Is chess rating inflation real?
We counted fourteen years of FIDE rating lists instead of citing anyone. Ratings are not inflating. They are very slightly deflating.
The complaint is old and specific: a 2000 today is not what a 2000 used to be, ratings are drifting upward, and the titles are getting cheaper. The obvious way to check is the table below — FIDE publishes every past monthly list, so we downloaded every January from 2013 onward and counted the active players in each. The pool median falls, from 1904 to 1711. That looks like an answer and it is not one, because the pool itself more than doubled over the same years, from 90,783 active players to 211,925. New players arrive weak and drag the middle down no matter what happens to anyone already rated. A median that falls while every single person is gaining points is entirely possible, and so is the reverse. The pool table cannot settle this argument, and neither can any version of it.
| January | Active players | Median rating | Top 100 average | Rated 2700+ |
|---|---|---|---|---|
| 2013 | 90,783 | 1,904 | 2,702 | 48 |
| 2014 | 101,330 | 1,873 | 2,704 | 50 |
| 2015 | 114,178 | 1,822 | 2,703 | 45 |
| 2016 | 134,689 | 1,770 | 2,702 | 42 |
| 2017 | 152,748 | 1,722 | 2,705 | 44 |
| 2018 | 161,955 | 1,692 | 2,704 | 42 |
| 2019 | 167,608 | 1,663 | 2,703 | 41 |
| 2020 | 173,542 | 1,636 | 2,699 | 37 |
| 2021 | 188,388 | 1,625 | 2,699 | 37 |
| 2022 | 149,377 | 1,643 | 2,697 | 38 |
| 2023 | 151,718 | 1,613 | 2,697 | 40 |
| 2024 | 174,450 | 1,572 | 2,693 | 36 |
| 2025 | 194,644 | 1,722 | 2,688 | 30 |
| 2026 | 211,925 | 1,711 | 2,682 | 33 |
So we asked a different question: what happened to the same people? Every player carries a FIDE ID that never changes, so a player in the January 2025 list can be matched to the same player in January 2026 and the difference read off directly. Nobody enters, nobody leaves, and the composition problem disappears by construction. Across ten usable years the median player loses three rating points a year. Not gains — loses. And the reason to trust that is the second table: the movement is not the same at every age. Under-18s gain 29 points in a typical year, which is what improving looks like. Every adult band loses, and the loss grows with age until players over 50 are down 11 points a year. Points flow from adults to juniors, because juniors are the ones getting better and the rating pool is close to zero-sum. That is the shape of mild deflation, and there is no reading of it in which the system hands out points it did not take from somewhere.
| Age | Median rating change |
|---|---|
| Under 18 | +29 |
| 18 to 29 | +5 |
| 30 to 39 | -4 |
| 40 to 49 | -6 |
| 50 and over | -11 |
Each figure is the median of 10 year-on-year measurements, every one of them made by matching FIDE IDs between two January lists. Three years were excluded and here is why. In 2020–2021, 45.4% of matched players had an identical rating a year later — against about 1% in a normal year. They were not steady, they were not playing: over-the-board chess had stopped. In 2021–2022, 35.4% of matched players had an identical rating a year later — against about 1% in a normal year. They were not steady, they were not playing: over-the-board chess had stopped. In 2024–2025 the median moved 155 points in one year, and the move ran from +367 at the bottom of the scale to -7 at the top. Playing games does not produce a gradient like that; a rule change does.
Two things this cannot tell you. It cannot say whether players are genuinely stronger than they were — a rating only ever measures you against the people you played, so a pool that improves uniformly shows nothing at all, and the engines-and-databases era plausibly did exactly that. And it says nothing about online ratings, which are a different scale entirely on every site; if your Chess.com number and your FIDE number disagree, that is not inflation, it is two different rulers. What it does say is that the specific fear — that FIDE ratings inflate, that the number gets easier to hold over time — is the wrong way round. The elite columns say it most bluntly: 48 players were rated 2700 or above in 2013, and 33 are today, with the top hundred averaging 20 points lower than they did fourteen years ago.
Common questions
So is a 2000 today the same as a 2000 in 2013?
Slightly harder to hold, by this measurement. A 2000-rated player in their thirties loses about four points a year to the pool without playing any worse, so holding the number over a decade means slowly improving. That is the opposite of inflation, though it is a small effect — nobody should feel robbed.
Then why does everyone say there are more 2700s than ever?
Because that was true for a long stretch before this window opened. Over the fourteen years we measured, the count went the other way: 48 players at 2700 or above in 2013, 33 today, and the top hundred averaging 2682 against 2702. An argument that was right in 2005 is being repeated in 2026 without anyone rechecking it.
Why throw away three of the fourteen years?
Because two of them measure a pandemic and one measures a rule change, and the data says so rather than us. In the two excluded pandemic years, 45% and 35% of matched players had a byte-identical rating twelve months later; every normal year sits near 1%. The third year moved everyone by +155 at the median, in a gradient running from the bottom of the scale to the top. None of the three is a measurement of ratings drifting through play, which is the only thing this article claims to measure.
Counted from FIDE’s own published standard rating lists, every January from 2013 to 2026, active players only. Year-on-year figures match players by FIDE ID between consecutive lists. Measured 2026-08-02; no player, ID or birth year is published — only counts and medians.