Gambler’s Fallacy in Craps: Why No Number Is Due
Independent rolls · Due-number myth · Streak probability
The gambler’s fallacy is the belief that a random outcome becomes more likely because it has not appeared recently—or less likely because it has appeared several times. With fair dice, the next-roll chance of 7 remains 6/36 = 1/6 after any previous sequence.
No number is due
If 7 has not appeared in 20 rolls, that history can make the next 7 feel overdue. It does not add a seventh combination to the dice. The next throw still has six ways to make 7 and 30 ways not to.
Independent versus dependent draws
| Experiment | What changes after an outcome? | Does history change the next probability? |
|---|---|---|
| Draw a marble and do not replace it | The bag’s composition changes | Yes |
| Draw a marble and replace it | The bag returns to the same composition | No |
| Roll two fair dice again | The same 36 combinations remain possible | No |
The original CrapsPit explanation used 50 red and 50 blue marbles. Without replacement, drawing one red leaves 49 red and 50 blue, so the next draw changes. With replacement, the bag resets to 50/50. Dice behave like the replacement example: the previous total is not removed from the next throw.
Next roll versus a future window
Two statements are often confused:
- Chance of 7 on the next roll: always 1/6, or 16.67%.
- Chance of at least one 7 in the next five rolls: 1 − (5/6)5 = 59.81%.
The five-roll window has more opportunities, so its cumulative probability is higher. But when roll five arrives, its individual chance is still 1/6 regardless of rolls one through four.
Streaks are part of randomness
| Sequence starting now | Probability |
|---|---|
| No 7 in the next 5 rolls | (5/6)5 = 40.19% |
| No 7 in the next 10 rolls | (5/6)10 = 16.15% |
| No 7 in the next 20 rolls | (5/6)20 = 2.61% |
| Three 7s in a row | (1/6)3 = 0.463%, about 1 in 216 |
A 20-roll stretch without 7 is uncommon but expected to occur occasionally across many tables and sessions. Once it has happened, it still provides no prediction for roll 21.
How the fallacy changes bets
| Fallacy | What actually changed |
|---|---|
| “Take Place bets down because 7 is due.” | Nothing in the next-roll probability; removing bets changes dollars exposed |
| “Press because this shooter is hot.” | Bet size and variance increase; dice probabilities do not |
| “Bet 12 because it has not rolled all night.” | 12 still has one combination out of 36 |
| “A come-out 7 makes another 7 less likely.” | The next 7 remains 1/6 |
You may change a bet for bankroll reasons. Taking money down because the exposure exceeds your limit is rational. Claiming recent totals predict the next roll is not.
Law of large numbers is not a correction force
Over many fair rolls, observed frequencies tend to move closer to theoretical proportions. The dice do not accomplish this by compensating for a past shortage. Future rolls contribute a much larger sample, making the old imbalance smaller as a percentage of the total.
Continue with craps variance, the 36-combination math guide, and table superstitions versus actual rules.
Rewrite the reason before changing a bet
When a streak prompts a bet change, state the reason without prediction language. “Seven has not appeared for 20 rolls” describes the past. “Seven is more likely on roll 21” makes an unsupported forecast. “I have more money exposed than my limit allows” describes a current bankroll fact and can justify reducing action.
- Name the completed sequence without words such as due, hot, cold, ready, or overdue.
- Write the unchanged next-roll probability for the outcome you are considering.
- Calculate how many additional dollars the proposed bet places at risk.
- Change the wager only if that exposure fits a rule set before the streak began.
This pause does not require ignoring what happened. It uses the history for recordkeeping while refusing to convert it into information the dice do not contain.
Loved your analogy, right out of my statistics books! Now do you believe that the dice can be thrown in such a way as to create enough of systematic perturbation so that the average value of the 2 dice thrown, say over 300 rolls does not follow the Central Limit Theorem? And hence we can avoid having that dreaded number I shall not mention be the mean value of the distribution?
Hi, jagxkr. Wow, that question was a mouthful.
For the benefit of our other readers, let’s keep it simple. For lessons on the Central Limit Theorem, mean, standard deviation, and other statistics terms, we’ll leave that to our readers to research. By “systemic perturbation,” I assume you mean shake up (or more accurately, influence) the results such that they don’t follow the normal distribution. By definition of the Central Limit Theorem, a large number of random independent rolls will have a normal distribution. In a word, my answer to your question is, “No.” If a large number of independent craps rolls are random without unnatural influence, the results will tend to follow the normal distribution (i.e., Central Limit Theorem). If you follow the casino’s rules for throwing dice, the rolls will, indeed, be random and independent. I think what you’re asking is what the dice-control cons try to sell people. In other words, I think you’re asking if there’s a way to influence the dice so their outcomes are not purely random such that we avoid the number 7 appearing as often as it should according to the normal distribution. As explained in our lesson on Dice Control, we here at the Crapspit strongly believe that “dice control” is a con designed entirely to sell you something (i.e., transfer your money from your pocket to their pocket), such as seminars, paid memberships, lessons, books, DVDs, and merchandise (e.g., practice tables, software to track your rolls, etc.). As explained in our lesson on Dice Control, there’s no way to throw a legal pair of dice such that your throwing technique will influence the outcome.
