The Gambler's Fallacy and Other Traps in Your Head
Why a coin that landed heads seven times is still 50/50, why twelve matches without a draw change nothing, and how hindsight, selective memory, anchoring and sunk cost quietly rewrite your record.
The gambler's fallacy is the belief that an event becomes more likely because it has not happened recently: a coin that landed heads seven times is "due" tails, a league twelve matches without a draw is "due" one. It is wrong for one reason — the coin and the fixture list have no memory, so the chance on the next attempt is what it was on the first. Below: the arithmetic on both, the mirror-image error that fires after a winning streak, and the biases that keep it alive in your own records.
In short.
- Seven heads is a 0.78% run, but the eighth flip is still 50/50. Rarity before the fact is not pressure afterwards.
- Twelve matches without a draw happens about 1.8% of the time and changes the next draw price by nothing.
- The hot hand is the same error facing forwards: three winners do not make the fourth likelier, only the stake bigger.
- Memory keeps winners and drops the rest — a month recalled as "just short" was −₦860 in the log below.
- The fix is procedural: state your probability before you see the price, and log every bet, refusals included.
What the fallacy actually claims
Stated plainly: a sequence of independent events must correct itself over a short horizon. The correction is real over very long horizons — flip a coin a million times and heads sits near 50% — but it works by dilution, not compensation. The next thousand flips do not repay the seven heads; they bury them.
Coins are perfectly independent. Football is close enough, and where it is not — an injury, a manager change — the effect reaches the odds long before you notice the run.
Coins keep no score
The left column is the chance of seeing that run before it starts. The right column is the chance the next flip is heads, given the run has already happened.
| Heads in a row | Chance of that run from the start | Chance the next flip is heads |
|---|---|---|
| 3 | 12.5% | 50% |
| 5 | 3.13% | 50% |
| 7 | 0.78% | 50% |
| 10 | 0.098% | 50% |
The whole fallacy lives in reading the left column and applying it to the right one. Ten heads is a one-in-a-thousand run to predict in advance; once nine have landed, the tenth is a coin. The unlikely part already happened, and it does not transfer.
The same error on a run without draws
Now the version that costs money. The draw in a league you follow is priced around 3.50, which implies 1 ÷ 3.50 = 28.6%, so 71.4% of matches finish decisively. Twelve pass without one.
| Matches with no draw | Chance of that run | Draw chance in the next match |
|---|---|---|
| 3 | 36.4% | 28.6% |
| 5 | 18.6% | 28.6% |
| 8 | 6.8% | 28.6% |
| 10 | 3.4% | 28.6% |
| 12 | 1.8% | 28.6% |
The right-hand column never moves, and neither does the board: no bookmaker shortens the draw because the last twelve matches were decisive. If anything the true chance sits below the implied one, since the price carries a mark-up.
Example. Say the real draw chance is 26% against a price of 3.50. A ₦500 bet is worth 0.26 × 3.50 × 500 = ₦455 and costs ₦500, so it loses ₦45 on average — a 9% leak per bet. Six flat bets stake ₦3,000 for an expected −₦270. Now stake as though the draw is overdue: ₦500, ₦700, ₦1,000, ₦1,400, ₦2,000, ₦2,800. That is ₦8,400 through the account for an expected −₦756, with 17% of a ₦50,000 bank in a market you would not have touched last week.
The hot hand: the same mistake facing forwards
The mirror image says a run will continue: three winners means you are "seeing it clearly", so the next stake goes to ₦1,500. Genuine team form exists and is already in the price; in your own record there is nothing to continue, because the fourth bet is a new bet on a new match.
The cost is arithmetic. If your log returns −4% on turnover, a ₦500 bet loses ₦20 on average and a ₦1,500 bet loses ₦60. Tripling the stake triples the leak, right after your most confident moments. The reverse move, cutting stakes after losses, is weighed up in betting through a losing run.
