Does the law of large numbers make black due after a roulette wheel lands red eight times?
answer
- the wheel keeps no ledger
- independence, not compensation
- dilution rather than correction
- later spins swamp the streak
- proportion converges, debt is never repaid
basics
~20 sNo. Spins are independent, so the chance of black is exactly what it was before the streak. The law of large numbers dilutes early results under a huge volume of later ones; it never reaches back to correct them.
solid answer
~40 sNo — this is the gambler's fallacy. Roulette spins are independent, so the eight reds change nothing about spin nine; on a single-zero wheel, 18 of the 37 pockets are black both before and after the streak. The confusion comes from reading the law of large numbers as a self-correcting force. It is not. The law works by *dilution*: as the number of spins grows, those eight reds become a vanishing share of the total, so the observed proportion of red drifts toward its true value without anything ever compensating for the streak. A useful test of the reasoning: ask what physical mechanism would make the wheel favour black next. There isn't one — the ball does not know the history.
go deeper
Answer no and say the word independence. Be ready to state that the probability on the next spin is unchanged by everything that came before, and avoid the phrase "law of averages" entirely.
Explain the mechanism you are rejecting. An interviewer wants to hear that the law of large numbers dilutes early results with volume rather than reversing them, and that this leaves the streak permanently on the books.
Show you can tell the fallacy apart from legitimate inferences on the same data — evidence that the device is biased, and reversion to the mean. Name the mechanism you are claiming rather than gesturing at balance.
Be ready to spot this reasoning inside business arguments — capacity plans and forecasts built on a metric being "due" for a rebound — and to insist that any claimed corrective force be identified explicitly before it is priced into a decision.
## The fallacy The gambler's fallacy is the belief that an outcome that has been rare recently becomes more likely soon — that after a run of reds, black is "due". People often justify it by invoking the law of large numbers, or its folk cousin, the law of averages. The justification is precisely backwards. ## Why the answer is no Roulette spins are independent: the outcome of one spin carries no information about the next. Formally, `P(black on spin 9 | eight reds) = P(black on spin 9)`. On a single-zero wheel there are 37 pockets — 18 red, 18 black, and one green zero — so black has probability 18/37, roughly 0.486, before the streak and after it. (Note that even without any streak, red and black are not a 50/50 proposition; the green pocket is where the house edge lives.) No physical mechanism connects spin eight to spin nine. The ball has no memory, the wheel keeps no ledger, and nothing in the apparatus is tracking a deficit of blacks to repay. ## What the law actually does The law of large numbers says that as the number of spins grows, the *proportion* of spins that come up black approaches 18/37. Compare two mental models of how that could happen: - **Compensation (wrong).** Future spins are biased toward black until the books balance. - **Dilution (right).** Future spins are drawn from the same unchanging distribution. After a million spins, an eight-spin excess of red is a rounding error in the proportion — swamped, not repaid. Only the second is what the mathematics says. This is why the law of large numbers is compatible with the streak persisting, or with a longer red streak arriving later: the law constrains the long-run *ratio*, not the running *difference*. ## The mirror image: the hot hand The opposite inference from the same data is that red is running "hot" and will continue. Applied to a fair, memoryless device that is equally wrong — nothing carries the streak forward either. But it is worth noticing that the two errors are not symmetric in their logic. "Black is due" asserts a force that does not exist. "The wheel favours red" is at least the right *shape* of inference: it treats the data as evidence about the mechanism rather than as a debt to be repaid. It just happens to be very weak evidence here. Eight reds in a row on a fair single-zero wheel has probability `(18/37)^8`, roughly one in three hundred — surprising for one particular table on one particular evening, nowhere near enough to overturn a strong prior that a regulated wheel is balanced, especially once you account for how many eight-spin windows exist across a casino floor. ## Where this shows up outside a casino The same reasoning error appears whenever someone treats an independent or near-independent series as owing them a correction: - "We've had four bad weeks, so we're due for a good one." - "This machine hasn't failed in a while, so a failure must be coming." - "Our metric is below target this month, it'll snap back next month." Each assumes a restoring force. Sometimes there genuinely is one — a real feedback mechanism, a seasonal cycle, or plain reversion to the mean when the earlier reading was an unusual draw. The discipline is to say which mechanism you are claiming. Reversion to the mean is a statement that the *next* draw is likely to be near typical because extreme draws are rare; it is not a statement that the next draw compensates for the last one. If you cannot name the mechanism, you are reciting the fallacy. ## How to answer it in an interview Say no, say why in one clause (independence), and then correct the underlying model out loud: the law of large numbers swamps early results rather than reversing them. If you can add that the observed *proportion* converges while the raw *count* difference need not shrink, you have shown you understand the law rather than merely rejecting the fallacy.
- Someone concludes instead that the wheel must be biased toward red — is that better reasoning?It is at least the right shape of inference: it treats the streak as evidence about the mechanism rather than as a debt to be repaid. But it is weak evidence. Eight reds on a fair single-zero wheel has probability around one in three hundred, which is unremarkable across the many eight-spin windows a casino floor produces in an evening.
- How is reversion to the mean different from the gambler's fallacy?Reversion to the mean says an extreme observation is likely to be followed by a more typical one, because extreme values are rare draws to begin with. It predicts the next value is near average. The fallacy predicts the next value overshoots in the opposite direction to repay the earlier deviation — a compensating force that does not exist.
- Where does this fallacy show up in analytics work rather than gambling?Whenever a run of below-average results is taken as raising the chance of an above-average one — "four bad weeks, so we are due for a good one". The fix is to name the mechanism you are claiming. If there is a real feedback loop or seasonal cycle, say so; if there is not, the run tells you nothing about the next period.
The wheel has no memory and no ledger. It cannot know what it did on the last eight spins, so it has nothing to make up for.
saying these in an interview costs you the question
- Says black is now more likely to balance the streak
- Invokes the law of averages as a corrective force
- Claims the law of large numbers makes short runs even out
- Treats past independent outcomes as changing future probabilities
- Assumes red and black are exactly 50/50 on a real wheel
- Confuses reversion to the mean with compensation