The move you shouldn't make

At the top of a sport you are often not playing the ball. You are playing someone's prediction of you, built from everything a good player is supposed to do. A goalkeeper facing a penalty commits to a side at about the same moment the kicker does, so each is guessing at the other.¹ The move that coaching calls wrong, the shot straight down the middle, draws its value from how confident everyone is about what you ought to do.

That seems to make "do what you shouldn't" a rule of elite play. It is really two rules, for two different moves you shouldn't make, and they want opposite treatment. One is wrong only in the textbook: it stays good when the opponent knows it is coming, so you make it every time it applies. The other is good only because it is unexpected: it works in small doses and dies when it becomes a habit. A move that fails both tests is not daring. It is playing badly.

Michael Bar-Eli and colleagues studied 286 penalty kicks from top leagues and championships.¹ The kick and the keeper's dive went the same way 43% of the time, far from the 0% or 100% that would follow if either side could see the other's choice first.¹ Keepers stayed in the centre on 6.3% of kicks and dove to a side on the rest.¹ Kickers sent 28.7% of their kicks down the middle.¹ Those centre kicks were stopped 13.4% of the time, the same rate as kicks to the right and lower than the 17.4% for kicks to the left.¹

So the kick that looks reckless, aimed at the spot where the keeper is standing, did as well as a kick into a corner.

The authors read their numbers as a goalkeeper's error. A keeper who stayed put stopped a third of the kicks he faced, 6 of 18, against about one in seven when he dove, and they argued that keepers dive because conceding while standing still feels worse than conceding in the air.¹ But a keeper does not face a fixed spread of kicks. A centre kick that met a keeper who stayed was stopped 60% of the time.¹ If keepers stayed more often, kickers would stop shooting down the middle, and the third of kicks a standing keeper saves would shrink. The diving is part of what makes the middle kick work, and the middle kick is part of what makes diving sensible.

Equal at the top

When each side must be unpredictable, the theory of mixed strategies says a server must have the same chance of winning the point whichever direction he serves, and that any significant deviation from the right mix "will generally be pounced upon by a sophisticated opponent."² Mark Walker and John Wooders tested this on first serves in ten matches between highly ranked players at the four majors and the year-end Masters, all but two of them finals.² Serves to the receiver's left and right won at rates consistent with that equality in all but one of the 40 cases they examined.² A later study of nearly half a million serves from more than 3,000 matches, tracked by Hawk-Eye, found men's win rates conforming "remarkably closely" to the theory, and higher-ranked players conforming more closely than lower-ranked ones.³

At that level the "wrong" serve and the "right" one are worth the same. What loses is being readable. Even the best players in Walker and Wooders' data switched from one side to the other too often to be truly random.² Their own example from another sport is blunt: "The pitcher who 'tips' his pitches is usually hit hard".²

That is the first kind of shouldn't, the surprise. Its value is borrowed from the opponent's expectation, and the right dose is whatever keeps the expectation in place.

The textbook error

The second kind looks the same from the stands and behaves in the opposite way. David Romer used play-by-play data to estimate when NFL teams should kick on fourth down and when they should try for a first down.⁴ On the 1,604 fourth downs where kicking was better on average, teams went for it nine times.⁴ On the 1,068 where going for it was better, they kicked 959 times.⁴ The departures ran toward the "conservative" choice, and Romer argued that ordinary risk aversion could not explain them.⁴

Going for it on fourth and short is not a bluff. It does not need the defence to be wrong about you; the defence can expect it and it is still the better choice on average. It also kept working as it spread.⁵ From 2008 to 2017 teams went for it on fourth and 2 or shorter 32.5% of the time; across 2018 and 2019 they went for it 43.6% of the time, and their success rate on fourth and short rose from 61.5% to 65.8%.⁵

The same years show a surprise being worn out. Passing on fourth-and-goal nearly doubled between 2008 and 2015, when 79.6% of fourth-and-goal calls were passes, and fourth-and-goal success rates "plummeted as pass rates soared."⁵ Those figures do not show that defences learned the goal-line fade, but they are what a surprise looks like when it turns into a habit.

The useful question about an unorthodox move is not whether it is bold. It is whether it would still be sound if the other side read it. If it would, it is a textbook error, and the right frequency is always. If it would not, it is a surprise, and the right frequency is low and irregular. If it would be bad even unread, it is a bad move that sometimes works, and the times it works are the ones people remember.

The slogan blurs the two, and each mistake has a cost. A team that treats a textbook error as a surprise rations a move it should make every time. A player who treats a surprise as a discovery makes it a habit and hands it to the other side.