The Gambler's Fallacy
After a coin lands heads five times in a row, what's the probability it lands tails on the sixth flip?
If you said "more than 50%" — or even if you felt it should be — you've experienced the gambler's fallacy. The correct answer is 50%, exactly as it was on the first flip.
Why the Brain Gets This Wrong
Independent random events have no memory. A fair coin has no obligation to "balance out" its results. The universe doesn't keep a ledger.
But human cognition evolved to detect patterns. When we see a long run of one outcome, our brain flags it as unusual and instinctively expects a correction. This is the representativeness heuristic — we expect small samples to look like the population distribution they came from.
The problem is that this heuristic, useful in many contexts, is simply wrong when applied to independent random events.
How This Shows Up in Betting
"Team X is due for a win." After losing five in a row, it feels like they must be about to turn it around. But past losses don't change the probability of future outcomes. If their true win probability was 40% before each of those losses, it's still 40%.
Chasing a losing streak. The Martingale strategy is partly built on fallacy — the intuition that after several losses, a win is overdue. It isn't. Each bet is independent.
"The number hasn't come up in ages." In roulette, every number has an equal probability on every spin, regardless of history.
Betting against streaks. Some bettors fade teams on long winning streaks, believing they're "due" to lose. Sometimes this aligns with regression to the mean (a real phenomenon). Often it's just gambler's fallacy dressed up.
Gambler's Fallacy vs. Regression to the Mean
These are related but distinct concepts, and confusing them is easy.
Regression to the mean is real: a player who shoots 45% from three-point range all season but goes 8/10 in last week's game is likely to shoot closer to 45% going forward. Not because of "balance," but because extreme outcomes contain randomness, and random noise regresses toward the true underlying rate.
The difference: regression to the mean works by recognizing that extreme samples contain random error. Gambler's fallacy incorrectly attributes past outcomes to future probabilities on processes with no true underlying drift.
For sports betting: it's legitimate to think a team's underlying quality will express itself over time (regression to the mean). It's fallacious to think past results directly cause future outcomes.
The Practical Test
Ask yourself: does the past outcome mechanically affect the future probability?
- Coin flip, roulette wheel, dice: No. Each is independent.
- A player who's overtired after a long road trip: Yes, that's causal.
- "The team is due for a win": No. That's fallacy.
- A team whose key player is injured: Yes, that's informational.
Genuine edges come from causal or informational factors, not from statistical "balance" that doesn't exist.