Backgammon ELO Ratings Explained

Beating a much stronger opponent is more surprising than beating a much weaker one, so a rating system normally rewards it more. Losing to a stronger opponent is less surprising than losing to a weaker one. The system compares the actual result with an expected result.

That expectation is not a promise about a particular match. Backgammon includes substantial short-term variance, especially in short matches. A lower-rated player can win without invalidating the ratings.

BlitzGammon applies rating changes to rating games and to eligible hosted matches with ELO enabled. Do not assume that every game changes your rating: check the chosen match mode. Ordinary practice against AI is not the same flow as a rated match.

A casual game is useful for experimenting without treating the result as a rating test. A rated game adds competitive consequences, but the same principle of good play applies: choose the move that best serves the position and score, not the one that merely feels less risky.

The current calculator uses a rating-difference divisor of 800 in its expected-win calculation, not the 400 found in many standard Elo examples. Its K-factor depends on the players’ average rating, and a match-length multiplier makes longer matches carry more weight.

For the match winner, the calculation is P = 1 / (1 + 10^((loser rating − winner rating) / 800)). The K-factor is 32 − 0.012 × average rating, limited to the range 5 to 16. Before any performance adjustment, the change is match length^0.75 × K × (1 − P). Here match length means the target points, not the number of games played.

The calculator rounds the change to two decimal places and rounds the new stored ratings to integers. It also applies a lower bound of 100 to the loser’s new rating. These details mean displayed rating differences need not behave like an unrounded textbook example.

The existing analysis flow can halve the rating change for an eligible human-versus-human rated match when the match winner loses the PR comparison. It requires the stored pre-match ratings and both analysed PRs. A match winner who also wins the PR comparison retains the full calculated change.

This is not a separate rating gain for losing the match with better PR: it reduces the magnitude of the change. Analysis and the eligibility checks must complete before such an adjustment can apply, so do not assume every result is immediately final or that every match mode receives it.

• Understand Performance Rating

A rating reflects the pool and format in which it was earned. Numbers from different sites are not automatically comparable, because starting ratings, formulas, player populations and performance adjustments differ.

Track sustained results rather than treating one rating swing as a change in your underlying skill. Use PR and match review alongside ELO: ratings describe competitive outcomes, while analysis can help explain where your decision-making needs work.

• Turn match results into a learning plan