Are Online Backgammon Dice Fair?

Sooner or later every online player has a game where the opponent rolls double after double and every blot gets hit. The natural reaction is to suspect the dice. This page explains how online dice are produced, why lopsided luck is normal, and how you can check the dice yourself instead of relying on a feeling.

How online dice are generated

Servers do not throw physical dice. They use a random number generator, a piece of software that produces numbers with no usable pattern. Each roll is turned into two values from 1 to 6, and every one of the 36 combinations should come up equally often over time.

There are two broad kinds of generator. A pseudo-random generator is an algorithm that starts from a seed and produces a long sequence that behaves like chance; a good one passes standard statistical tests. Some systems also mix in physical sources of randomness, such as hardware noise. Either way, the key properties are the same: each roll is independent of the ones before, and neither player can predict or influence the next roll.

Why streaks happen

Random does not mean evenly spread. In a fair sequence, clusters are expected. Doubles come up on 6 of 36 rolls, about one roll in six, but that average hides long gaps and short bursts. Two or three doubles in a row will happen to everyone fairly often across hundreds of games. A dozen rolls with none at all is just as normal.

The same goes for hits. A single direct shot from 6 pips away is hit by 17 of 36 rolls, a bit less than half the time. Leave three such shots in one game and it is quite likely that at least one will be hit. It feels personal; it is just arithmetic. The shot tables on the pip count page show how often each distance gets hit.

Confirmation bias

Memory is a poor judge of dice. We remember the joker that beat us and forget the dozens of ordinary rolls in between. We also tend to notice our opponent's lucky rolls more than our own. Over many games this builds a strong but misleading impression that the dice favor the other side, the stronger player, or whoever is behind.

A fair test needs records, not memories. It also needs a rule decided in advance: choose what you will count before you look at the results, or it is too easy to find a pattern that is not there.

How to check with your own statistics

Many servers let you save or export your matches. If yours does, you can test the dice yourself:

  1. Collect a large sample. A few games prove nothing. Aim for thousands of rolls.
  2. Count each combination. Tally how often each of the 21 distinct rolls appears. Non-doubles such as 6-5 should appear about twice as often as any one double, because 6-5 and 5-6 are two of the 36 outcomes.
  3. Compare both players. Count doubles and total pips rolled for you and for your opponents. Over a large sample the totals should be close.
  4. Look at specific moments. If you suspect a bias such as more hits when you leave a shot, record every time you left one and how often it was hit, then compare with the expected numbers.
  5. Accept the noise. Small differences are normal. Only large, persistent gaps in a big sample are worth worrying about.

Analysis tools can also measure luck directly by scoring each roll against the average. If you import your matches into one of the programs on the analysis software page, you will usually find that over time your luck evens out.

What reputable servers publish

A well-run server is open about its dice. Signs to look for:

If a site says nothing at all about its dice, that does not prove they are unfair, but it gives you less reason to trust them. When you start playing online, a few minutes reading how the dice work is time well spent.