Elo Rating

The Elo rating is chess's standard measure of playing strength: a number, typically between about 400 and 2900, that rises when you win and falls when you lose, calibrated so that the gap between two players predicts their expected score against each other. Named after the physicist Arpad Elo, the system was adopted by FIDE in 1970 and now underpins virtually every chess rating you will encounter, over the board or online.

Ratings are wonderful servants and terrible masters. As a grandmaster, I spent years of my life watching these numbers move, and I have seen what a healthy relationship with a rating looks like — and what an unhealthy one does to a player's chess. This article explains how the system actually works, why your online rating doesn't match your FIDE rating, and how to think about the number without letting it think for you.

How the Elo System Works

The core idea is beautifully simple and requires no heavy math: a rating difference is a prediction. If two players are rated equally, the system expects them to score 50% against each other. As the gap grows, the expected score of the stronger player rises along a smooth curve — roughly 64% with a 100-point advantage, roughly 76% with 200 points, and over 90% with a 400-point gap. (Expected score counts a draw as half, so "76%" might mean winning many games and drawing others, not winning 76 games in 100.)

Every rated game is then a small experiment testing that prediction. After the game, your actual score is compared to your expected score, and your rating moves in proportion to the surprise:

• Beat a much higher-rated player, and you gain a lot — the result was very unexpected.
• Beat a much lower-rated player, and you gain almost nothing — the result was priced in.
• Draw someone rated 200 points below you, and you lose points, because you were expected to score well above 50%.

The size of the adjustment is scaled by the K-factor, a multiplier that controls how fast ratings move. New and young players get a high K (their rating should find its level quickly), established players a moderate one, and elite players a low K (their strength is well-measured, so single results should barely move the estimate). Under current FIDE rules, K is 40 for newcomers and most juniors, 20 as standard, and 10 for players who have ever reached 2400 — the exact bands have been tweaked over the years, but the principle is constant: uncertainty high, K high; uncertainty low, K low.

That is genuinely the whole conceptual machine: expected score from the rating gap, points exchanged in proportion to surprise, scaled by K. Everything else is implementation detail.

What a Rating Really Measures

A rating is a statistical estimate of past results against a particular pool of opponents. Three consequences follow, and each corrects a common misconception.

First, ratings only compare players within the same pool. A number has no absolute meaning; it is a position relative to the people you actually play. This is why comparing across systems is treacherous — 1500 on one server, 1500 in another country's federation, and 1500 FIDE are three different claims about three different populations.

Second, a rating trails your true strength. Improving players are systematically underrated: the number reflects the games you played months ago, not the player you are today. This is why juniors on the rise beat their rating expectation over and over — and why K-factors are higher for them.

Third, ratings measure results, not knowledge. You can know a great deal of theory and be rated modestly, because ratings only pay for points on the board. In my coaching experience, rating gains come from removing the causes of lost half-points — chiefly blunders and time trouble — long before they come from adding new knowledge. The rating rewards the whole system: openings, calculation, endgames, nerves, and clock handling, weighted exactly as they affect results.

FIDE, National, and Online Ratings — Why the Numbers Differ

Players are often startled that they carry three or four different ratings. All are Elo-style systems; they differ in pool and parameters.

FIDE ratings are the international standard, earned in officially rated over-the-board events, with separate lists for classical, rapid, and blitz time controls. The classical FIDE rating is the reference number for titles and world rankings.

National federation ratings (USCF in the United States, and equivalents elsewhere) cover domestic events. They use their own formulas and pools, so they drift relative to FIDE — a national rating may run somewhat higher or lower than the same player's FIDE rating depending on the federation and level, and conversion is always approximate.

Online ratings typically run substantially higher than FIDE ratings, and the reasons are structural, not flattering. Online pools start new accounts at higher baselines, include a vast population of casual players, and use different rating engines with different spreads. The same person might plausibly be 1600 FIDE and 1900–2000 online in rapid — but the offset varies wildly by site, time control, and rating range, so treat any conversion chart as folklore. The number that matters is your trend within one pool, not the comparison across pools.

One more distortion worth knowing: different time controls measure different skills. Your blitz and bullet ratings reward pattern speed and clock handling; classical rewards depth and stamina. A large gap between your slow and fast ratings is diagnostic information, not an error.

