What "vs. expected" means

"vs. expected" compares how many sets were actually won with how many the rating model said should be won. It answers one question: is this character, pairing or team doing better or worse than the players' ratings alone would predict?

How it is worked out

Before every set, the model gives each team a chance of winning based on the four players' ratings at the time. A team that is a 70% favourite is "expected" to win 0.7 of that set. Adding those chances up over every set gives the expected number of wins.

The number shown is the gap between actual and expected wins, as a percentage of the expected wins. +10% means 10% more wins than the ratings predicted, and -10% means 10% fewer. Around zero means the results match the ratings.

Why it is useful

A plain win rate mostly tells you how good the players are: a pairing used by top players will win a lot whatever the characters are. "vs. expected" removes that, because strong players are already expected to win. What is left is the part the ratings do not explain, such as two characters covering each other's weaknesses or a matchup being lopsided.

Things to keep in mind

Small samples are damped. A few "as expected" sets are blended into every figure, so a pairing that is 3-0 does not show up as wildly above expectations.

It is about results, not a guarantee. A positive number shows what has happened so far in the data, from 2015 to today, and it can move as more sets are played.

Characters are each player's current main. The data does not record who played which character in each doubles set, so every player is counted as their main.

You can see it in use on the Stats page and the Character Pairing Records page.