Strategy · Part 3

Doubling cube theory

Nothing in backgammon is worth more points than cube judgment — and nothing is misplayed more. The mechanics take a minute (rules §15); this page is the theory: when to double, when to take, and why the match score changes both answers.

The 25% take point

Ignoring gammons, the money-play take point is 25%. The arithmetic:

Take EV = −2 × (1 − p) + 2 × p   →  0 when p = 0.25
Pass EV = −1

Passing always costs exactly 1 point. Taking a 2-cube risks 2 to win 2 — which breaks even when you win one game in four. So: take whenever you can win more than 25% of the time, adjusted up when the opponent threatens gammons (their wins cost you 4, not 2) and down when you have gammon chances of your own.

Cube actions by winning chances (money play)

Your winning chancesCorrect action
Below 50%No double — you're not favored
50–60%Some doubles in volatile positions
60–69%Double / take territory begins
~70%The doubling window opens wide
70–80%Double / take (the taker's edge runs out near 75%)
80%+Double / pass — or too good (below)

The doubling window

The window is the equity range where doubling is correct and the opponent should take. Double too early and you hand over a cube they'll happily use against you; double too late and they simply pass, capping a big win at 1 point. Volatility is the trigger: the swingier the next exchange of rolls, the earlier you should turn the cube — “double when this roll might make them pass next turn.”

Too good to double

When your position is so strong that a double forces an instant pass, doubling can be a mistake: you bank 1 point while the position was on its way to a gammon worth 2. Playing on is correct when your gammon chances outweigh the risk of the game turning around — and holding the cube means the opponent can never re-enter the auction.

Cube ownership is an asset

A double-and-take hands the cube to the taker — only they may redouble (rules of ownership). That option has real value: live-cube equity, blended from the cubeless and dead-cube extremes by Janowski-style formulas. Practical upshot: the taker can play slightly worse positions than raw percentages suggest, because the recube threat is working for them.

Match play: everything bends to the score

In matches, points are not equal — going from 6 to 7 in a 7-point match is worth far more than going from 0 to 1. Match equity tables (MET) convert scores into match-winning chances; Kazaross-XG2 is the modern standard. Sample values to 7 points:

Score situationLeader's match equity
1-away vs 1-away (DMP)50%
2-away vs 2-away50%
Crawford game, leader 1-away≈70%
2-away vs 5-away≈78%

Take points now come from the MET rather than the flat 25%: a trailer facing a cube that puts the match on the line needs far more than 25%, while a comfortable leader should take conservative cubes and avoid gammonish ones. The Crawford rule and post-Crawford free drops are the score-based extremes of the same logic.

Race cubes in one line

Pure races have formula answers — lead of about 8% of your count to double, 10% to redouble, take within about 12%. The counting methods and the 8-9-12 rule live in pip counting.


The vocabulary — beaver, cube life, EMG — is in the glossary. For where cube rules differ by format, see match vs money play.