Why Dice Feel Unfair

Almost everyone who uses a digital roller for long enough concludes at some point that it is broken. The rolls felt wrong. There were too many ones. The good numbers came when they did not matter. This is worth taking seriously, because the feeling is real even when the roller is fine.

Clustering is what random looks like

If you ask someone to write down a made-up sequence of fifty coin flips, they will spread the heads and tails out too evenly and avoid long runs. Genuine sequences are lumpier. Runs of four or five in a row show up routinely, and their absence is what would be suspicious.

On a d20, the same number comes up twice in a row 5% of the time. Across a session of about fifty rolls, that is roughly 2.5 back-to-back repeats. Not unusual, but memorable enough that people notice and file it as evidence.

Three bad rolls in a row is not rare

Rolling 5 or lower on a d20 happens 25% of the time. Three of those in succession happens 1.56% of the time, which sounds small until you count how many sets of three consecutive rolls occur in an evening. In fifty rolls there are forty-eight of them, so a run of three bad results is more likely than not to happen at least once every session.

The gambler’s fallacy

Dice have no memory. After four ones in a row, the chance of a fifth is still exactly the same as it always was. Nothing is due, nothing is owed, and the die does not know what it did last time.

This is genuinely counterintuitive, because the odds of rolling five ones in a row from a standing start are very low indeed. But once four have happened, they are no longer a prediction. They are history, and the fifth roll starts from nothing.

Multiple dice are a different shape entirely

Some of the sense of unfairness comes from expecting a flat die to behave like a summed one. A single d20 is flat: every result is equally likely. Add dice together and the middle takes over completely.

On 3d6 the extremes are genuinely rare, so if you are used to rolling three dice and you switch to one, the swinginess of a single die feels like something has gone wrong. It has not. It is just a different distribution. The dice sum page covers why.

What would actually indicate a problem

Not a bad evening. You would need thousands of rolls and a statistical test, and even then the honest answer is usually "no strong evidence either way". This is precisely why our fairness page explains the method rather than claiming a result, and why the randomness article spends its time on the generator rather than on the output.

The reassuring version: on this site the number is decided before the animation starts, by your own browser’s cryptographic generator, and nothing about the dice you see on screen can change it.