Tools › Race doubling
Race-doubling calculator
Once a game becomes a pure race, the cube decision is just arithmetic. Enter both pip counts — you are the player on roll — and see whether to double, whether to redouble, and whether your opponent should take or pass.
How it works
The tool applies Axel Reichert's iSight count-difference criteria for races. It increases your pip count by one sixth (because you are on roll), then compares it to your opponent's: you have a double if your raised count is no more than 6 above theirs, a redouble if no more than 5, and your opponent has a take as long as your raised count is at least 2 above theirs. Below that, you are strong enough that they should pass and you cash the game.
Two honest caveats. First, this is for pure races only — once contact is gone. Second, full iSight adjusts each raw pip count for high stacks, gaps and extra crossovers before comparing; this tool works from the numbers you enter, so for tournament-grade accuracy adjust your raw counts first (add roughly: 2 per checker more than two on the 1-point, 1 for each gap on the 4-, 5- or 6-point your opponent doesn't share, and 1 per extra crossover). For a quick read, raw counts are usually close. It is an estimate, not a solver.
Pair it with the doubling-cube trainer for contact positions and the cube guide for the ideas behind it.
Method: Axel Reichert, Improved Cube Handling in Races: Insights with iSight (Backgammon Galore), building on Tom Keith's race-cube work.
Common questions
Does this work for contact positions?
No — it is for pure races only, where all contact has been abandoned and no hits are possible. Cube decisions in contact positions need a neural-net engine; this tool deliberately covers the race case, which can be solved with a simple formula.
What method does it use?
Axel Reichert's iSight count-difference criteria: increase the on-roll player's pip count by one sixth, then double if that exceeds the opponent's count by at most 6 (redouble at most 5), and the opponent takes if the margin is at least 2. It is among the most accurate published race methods.
How accurate is it?
It is a fast estimation, not a solver — it disagrees with a perfect engine on roughly 7% of close races. For the best accuracy, adjust each raw pip count for stacks, gaps and crossovers before entering it (see the note below); for a quick over-the-board read, the raw counts are usually fine.
Why is my count increased by a sixth?
The player on roll moves next, so they are effectively a little further ahead than the raw pips suggest. Inflating their count by one sixth corrects for that on-roll advantage before comparing the two sides.