How much of your rating is just noise?
About 95 points of it. Fit a line through fourteen years of a player's ratings and the typical one sits 24 points away from that line in any given year — which means a 40-point swing is not news about you.
The question nobody answers
Every player has had the experience of watching a rating fall sixty points over a few weeks and wondering what went wrong. The advice on offer is always qualitative — do not read too much into it, ratings fluctuate — and never comes with a number, so it is impossible to know whether sixty points is nothing or something.
It is answerable. Each player in FIDE's fourteen-year archive has fourteen January ratings, which is a trajectory. Fit the straight line that best describes it and the residual — how far each year sits from that line — is the part of the rating that is not trend.
The trend has to be removed first, and this is the step that decides whether the answer is any good. Most players are drifting gently one way or another; measuring the spread around a flat average would count a declining career as noise and roughly double the number. Subtracting the line leaves the wobble.
Twenty-four points, and a band about 95 wide
Across 4,613 players with enough listed years to fit, the median residual spread is 23.6 rating points. Half of players fall between 14.5 and 35.4, and a tenth are above 52.
Two standard deviations either side of the trend covers the great majority of years, which puts the typical player inside a band roughly 95 points wide. Your rating is not a point. It is a region, and the number on the screen is wherever inside that region you happen to be standing this January.
So the practical rule is blunt. A forty-point move is inside the noise for almost everybody. A hundred-point move is outside it for most people and worth taking seriously. Between those two, it depends on which end of the table you sit at.
| Average level | Players | Typical wobble | Middle half | Band width |
|---|---|---|---|---|
| 0–2199 | 947 | 34.4 | 21.7 to 50.7 | 138 |
| 2200–2299 | 1,000 | 24.1 | 15 to 35.5 | 96 |
| 2300–2399 | 1,182 | 21.7 | 12.1 to 31.8 | 87 |
| 2400–2499 | 865 | 20.9 | 12.9 to 30.2 | 84 |
| 2500–3000 | 585 | 19.6 | 13.3 to 28.3 | 78 |
A least-squares line fitted to each player's January ratings, 2013 to 2026, and the spread of the residuals around it. 4,613 players had enough listed years to fit. Band width is four standard deviations — two either side of the trend.
Stronger players are steadier
The wobble shrinks as the level rises, and it does so consistently. Players averaging below 2200 across the window have a median residual spread of 34.4 points. At 2200 to 2299 it is 24.1, at 2300 to 2399 it is 21.7, and at 2500 and above it is 19.6 — a little over half the figure at the bottom.
Part of that is games played. A rating built on forty games a year moves less per game than one built on twelve, and the strongest players in this sample are the ones competing most — the article on annual movement finds their flat-year rate is half everyone else's.
Part is that results get more predictable as the gap between opponents narrows relative to the level. Whatever the mix, the practical effect is that a fifty-point move means more if you are 2500 than if you are 2100, and most advice about reading ratings ignores that entirely.
What to do with this
Stop comparing your rating with your rating from last month. The comparison that survives this measurement is against your own trend over a year or more, and there is no shortcut: fewer than a few dozen games and you are mostly looking at where in the band you landed.
It is also the reason this site reports a percentile rather than dwelling on the rating. A percentile moves far less than the number underneath it, because the whole pool wobbles together — everyone is bouncing inside their own band at the same time, and your position among them is the steadier statement.
And it puts the age findings in proportion. Players over fifty lose about 11 rating points a year, which is less than half the year-to-year wobble of a single player. The decline is real and it is only visible across many players or many years, which is exactly why individuals rarely notice it happening to them.
What this cannot see
Annual snapshots, so it is a floor rather than the whole of the movement. A player who dropped eighty points in March and had recovered by December looks perfectly steady here, and within-year swings are certainly larger than these figures.
Players with no variation at all are excluded, because a perfectly flat line over fourteen years is an inactive rating carried forward rather than supernatural consistency. Players listed in fewer than eight of the fourteen years are excluded too — a line through five points is not a trend.
And the sample is the 5,551 titled players this site can follow across the whole archive, of whom 4,613 qualified. It says nothing directly about a 1200-rated club player, though the pattern across levels suggests their wobble is larger rather than smaller.
Common questions
Is a 50 point rating drop something to worry about?
Usually not. The typical player's rating sits about 24 points away from their own trend line in any given year, so a 50-point move is close to ordinary variation — especially below 2200, where the typical wobble is 34 points. A hundred-point move is outside the noise for most players.
How much do chess ratings fluctuate normally?
The median player strays 23.6 points from their own trend, and half fall between 14.5 and 35.4. That puts a typical rating inside a band roughly 95 points wide. Stronger players are steadier: about 20 points at 2500 and above against 34 below 2200.
Why does this site show a percentile instead of just my rating?
Because a percentile is the steadier number. Every player is bouncing inside their own band at the same time, so the whole pool moves together and your position among them changes far less than the rating does.
Data updated 2026-08-02
Chess.com publishes no rating distribution. Those figures are derived from the Lichess distribution mapped through the conversion anchors, and are rounded accordingly.