So, let’s paraphrase your question and then reiterate our answer. You asked (paraphrased), “Do you believe that the dice can be thrown in such a way as to influence their outcome to land more or less on certain numbers. For example, do you believe you can influence the outcomes such that the number 7 won’t appear as often as it should according to the normal distribution?” For the reasons explained in our lesson on Dice Control, the answer is simply, “No.” However, there are legitimate ways (i.e., dice control is not legitimate) to influence the outcome of dice rolls. For example, you could sneak illegal dice into the game, as described in one of our other lessons. But by sneaking illegal dice into the game (e.g., using loaded dice), the results are not purely random. So, in summary, your question is based on whether you use legal or illegal dice. If you use legal dice, then the rolls will be purely random (and independent), so the outcomes over a large sample size (i.e., a large number of rolls) will resemble the normal distribution. Accept it, there’s no way you can gain an advantage other the casino when playing craps with legal dice (i.e., we believe you can’t create any “systemic perturbation” to affect the outcome of a pair of legal dice such that you decrease the appearance of the number 7). If you experience someone at the table (regardless of their throwing technique) who hits 15 points in a row, the reason they hit 15 points in a row is because of distribution variance (assuming legal dice are used), not because of some bogus dice-control skill.
Good luck and have fun at the table!
Editorial clarification, September 6, 2026: The statistical explanation above needs correction. One roll of two fair dice has a discrete distribution: totals 2 through 12 have 1, 2, 3, 4, 5, 6, 5, 4, 3, 2 and 1 combinations out of 36. Observing more rolls estimates that same distribution; individual totals do not become normally distributed. The central limit theorem instead describes how appropriately standardized sums or averages of many independent, identically distributed observations with finite, nonzero variance approach a normal distribution. It does not make a seven due or prove dice-control claims from a short sample.
To be profitable just need to roll numbers, don’t need to make passes. Some of my most profitable excursions at the doc table were when I never actually rolled a point but did roll many numbers before severing out. Place bets pay same as pass line and buying 4 and 10 pay better than pass line net take for same number
dude your logic makes no sense at all. read what he said to you and youll see that you cant control the dice enough to effect the house edge. why you said all that nonsense afterward is anyones guess.
Have you lost your marbles?
Just poking a bit of fun at your example, but… why not have a bag of 100 dice? You reach into the bag and being blindfolded, place the 50 pairs in a row. Take the blindfold off, take pencil and paper and note the dice total outcomes for each pair. You need not shoot the dice to get accurate results, which also contradicts the Dice Influencing / Dice Control promoters outcome expectations.
This would then contradict Scenario #1 using dice immediately.
Scenario #2 example as an exercise of 50%/50% marble outcomes are no comparison to possible random dice outcomes. Putting 100 coins in a bag with the addition of heads or tails, removing two at a time and lining them up as pairs would create more variance… and a slightly better example of randomness over the marble example. Repeat if necessary, but after over 1,000,000 trials, the results should be getting closer to 50% / 50%.
Short trials prove that a Craps Player who plays with the idea of short term variance of outcomes IS better prepared to leave the Craps Table as a winner. More time at the table results in time result in outcomes that were expected through mathematics.
Removing 50 pairs of dice a second time to compare with the first 50 pair results would prove the Gambler’s Fallacy for what it is… bad theory, poor mathematics, unreasonable statistics and pure nonsense.
Successful Craps Players expose their bankrolls over short periods of time and wager for beneficial random dice outcomes that defy the mathematical predictions that depend on infinite time and outcome statistics. Otherwise there never could be a winning Craps game, a winning Craps System nor any chance to beat a regressive Casino game if short term dice outcomes mirrored long term expectations.
SevenOut, you completely miss the point of the article, and you apparently completely misunderstand the concept of the Gambler’s Fallacy since you stated that the concept is “bad theory, poor mathematics, unreasonable statistics, and pure nonsense.”
By making that statement, it’s easy to conclude that you don’t know what you’re talking about. Since this will be a long post, I decided to simply create a new page due to the length of the reply and link it from here: https://www.crapspit.org/gamblers-fallacy-bad-theory-statistics/
Editorial clarification — September 7, 2026: The personal criticism in the historical reply above does not meet CrapsPit’s current editorial standards. A useful response should address the claim: past outcomes from fair independent rolls do not change the probability of the next total, while larger samples tend to make observed proportions more stable around their long-run values. Disagreement should be answered with definitions, assumptions and reproducible math. The original reply is preserved for context.