How the skill rating works

The formula, how reliability sets the step size, a worked example computed by the live engine, and what the rating deliberately ignores.

The formula

Every player starts on 1500. After each completed game the two teams are compared on their average rating, that comparison produces an expected result, and the points that change hands are the gap between the expected result and what actually happened.

The rule

Winning is worth 1 and losing is worth 0. The score margin does not affect the rating change — a 21–19 win and a 21–4 win move the same number of points. The winning team always gains and the losing team always loses; how much depends on the expectation and on how settled each player's own rating is. A best-of match (best of 3 or 5 games) is still one result, but it is compared against the chance of winning the whole match rather than one game, and it moves ratings as much as the games needed to win it — a best of 3 like two games.

You carry one rating per sport, wherever you play it. A club night, a cross-club night and a game at another venue all feed the same number, and every board — club, venue, city, country — is a view of that number with its own listing gate. Doubles and singles are rated separately: the headline is your doubles rating, and singles games move a singles rating of their own.

How fast a rating moves

Each player carries a reliability: how sure the rating is of them. It starts at 0% for a newcomer, climbs with every set they play (about 63% after ten sets, 95% after thirty) and drifts back down when they stop playing — it halves for every three months away. The step size follows it: a brand-new player moves 40 points per full surprise, a fully settled player 18.

Sets of experienceReliabilityStep size
00%40
1063%26
3095%19
60 (the cap)99.8%18

Each player moves by their own step, not a court average: a newcomer partnering a regular takes the bigger share of the same result. And a result against players who are still settling in counts a little less for you — from 60% of a normal game when the other side is brand new, up to 100% against settled opponents — because the less sure we are of them, the less their result says about you.

Reliability measures how much the rating has been tested, not how good the player is. A returning player with a rock-solid rating from last year is treated as less certain simply because a year has passed; a few nights back on court restore it.

How a doubles result is shared

A doubles game has one result and four players, so the calculation has to decide who the result belongs to. It belongs to the pair.

  1. Each side's strength is the average of its two players' ratings going into the game, after every earlier game has counted.
  2. The expected result comes from comparing those two averages.
  3. Both players on a side are judged against the same expectation; each then moves by their own step size, scaled by how settled the opponents are.

The rule

Partners always move in the same direction from the same expectation. Only the size of the step differs, and only because a rating that is still settling in moves faster. Singles works the same way with one player per side, into a separate singles rating.

There is no attempt to credit one partner more than the other for the play itself. A 1700 player and a 1300 player on the same side face the same expectation, whoever carried the game — the score cannot say.

Someone who has never played starts on 1500 with no reliability. Their first results move them quickly, and results against them count less for their opponents until they have settled.

The calculation uses each player's rating going into the game. Team balance, when a club turns it on, compares the ratings players joined the session with instead, so the two can differ late in a long night.

A worked example

Your pair averages 1490. The pair across the net averages 1580. The four of you have about fifteen sets behind you, so every rating is fairly settled (reliability around 78%).

ResultRating change
Expected chance of winning37%—
You win 21–18Win+13 each
You lose 18–21Loss-8 each

Beating the stronger pair is worth more than beating an equal one, and losing narrowly to a stronger pair still costs points — the margin is not a consolation. These figures are computed by the same function the app runs after a real game, not written into this page, so they cannot drift from the engine.

One number everywhere

There is no separate club rating, venue rating or city rating. Every rated game you play — at your club, on a cross-club night, at a venue you visit — moves the same number, and each board simply lists the players who have played enough at that scope recently. Your position can differ from board to board because the company differs, not because the number does.

The rule

A board lists a player once they have ten games at that scope in the last ninety days and a linked account. Everyone who was on court still counts in the maths, listed or not.

Ratings only compare people who are connected by games. Two groups that never meet are each internally consistent but not comparable with each other until players start mixing — a 1650 in one town and a 1650 in another say the same thing only once enough games join the two.

When the formula itself changes, every rating is rebuilt from the full history under the new rules rather than carried forward from the old ones, so nothing in the ledger ever depends on which version was current on the night.

See also Seasons, the points race and awardsAchievements overview