How much does a chess rating move in a year?

Most years, nothing happens

In 27.4% of the rating-years measured here the number did not move by a single point. That is not a run of drawn games or a season of holding steady. FIDE carries a rating forward unchanged when a player enters no rated events, so an exact zero almost always means the player did not compete.

This has to be separated out before anything else can be read, and most published versions of this figure do not do it. Leave the flat years in and they pile a spike onto zero, drag every median onto it, and produce the useless answer that a rating moves by nothing at every level.

So a quarter of the answer to how much a rating moves in a year is: it does not, because most rated players are not playing rated chess most of the time.

And when it does move, it moves a little

Of the 43,258 years in which the rating changed, the median change was minus 3 points. The middle half fall between 20 points down and 15 points up. Only 45.4% of active years were gains at all.

A twenty-point year in either direction is therefore an ordinary year, and the typical active titled player drifts very slightly downward. That last part is worth sitting with, because it is the opposite of how people talk about ratings — as something you climb.

Most of it is age. This is a sample of titled players followed across fourteen years, and the median one of them is past their peak, which the companion article on peak age puts at 33. A population that is mostly on the far side of its peak drifts down by a few points a year, and that is exactly what appears here.

Only the lowest band actually gains

Split by the rating a player started the year at, and one band behaves differently from all the others. Below 2100, the median active year is a GAIN of 7 points and 56.6% of years are gains. Every band above it has a negative median and fewer than half its years going up.

Two things produce that and they cannot be separated here. Players below 2100 in a titled sample are disproportionately young and improving. They are also disproportionately players having a bad year, who will tend to bounce back for statistical reasons alone — a rating observed unusually low is partly bad luck, and bad luck does not repeat. That second effect is regression to the mean, and it is a property of measurement rather than of chess.

The flat column is the other story in the table. Below 2300 roughly 30% of years are flat; at 2500 and above it is 16.1%. The strongest players are close to twice as likely to actually play in any given year, which is most of what separates a professional from a titled amateur.

How rare a big year is

Among years where the rating moved at all, 8.1% gained 50 points or more. Gaining 100 or more happened in 2.4% of them, 150 or more in 1%, and 200 or more in 0.46%.

So a hundred-point year is not unheard of — it happens to about one active titled player in forty, every year — but it is not something to plan around. Roughly forty such years appear in this sample annually, out of some four thousand active ones.

The useful version of that for anyone measuring their own progress: if your rating is a few dozen points different from last January, you have had a completely normal year, and reading anything into it is reading noise.

What this measures and what it does not

Titled players, standard ratings, and the FIDE January lists only. A rating that rose and fell inside a year shows up here as the difference between two Januaries, so this measures net annual movement rather than volatility within the year, which would be larger.

Improving juniors are underrepresented in a way that matters. This follows players who already hold a title, so the fastest-moving people in chess — those going from 1200 to 2000 in a few years — are almost entirely outside it. The numbers describe a settled population, not an aspiring one.

The population is narrower than "titled players" as well. The series covers the 5,551 who could be matched to an online account, all of whom are on the current FIDE list — so nobody who stopped playing altogether and fell off it is represented, and the flat-year share is if anything an understatement of how much of rated chess is dormant.

A flat year is also inferred rather than observed. A rating that did not move is treated here as no rated games, which is overwhelmingly what it means, but the archive records ratings rather than games and cannot confirm it player by player.