Hindsight bias: the bet you say you knew
Once a result lands, the reasons for it look obvious and the alternatives become hard to picture. That corrupts the Sunday review: every winner reads as a call you made, every loser as a freak event.
The tell is "I knew that was coming" with nothing written before kick-off behind it. The correction is dull and effective: record the reason and your own percentage on the slip. A month later those are the only evidence that survives.
Selective memory and the month it hides
Memory keeps winners in detail and losers as a blur, so a recalled month beats the settled one.
| Version | Bets | Staked | Returned | Result |
|---|---|---|---|---|
| Remembered: five good calls | 5 | ₦2,500 | ₦6,925 | +₦4,425 |
| Logged: nine winners, fifteen losers | 24 | ₦12,000 | ₦11,140 | −₦860 |
The five remembered bets — priced 2.40, 3.10, 2.00, 4.50 and 1.85 — really did return ₦6,925 on ₦2,500, and that is the month you describe to a friend. All 24 returned ₦11,140 on ₦12,000, a loss of 7.2% on turnover. Nothing in the top row is invented; it is the part that survived recall.
Three more traps that share the wiring
The fallacy rarely arrives alone. Three relatives do most of the remaining damage, and each has a number that ends the argument.
| Trap | How it sounds | What it does | The number that ends it |
|---|---|---|---|
| Anchoring | "It was 2.40 this morning, 1.95 is robbery" | Judges the price against an earlier price, not against the match | 1.95 means 51.3%. Is your estimate above that? |
| Loss aversion | "Take the ₦1,100 now, before it goes" | Closes winners early and holds losers, because a loss stings more than a win pleases | ₦500 at 4.00 with the lead is worth over ₦1,100 if the side wins more than half the time |
| Sunk cost | "I'm ₦6,000 down, I can't stop now" | Treats past losses as an investment later bets can protect | The ₦6,000 is gone either way. Is the next ₦500 bet worth ₦500? |
All three, and the fallacy itself, work through one channel: the stake. That is why stake discipline fixes more psychology than self-examination does, as the psychology of betting guide sets out.
Five countermeasures that survive a Saturday
- Estimate before you look. Write your own percentage, then open the price. A number formed after seeing 1.95 is not an estimate.
- Ban streaks as reasons. If the written justification says "due", "on fire" or "unbeaten in five", the bet is not placed. Real reasons get written instead.
- Log the refusals. A record of the bets you passed is the only defence against remembering yourself as the person who called them all.
- Review at 100 bets, not 10. Judging a method on one weekend is the fallacy in a spreadsheet.
- Keep the unit fixed. Every bias above needs a variable stake to do real damage.
To practise on live prices, write your estimate for two fixtures on today's predictions before opening the coupon, and keep both numbers.
FAQ
Is the gambler's fallacy the same thing as the law of averages?
No, and the confusion is the point. Long-run averages settle because later results outnumber earlier ones, not because a balancing force pushes back. Seven heads never get repaid; after a million flips they are simply too small to see. Nothing acts on the eighth flip, which stays at 50%.
If a team has drawn none of its last twelve, is the draw not more likely now?
Not from the streak alone. A run of twelve decisive matches is unusual before the fact and irrelevant after it, so the draw chance stays around the 28.6% that a price of 3.50 implies. Real reasons to expect a draw — squad news, both sides content with a point — are visible in the match, not in the sequence.
How do I know whether a bias is actually costing me money?
Split the log. Compare bets that cited a streak against the rest, and bets placed after a winner against bets placed after a loss. If a split still shows a gap after 100 bets or more, it is a leak worth closing; below that it is a hint, and treating a hint as proof is a bias in itself.
Read next
- 15 mistakes beginners make — what these biases produce in practice.
- Bankroll management — the fixed unit every bias needs you to abandon.
This article is for information only and is not an inducement to gamble. Betting involves the risk of losing money — never stake more than you can afford to lose. 18+. If gambling stops being entertainment, read our responsible gambling guide and seek help.
