Penalty aiming
A Socratic walk-through of penalty aiming — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why should a penalty taker sometimes deliberately aim at their weaker side?
A penalty taker has a stronger side. Everyone knows it: the striker, the coach, the goalkeeper, the analyst with the video. If you were advising him, the obvious counsel would be to play to his strength — put it where he puts it best, every time.
And yet elite takers demonstrably do not do that. They go the other way often enough that keepers cannot simply commit. Is that a failure of nerve, or a distrust of one's own strength? Or could deliberately using the weaker side be the correct advice, in a sense we can actually derive?
Reasoning it through
REASONING #Start with the timing, because it determines the whole structure of the problem. The ball travels twelve yards in roughly four tenths of a second — less time than a goalkeeper needs to see the direction, decide, and get a hand to a corner. So the keeper must commit before the ball is struck, or at latest while it is being struck. Neither player can observe the other's choice in time to respond to it.
What kind of situation is that? Two people choosing at the same instant, each longing to know the other's choice. Now the crucial question: is there a best choice for the taker independent of what the keeper does? Suppose he always uses his stronger side. The keeper, who has seen the video, always dives there. The taker is now scoring at his worst rate against that side — worse, quite possibly, than his weaker side would deliver against a keeper diving the wrong way.
So the value of "aim at my strength" depends on how often the keeper expects it. The taker's strength is not a fixed quantity at all; it is a resource consumed by being used predictably.
Follow that where it leads. If always going strong is bad, and always going weak is obviously worse, the taker must mix. And here is the genuinely surprising part: consider what the mixture must look like at rest. Suppose the mix left his stronger side yielding a higher success rate than his weaker side, given the keeper's behaviour. Then he should use the strong side more — but using it more makes the keeper cover it more, which pushes its success rate down. The adjustment only stops when the two sides deliver equal success.
That is the strange and beautiful prediction. Not that the taker should favour his strength in proportion to how strong it is — but that in equilibrium, the very thing he is doing makes his two options equally good. And the same reasoning applies to the keeper: his mix must be one that leaves the taker indifferent, or the taker will exploit it.
Notice that this is a testable claim, not just algebra. It says that if you gather real penalties and split them by the side the taker chose, the scoring rates should come out about equal — and the sequence of choices should show no exploitable pattern. Both have been checked. Chiappori, Levitt and Groseclose analysed professional penalties and found behaviour consistent with the equilibrium; Palacios-Huerta assembled around fourteen hundred kicks and found scoring rates statistically indistinguishable across the taker's choices, with sequences passing tests for serial independence. It stands as one of the cleanest confirmations of mixed-strategy play outside a laboratory — notable partly because ordinary people in lab experiments are conspicuously bad at randomising, while these professionals were not.
The analogy
THE ANALOGY #Think of a shopkeeper who must decide each morning how much of the shelf to give a popular line and how much to a slower one. If the popular line is left permanently in the prime spot, competitors set up next door selling exactly that. The stock that earns most is not the stock that sells best in isolation — it is the allocation that leaves a competitor no obviously best place to attack.
A shopkeeper can revise the display after seeing what the competitor does, whereas the penalty taker and keeper commit simultaneously and only once — and the shop's customers are indifferent, whereas the keeper is actively studying you.
Clarifying the model
THE MODEL #Three refinements, because the tidy result is easy to over-read.
The equal-payoff conclusion does not mean the taker aims at each side half the time. Equilibrium equalises the payoffs, not the frequencies: a taker with a large asymmetry should still use his stronger side more often, just not so much that the keeper can profitably camp on it. The observed frequencies in real data are indeed lopsided, and that is entirely consistent with the theory.
Nor is "sometimes aim weak" a recommendation to be reckless. It is a recommendation to be unpredictable in a specific proportion — and the proportion, not the individual choice, is what carries the value. A single kick to the weaker side that is saved was not thereby a mistake, any more than an insurance premium is a mistake because the house did not burn down.
And the two-option picture is a simplification. Takers can shoot down the middle, keepers can stand still, height matters, and some techniques delay the visible cue and shrink the keeper's information further. Adding those actions changes the arithmetic without changing the logic. It is also worth saying that these findings come from elite professionals with enormous practice at exactly this task; there is no guarantee that amateurs, or players in other one-shot strategic settings, mix nearly so well.
A picture of it
THE PICTURE #How to readThis is the choice grid, not a sequence: the two columns are the keeper's simultaneous options and the two rows are the taker's. Read a cell as the outcome when those two choices meet. The point is what the grid does not contain — no row is best in both columns, so there is no safe pure choice for either player. Each is trying to land in the cell the other is not covering, and since neither can see the other's commitment in time, the only stable answer is a mixture rather than a rule.
What became clearer
WHAT CLEARED #A penalty taker's stronger side is not a fixed advantage but a predictable one, and predictability is exactly what the keeper is paid to exploit. Because both commit before either can react, no single choice can be best — so the taker must randomise, and at the mix where randomising is stable, both sides yield the same expected success. That is why deliberately aiming at the weaker side is correct play rather than self-sabotage, and why real penalty data show near-equal scoring rates across the sides takers choose: the equality is not a coincidence but the signature of equilibrium.
Where to go next
ONWARD #- Why keepers who dive early or stand still fare differently from what the simple grid predicts.
- Whether the same equalisation shows up in tennis serve placement, where the data are also rich.
Key terms
TERMS #| Term | What it means |
|---|---|
| Mixed strategy | a way of playing that randomises over several actions in fixed proportions, rather than always choosing one. |
| Minimax or Nash equilibrium | a pair of strategies from which neither player can profit by unilaterally changing, which in this game requires both to mix. |
| Indifference condition | the equilibrium requirement that every action a player actually uses yields the same expected payoff. |
| Serial independence | the absence of exploitable pattern across a player's successive choices, a further prediction of equilibrium mixing. |
Every term the collection defines is gathered in the glossary.