Sandhya Indurkar

Math, Applied

The Gambler's Fallacy: Streaks Do Not Load the Next Trial

Coin flip streak does not change next flip odds

The idea

Five losing trades in a row feels like a win is due. Four red spins on roulette feels like black is loaded next. A sales rep misses quota three weeks and leadership expects a bounce simply because the streak has gone on long enough.

For independent trials with a fixed probability, the next outcome does not compensate for the past. The streak was unlikely. The next flip is still the same odds.

Gambler's fallacy answers: Does this process have memory, or are we imposing a story on independent noise?

Example: streaks do not change the next trial

A long run of losses feels like balance is due. For independent trials, the next probability stays the same.

Five tails in a row does not make heads more likely on flip six.

Streak probability

3.1%

5 wins in a row

Next trial

50.0%

Unchanged by the streak

A 5-win streak has probability 3.1%. The next trial is still 50.0%. Past outcomes do not change a fair process.

The math

Streak probability

P(k wins in a row) = p^k

A five-win streak at p = 50% has probability about 3%. Rare, but not evidence that the process changed.

No memory

P(next win | past streak) = p when independent

Fair coins, fair dice, and stable conversion rates with enough volume do not owe you a reversal after a run of losses.

Do not confuse

regression to the mean ≠ gambler's fallacy

Extreme performers often snap back because luck mixed with skill. That is a different post. Gambler's fallacy is claiming the next independent trial must balance the last few.

A simple application: quota pressure

Managers sometimes push harder after a slump as if outcomes must rebalance within the month. Check whether the underlying win rate changed or the streak is normal variance on a small sample. Pair with sample size and regression-to-the-mean posts before you reorganize the team.

Sales quota: streak vs stable win rate

Move streak length. The next week stays at the baseline win rate unless the process actually changed.

4-week miss streak: 11.3% likely · next week still 42% hit rate

Probabilities (%)

Miss streak: 11.3% · Next week: 42.0%

Baseline hit rate

42%

Miss streak odds

11.3%

Streak length

4 wks

Optimize (move here)

  • Separate streak stories from measured win-rate shifts
  • Pair with sample size on weekly KPIs

Hold (do not over-react)

  • Expecting automatic rebound after a miss streak

Escalate if

  • Hit rate drops 15+ pp vs trailing quarter baseline

A slump feels like a rebound is due. On independent weeks, the baseline rate has not changed. Check for real process shifts before punitive targets.

The habit: when someone says we are due, ask whether trials are independent and whether p actually moved. Stories about balance are not probability models.