PR stands for performance rating, and it is the modern standard for measuring backgammon skill. In one number, it answers the question that win-loss records cannot: how well did you actually play? Because dice decide individual games, you can play brilliantly and lose, or play poorly and win. PR cuts through the luck and grades your decisions. Here is exactly what it means and how to bring yours down.
Every roll is committed in advance and provable afterwards. Every match is rated. Every game can be replayed move by move with a computer's assessment beside your own. Nothing about how you are judged here is hidden from you.
Play ranked, freeA backgammon engine knows the best available play in any position. After a game, it compares every decision you made, each checker play and each cube decision, to that best play, and measures how much equity (your expected outcome) you gave up by choosing differently. It adds up all that lost equity, divides by the number of decisions, and scales the result. That average error, per decision, is your PR.
Two things follow from this. First, lower is better: a PR of 0 would be perfect, computer-level play. Second, PR is about consistency: a single huge blunder hurts, but so does a steady trickle of small ones, because it is an average across every move you made.
PR is a continuous scale, but these rough bands are a useful guide:
| PR | Level |
|---|---|
| Under 5 | Expert to world-class |
| 5 to 8 | Advanced |
| 8 to 12 | Solid intermediate |
| 12 to 18 | Developing player |
| Above 18 | Casual or beginner |
Do not be discouraged if your early numbers are high. Most strong club players live in the 6 to 10 range, and the very best players in the world rarely average below 3 or 4 over a long match. The point of PR is not to hit a magic number; it is to give you a trustworthy measure that goes down as you improve.
Win rate is noisy. Over a handful of games, the dice dominate, and a beginner can beat an expert. PR is far more stable because it grades the quality of your choices rather than the outcome. That makes it the honest mirror: it will not flatter you after a lucky win, and it will not punish you for a well-played loss. If you want to know whether you are genuinely getting better, watch your PR trend over many games, not your last result.
Every PR is an average, and an average depends on what you divide by. That is where the programs part company. GNU Backgammon's own error rate divides your total error by your unforced moves only, the turns where you had a real choice. The Snowie error rate divides by every move made by both players, forced ones included, which spreads the same errors over more turns and produces a smaller number. eXtreme Gammon's PR was scaled to sit close to the Snowie figure, but it uses its own definition of a decision, so the match is not exact.
Depth matters too. A program can look at a position quickly or think several moves ahead, and on close plays the two can disagree about which move was best. Analysis is a very strong opinion, not an oracle, and in positions where rollouts are close, a "mistake" of a tiny size may not be a mistake at all.
The practical rule is simple: compare a PR with other PRs from the same program and the same settings. Comparing your number here with a friend's number from a different app tells you roughly where you both stand, and no more than that.
One short game does not contain many decisions. A large chunk of any game is forced or nearly forced: dancing on the bar, a bear-off with one sensible play, a race where every move is obvious. Strip those out and a single game can rest on a fairly small number of real choices, so one bad cube decision can move the number a long way.
That is why you should read a single game's PR the way you would read one round of golf: interesting, not conclusive. The number starts telling the truth as it accumulates. Look at your average over many matches, and look at whether the average is moving, rather than celebrating or worrying about any one result.
PR measures error, and some games are simply easier to play without error than others. A smooth race where you were ahead from the start gives few chances to go wrong. A tangled back game, a prime-versus-prime battle or a blitz is full of hard decisions, and even strong players make measurable errors in them. A player who steers every game toward simple positions can post a lower PR than a stronger player who keeps getting into complicated ones.
None of that makes PR useless. It makes it a measure to read over many games, across many kinds of position, which is exactly the condition under which the effect evens out.
A ladder or Elo-style rating moves on results: you beat someone, you go up by an amount that depends on how strong they were. It answers "how do my results compare with my opponents' results?", and because results include luck, it settles slowly. PR ignores the result entirely and answers "how well did I play the rolls I got?"
The two are most useful together. A rising PR with a flat rating usually means you are playing better and the dice have not caught up yet. A rating that climbs while your PR stays flat usually means you have been lucky, and it is worth knowing that before you stake a lot on it. Ranked backgammon explains how the ladder here works, and skill or luck covers why results alone take so long to separate the two.
Because PR is an average error, the fastest way down is to cut your biggest and most frequent mistakes first, not to polish moves that are already good. For most players that means:
Fix one leak at a time, and watch the number fall.
You cannot improve a number you never see. On Backgammon Battles, every completed match against a person is analysed by GNU Backgammon, the free open-source engine, and you can ask for the same review of a finished practice game against the bot. The number we show as your PR is the Snowie-style error rate from that analysis, so lower is better and it sits on the same rough scale as the bands above.
Alongside the number, the review counts your unforced moves and marks each mistake as doubtful, bad or very bad, so you can see whether your PR came from one blunder or a steady trickle of small errors. Your profile keeps your average PR and your best, which is the trend the sections above say to watch. The analysis runs at a quick setting so results arrive straight after the match; treat a borderline call on a close play as an opinion, as you would from any engine.
One thing PR cannot do is tell you whether the dice were fair. That is a separate question, and here it has a separate answer: every roll is committed before you move and verifiable afterward. If you want results to carry some weight while you measure yourself, matches can be played for Battle Bucks, which put stakes on a game with no real money involved.
Play ranked matches and see exactly how well you played, move by move, with a performance rating on each. Free to play.
Play backgammon freePerformance rating: your average error per decision versus the computer's best play. Lower is better. It measures how well you played, independent of whether the dice won you the game.
Roughly: under 5 is expert, 5 to 8 advanced, 8 to 12 solid intermediate, 12 to 18 developing, above 18 casual. Most strong club players sit around 6 to 10.
Fix your largest, most frequent errors first, usually cube decisions, then loose blots and missed key points. Eliminating a few big blunders per game lowers your number fastest.
Not exactly. eXtreme Gammon, GNU Backgammon and Snowie all grade your moves against their own best play, but they count decisions differently, for example whether forced moves and the opponent's moves go into the average, and they analyse at different depths. Numbers from the same program are comparable with each other; numbers from different programs are only roughly comparable.
More than one. A single short game has few real decisions, so one blunder can swing the number a long way. Treat any single result as a rough reading and watch the average across many matches, which is what your profile shows alongside your best result.
Cube errors hurt PR most, start here →
Why backgammon is a game of skill →
© 2026 Backgammon Battles