Fulham2–3Chelsea

How to read it

A forecast opened up, beside what followed

A player’s figure is his rating before the match against the average of his side’s: +8% is eight percent above it, −6% six below. The average is over the players on the pitch that participant had rated before, and his rating before the match is the average of what he was given in earlier matches, weighted by the minutes he played — nothing about this match is in it. His bar is centred on his side’s average and its edges are 15 percent either way.

A side’s share is its eleven’s average as a share of the two. It is what the board scores, home against away — so a substitution that takes off a player above his side’s average and brings on one below it moves the side’s share down, and the two tables show which player did it.

The bar leans rather than fills. Ratings are bunched — nobody is rated 2 — so a clearly stronger eleven still reads about 51 to 49. The bar is centred on 50 and its edges are 45 and 55; for plus-minus the centre is level and the edges are half a goal a segment.

Relative, not ratings. A player is published against his side to a whole percent and a side’s share to one decimal, and no rating is published at all: the rows entered by the Arena are a vendor’s own published ratings, and the page is not a way to rebuild them match by match. Plus-minus is the Arena’s own and is raw; a figure marked peers belongs to a player with no segment of his own yet, who takes his club-and-role peers’.

One segment proves nothing. Many end goalless, and a single one resolves almost nothing about a rating. Nothing here marks a segment as called right or wrong; the board pools thousands of them, within the same segment and the same score state, before it says anything.