# Why make the dummy run? Bluffing, decoys and how often to fake

Source: https://www.footballdatascience.co.uk/learn/bluffing-dummy-runs
Published: 2026-10-04

> A striker who only runs when the pass is coming is easy to mark. Mix in runs that are never meant to get the ball and the defender can't tell which to follow. Game theory says exactly how often to bluff, and shows the dummy runs pay off through the real ones.

**On the terraces:** Fans moan when a striker makes a run and doesn't get the ball. But a striker who only runs when the pass is coming is easy to mark. This piece shows how often to make a run that's only a decoy, and why the dummies pay off through the real runs.

## The football question

A striker darts in behind and the ball never comes. Was it a waste of energy? Often it's the point. **A run that's never meant to get the ball can still change the game, by making the defender guess.** But how often should a striker make one?

## The concept

This is **bluffing**, and game theorists have long studied it in poker. A player who only bets on good cards is easy to read: fold whenever he bets. So good players sometimes bet on bad cards, just often enough that their opponents can't tell. The same idea is in [part 9](/learn/signalling-is-he-really-hurt), from the other side: there, the question was when a signal can be trusted; here, it's how often to send a false one.

The answer uses the same trick as the [penalty equilibrium](/learn/penalties-nash-equilibrium) in part 1: **mix your choices so the other side gains nothing by guessing either way.** For a bluffer, that means bluffing just often enough that the defender can't tell whether to follow.

## The run in numbers

The numbers are made up, in [expected goals](/learn/expected-goals-from-scratch):

- **A real run** comes when the pass is on. If the defender doesn't track it, it's a chance worth **0.30**. If he does, it's worth nothing.
- **A dummy run** is never meant to get the ball. If the defender tracks it, he leaves space worth **0.10** to a team-mate. If he doesn't, nothing happens.
- **A dummy costs the runner 0.02**, in legs and position.

The defender sees a run and decides whether to follow. Following a real run saves 0.30; following a dummy gives away 0.10. So he follows if real runs are common enough, and the attack controls how common they are by deciding how many dummies to make:

<figure class="rank-chart">
<div role="img" aria-label="Bar chart of xG gained per real run against dummy runs per real run, from 0 to 8: 0.00, 0.08, 0.16, then a peak of 0.24 at 3, then 0.22, 0.20, 0.18, 0.16, 0.14.">

</div>
<figcaption>Too few dummies and the defender follows every run; too many and he ignores them, and the dummies just cost legs. Three for every real run leaves him unable to tell. Made-up numbers.</figcaption>
</figure>

- **No dummy runs:** every run is real, the defender follows every one, and the real runs are worth **nothing**.
- **One or two dummies per real run:** real runs are still common enough that following pays. The defender follows everything; the attack gains only the space the dummies open (**0.08** and **0.16**).
- **Three dummies per real run:** now only a quarter of runs are real, and following a run gains the defender exactly as much as it costs him. He can't tell. At that point each real run is worth **0.24**, the most the attack can get.
- **Four or more:** real runs are so rare the defender stops following at all. The real runs are free, but the extra dummies just burn legs (**0.22**, falling).

$$\text{dummies per real run} = \frac{\text{chance}}{\text{space}}$$

<div class="plain" markdown="1">
In plain football

- **Chance** is what an unmarked real run is worth, here 0.30.
- **Space** is what the defender gives away by following a dummy, here 0.10.
- 0.30 ÷ 0.10 = **3** dummies for every real run. At that mix, following a run and ignoring it cost the defender the same, so he can't gain by guessing.
</div>

## The dummies pay through the real runs

At the best mix the defender follows **20%** of runs: just often enough that a dummy run gains exactly what it costs, 0.02. So the dummies earn nothing on their own. Their whole value is that they make the real runs work: without them each real run is worth 0, with them 0.24.

That's poker's lesson too: bluffs don't win by themselves; they make opponents pay off your good hands. It's also why a striker who's had a quiet game may have played well. The runs that came to nothing were what made the one chance possible.

The mix moves with the stakes:

- **A bigger chance in behind**, worth 0.40, means **4** dummies for every real run: when the real thing is more dangerous, the defender has to respect more runs, so the attack can bluff more.
- **More space for a team-mate**, worth 0.20, means **1.5**: when following a dummy hurts the defence more, the defender follows less readily, and fewer dummies keep him guessing.

