What is a PERA rating? The badminton number that reads every point you play

Ask a club player what their rating means and you will usually get a winning percentage in disguise. Most rating systems — chess Elo and its many descendants among them — treat a match as a single binary event: you won or you lost, and everything that happened between the first serve and the handshake is discarded. PERA, the Performance Evaluation Rating Algorithm behind competitive play on app.hmbrsports.com, starts from a different premise. The outcome is an event. The performance is the signal.

In practice, that means PERA reads the score itself. Every rated player carries a number, starting at 1500, that moves after each approved match by one formula: R_new = R_old + K(S − E). Three letters doing specific work, and together they decide what a Tuesday night match is actually worth.

S: the score inside the score

S is a player's actual performance, defined as points won divided by total points played. Win a game 21-19 and your S for that game is 0.525, barely above an even split. Win it 21-5 and S climbs to 0.808. Both results count as victories, but they are not the same evidence about how good you are, and PERA declines to pretend otherwise.

This is the system's central design choice. A 21-5 win and a 21-19 win are different performances, and the rating treats them differently. The dominant win moves your number further. The narrow one moves it less — or, as we will see, sometimes in the other direction. Every point you concede or claw back is information, and none of it is thrown away.

E: what the gap predicted

Against S, the system sets E, the share of points you were expected to win given the rating gap between you and your opponent. E comes from a logistic curve with a scale constant of 500, deliberately flatter than the 400 used in chess Elo, because individual badminton points are noisy in a way that whole chess games are not. Equal ratings produce an expectation of exactly half the points.

The update moves on the difference between S and E, which produces PERA's most discussed property: you can win a match and lose rating. An 1800-rated player who edges a 1700 opponent 21-19, 19-21, 21-19 took barely half the points in a match the gap said they should control, and drops roughly two rating points. The win stands. The performance underdelivered, and the number says so.

K: a factor that knows how sure it is

K decides how far each result moves the number, and in PERA it is not fixed. It adapts between 16 and 32, driven by an uncertainty estimate the system maintains for every player based on how well recent results were predicted. A new player's rating is mostly a guess, so K runs high and the number moves quickly toward the truth. An established player's rating has earned its stability, so K settles near the floor and single results barely nudge it.

The adaptation is bounded by construction: K can at most double its base value, so a few strange results cannot send a rating into a spiral. The system also surfaces that confidence to players as a status ladder, from provisional through developing to established and stable, so a number is never presented as more reliable than the evidence behind it.

The approval gate

None of this math runs on a match the system does not trust. PERA updates only on approved matches, which arrive two ways: scored live by an umpire, or self-scored and then confirmed by both opponents. A self-reported result your opponent disputes, or simply never approves, does not touch either rating at all.

This is a binary gate, not a discount. PERA does not down-weight an unconfirmed score to half credit or apply a suspicion multiplier; it excludes the match entirely. That removes any incentive to game partially counted results, and it keeps the rating pool clean: every point inside it was either officiated or agreed on by the two people with the most reason to argue.

Why this suits club players

Win-loss systems were built for populations with deep match histories. Club players are not that population. Most rated badminton happens in ones and twos per week, and a system that extracts a single bit of information per match converges slowly on the truth. By reading point share, PERA gets far more signal out of every match a club player actually plays.

It is also fairer to the most common club experience: the close loss. Drop a 19-21, 19-21 match to a clearly stronger player and your point share likely beat the expectation, so your rating goes up. You lost the match and earned credit for the performance, which is exactly the order of priorities the system was built around.

Those numbers are not decorative. On app.hmbrsports.com, organizers seed tournament draws directly from PERA ratings, so the quality of the bracket you walk into is downstream of every approved point you have played.