a bit of the decision to run or pass I detailed in this post.

What it does is it gives you a place to start. You should have a general equilibrium strategy based on your talent and what you emphasize going in week to week. You can hope to be a 50/50 run/pass on 1st and 10 team, with focuses on quick and intermediate passes and power runs. This is your so-called "identity" and your practices will focus there because it is what you do the most. Then you "kink-it," or skew your weekly plan to the things the defense is weakest against. Who do we run against? What coverages will we see the most? Do they blitz a lot?

Another important application is the "intelligent" mixed-strategy. For example, you face a team that runs the gamut of coverages: Cover 1, 2, 3 and 4 and man and zone and every kind of blitz and they also drop 8 guys into coverage sometimes. But you notice that if you line up in a "trips formation" they will only play Cover 1 or 3, then you have significantly improved your chances. You still don't know for sure if they will be in Cover 3 or 1, or if they will or won't blitz, but you r mixed strategy has been narrowed to a better range of possibilities.

Yet, most teams know their own weaknesses. Most defenses match their weakest defenders with their strongest, not content to let half their defense get run over every week. Further, you get into that neverending mental game: I want to throw quick routes because he likes to blitz. But he knows I know he likes to blitz, so maybe I will throw off deeper drops because his defenders will be looking for my quick passes. But then maybe he knows that I know that he knows that I know he likes to blitz, and thus will blitz anyway countering my counter. And so on. Do I have any special proficiency for this? What if the defensive coordinator is straight out of the Purloined Letter? Remember Norm Chow: if you are so certain of what the other team will do or you have a true read on the opposing coach, it's probably just you being arrogant.

Conclusion

Imagine you are a wing-T youth coach, and you have only three plays: the dive, the bucksweep, and the waggle (bootleg). You can win a lot of games simply by selecting those three plays practically at random; each perfectly counters the other. Then, every so often, you'll see that moment when you know that the waggle will be there. The corners are coming up for the run, the receiver has a mismatch, you know the QB will break contain, so you call it--TD.

Simplified, this is where gameplanning, play-calling, and deception all intersect. Although I've focused on play-calling from the sidelines, I recognize that in modern football playcalling differs from rock-paper-scissors in that it is not a static, simultaneous "now show it" game.

In football you call the play, then show a formation--thus narrowing the range of possibilities--then the play begins, and with good recognition both the offense and defense can react to what the defense is doing and put themselves in position to win. Many very good offenses try to "cheat" on good playcalling by calling everything from the line of scrimmage, and the run and shoot and the triple option try to "cheat" even further by putting a premium on "reading the defense" to make themselves right all the time. Many good defenses operate on similar principles. The important thing to remember for now is that deception and duplicity are your best weapons to prevent this kind of targeting, and once you've done that, you tilt the advantage back in your favor, and the "mixed strategy" reemerges as your best course. And again, if you can limit their strategies by formation or design, then you can improve your mixed strategy by being able to choose the things that defeat their known range of possibilities, rather than than having to be totally random.