Rating Classes, Milestones, and Titles

Rough strength bands help you locate a number on the map — with the caveat that all such labels are informal and pool-dependent. Around 800–1200 covers improving beginners; 1200–1600, developing club players; 1600–2000, strong club players; 2000–2200, expert level; and 2200+ is master territory, where formal chess titles begin — FM at 2300, IM at 2400 with norms, GM at 2500 with norms. At the summit, the world elite lives above 2700 ("super-grandmasters"), and Magnus Carlsen's peak of 2882 remains the highest official FIDE classical rating ever achieved.

For perspective on the whole distribution: the median FIDE-rated tournament player sits far below the numbers that dominate chess headlines. If you are rated 1800 FIDE, you are already stronger than the large majority of rated tournament players worldwide — approximately speaking — and stronger than almost everyone who plays casually. The internet's rating discourse, saturated with grandmaster commentary, badly warps perception of what "good" is.

Rating Anxiety and the Healthy Improvement Mindset

I want to finish with the part no formula covers. Rating obsession is the most common psychological ailment in chess, and it actively damages both enjoyment and results. Players refuse to play when their rating might drop, quit sessions after one loss to "protect" the number, choose safe openings against weaker players to avoid variance, and tie their self-worth to a noisy statistic's daily wobble.

Here is the reframe I give every student. The rating is a thermometer, not a goal. You cannot improve the thermometer directly; you improve the underlying player — calculation, endgame technique, openings, blunder-checking discipline, clock management — and the number follows with a delay. Short-term rating movement is mostly noise: swings of a few dozen points mean nothing, and losing streaks happen to everyone at every level. What carries signal is the six-month trend.

Practically: play enough games for the statistics to work (hiding from rated games starves the estimate), review your losses for causes rather than mourning the points, and do structured training on your actual weaknesses — for most club players that means tactics plus middlegame understanding, which you can train systematically in our middlegame courses. Chasing points punishes you twice: it adds anxiety that costs points, and it substitutes score-watching for the work that earns them. The players I've seen improve fastest are, almost without exception, the ones who checked their rating the least.

Frequently Asked Questions

What is a good chess rating?

It depends entirely on the pool and your reference group. As rough, approximate guides: 1200 FIDE-level strength already beats most casual players; 1600–1800 is a solid club player; 2000+ (expert) puts you in a small minority of tournament players; 2200+ is master level; 2500+ with norms is grandmaster territory. Online numbers run higher — a "good" online rapid rating for a committed amateur might be 1500–1800 depending on the site. Compare yourself to your own six-month trend, not to streamers.

What is the average chess rating?

For casual online players, typical ratings cluster in the several-hundreds-to-low-thousands range depending on the platform's system — most sites report averages somewhere around 600–1200 for their general pool, approximately. Among FIDE-rated tournament players — already a self-selected serious group — the distribution centers roughly in the 1400–1800 region. All these figures are approximate and shift with rating-floor policies and pool growth.

Why is my online rating higher than my FIDE rating?

Different pools, different baselines, different formulas. Online sites start accounts at generous defaults, their pools include vast numbers of casual players, and their rating engines have different spreads. None of this means online ratings are "fake" — they are accurate within their own pool. There is simply no fixed conversion; typical offsets of a few hundred points are folklore averages with huge individual variance.

How many points can I gain from one game?

It depends on your K-factor and the rating gap. With FIDE's standard K of 20, beating an equally rated player gains 10 points; beating someone 200 points higher gains roughly 15; beating someone 200 points lower gains about 5. Newcomers and juniors with K of 40 see swings twice that size. Elite players at K of 10 move in small steps — which is why their ratings are so stable.

Who has the highest chess rating ever?

Magnus Carlsen holds the highest official FIDE classical rating ever recorded — a peak of 2882, first reached in 2014. Historical comparisons across eras are tricky because rating pools inflate and deflate over time, but Carlsen's peak, along with Garry Kasparov's long reign above 2800, marks the summit of the official record books.

Related Terms

• Chess Titles — the formal ranks (GM, IM, FM, CM) that rating thresholds and norms unlock.
• Time Controls — why you have separate classical, rapid, and blitz ratings measuring different skills.
• Blunder — the single biggest rating leak at club level, and the first thing to fix.
• Time Trouble — the second biggest leak: points lost not to bad moves but to the clock.
• Armageddon — the tiebreak format that decides events when ratings and scores can't.
• Tempo — the in-game currency of time, as fundamental on the board as Elo is off it.