## The penalty run-up

The other famous bluff is the penalty run-up: the stutter, the pause, the glance one way. The Laws of the Game draw a line through it. [Law 14](https://www.theifab.com/laws/latest/the-penalty-kick/) says **feinting in the run-up is allowed**, but feinting to kick once the run-up is finished is an offence: the kicker is cautioned and the defending side gets an indirect free kick. Football allows the bluff while the keeper can still react, and bans it at the moment the keeper has to commit.

The logic is the same as the dummy run. A taker who always stutters, or never does, is easy to read. Mixed in at the right rate, the stutter keeps the keeper from guessing early, which is exactly what the takers and keepers in [part 1](/learn/penalties-nash-equilibrium) are trying to manage.

## Why it matters

- **Predictable is beatable.** If you only act when you mean it, opponents learn to ignore everything else.
- **Bluff often enough that they can't tell, and no more.** Too little and they ignore your bluffs; too much and they ignore your threats.
- **Bluffs are paid for by the real thing.** Judge a decoy by what it makes possible, not by what it gets.
- **Rules decide where bluffing is fair.** The run-up feint is allowed; the late feint isn't. A rule can keep bluffing in the game without letting it decide everything.

## Limitations

- **The numbers are made up.** The data can't see which runs were meant to get the ball.
- **Defenders read more than the run.** Body shape, the passer's head, the game state: real defenders have better clues than a run alone.
- **Legs run out.** A dummy run in the 85th minute costs more than one in the 5th, so the best mix changes through a match.
- **Opponents learn.** Over a season, a striker's habits get scouted, which makes it a [repeated game](/learn/repeated-games-ball-back).

## Try it yourself

Change the numbers in the snippet: make an unmarked real run worth 0.50 and see how many dummies the attack can afford. Or, next match, pick one striker and count his runs and the passes he gets. The theory says most of his runs should come to nothing, and that's fine.

## Reproduce the analysis

This needs nothing but Python. All the numbers are made up.

```python
# From Football Data Science by Bryan McGuire. Free to use with credit.
# https://www.footballdatascience.co.uk/learn/bluffing-dummy-runs
# Dummy runs as bluffing. All the numbers are made up, in expected goals (xG).
CHANCE = 0.30   # a real run the defender doesn't track: the pass comes and it's a chance worth 0.30
SPACE = 0.10    # tracking a dummy run pulls the defender away and leaves space worth 0.10 to a team-mate
EFFORT = 0.02   # what a dummy run costs the runner, in legs and position


def defender_tracks(real_share):
    """The defender tracks a run if the chance he stops is worth more than the space he gives away."""
    return real_share * CHANCE > (1 - real_share) * SPACE


def per_real_run(dummies):
    """What the attack gains, for each real run, when it adds `dummies` dummy runs for every real one."""
    real_share = 1 / (1 + dummies)
    if abs(real_share * CHANCE - (1 - real_share) * SPACE) < 1e-12:   # the defender can't tell: he mixes
        track = EFFORT / SPACE        # tracking just often enough that a dummy run gains nothing, and costs nothing
    else:
        track = 1.0 if defender_tracks(real_share) else 0.0
    return (1 - track) * CHANCE + dummies * (track * SPACE - EFFORT), track


print("dummy runs for every real run: what the attack gains per real run (xG), and how often the defender tracks")
for k in range(9):
    gain, track = per_real_run(k)
    print(f"  {k}: {gain:.2f}, tracks {track:.0%}")

best = CHANCE / SPACE   # the defender can't tell when real runs are SPACE / (CHANCE + SPACE) of all runs
print(f"best: {best:.0f} dummies for every real run, so a quarter of runs are real; the defender tracks "
      f"{EFFORT / SPACE:.0%} of runs and each real run is worth {(1 - EFFORT / SPACE) * CHANCE:.2f}")
for why, chance, space in (("a bigger chance in behind (0.40)", 0.40, 0.10), ("more space for a team-mate (0.20)", 0.30, 0.20)):
    print(f"{why}: {chance / space:.1f} dummies for every real run")
```

## Further reading

- [Bluff (poker)](https://en.wikipedia.org/wiki/Bluff_(poker)), Wikipedia: how and why poker players bluff, including the game-theory view that the best bluffing frequency leaves opponents indifferent to calling.
- [Law 14, The Penalty Kick](https://www.theifab.com/laws/latest/the-penalty-kick/), IFAB Laws of the Game: the rule allowing feints in the run-up and penalising them once it's complete.
- [Mixed strategy](https://en.wikipedia.org/wiki/Strategy_(game_theory)#Mixed_strategy), Wikipedia: the idea behind both the bluff and the penalty, choosing at random in proportions the other side can't exploit.