A chess rating system is a mathematical formula that turns your wins, losses, and draws into a single number that predicts how you’ll perform against other rated players. The big three are Elo (the original, used by FIDE), Glicko, and Glicko-2 (used by Chess.com, Lichess, and most apps). Higher number, stronger player — and the gap between two ratings tells you the expected score.
That’s the headline. Now let’s pop the hood, because once you get how the math thinks, your own rating stops feeling like a mysterious slot machine and starts making perfect sense.
The 30-second version
- Elo: Beat someone stronger, gain a lot. Beat someone weaker, gain a little. Lose to someone weaker, lose a lot. Simple, battle-tested.
- Glicko / Glicko-2: Elo’s smarter cousin. It also tracks how confident it is in your number (a “rating deviation”), so new or rusty players move faster.
- A ~400-point gap means the favorite is expected to win about 10 games out of 11.
How does the Elo rating system work?
Elo (named after physicist Arpad Elo, so it’s a name — not an acronym, please don’t write “ELO”) is beautifully simple. It assumes that on any given day your real strength wobbles a bit around your “true” level, like a dart player who’s roughly consistent but not a robot.
Before each game, the system calculates an expected score based on the rating gap. If you’re rated exactly the same as your opponent, you’re each expected to score 0.5 (a coin flip). If you’re 200 points higher, you’re expected to score about 0.76 — roughly three-quarters of a point per game on average.
Then it compares what actually happened to what was expected:
- Outperformed expectations? Your rating goes up.
- Underperformed? It goes down.
- Exactly as expected? Barely moves.
The change is your K-factor (how big a swing each game causes) multiplied by the difference between your actual result and your expected result. FIDE uses a K of 40 for newcomers, 20 for most established players, and 10 for elite players over 2400 — so the higher you climb, the slower your number drifts. That’s on purpose: a 2700 player has proven their level over hundreds of games, so one bad day shouldn’t yo-yo their rating.
How are chess ratings calculated, step by step?
Let’s walk one real example so it sticks. Say you’re rated 1500 and you play someone rated 1700 — a 200-point favorite.
- Expected score: With a 200-point gap, the system expects you to score about 0.24 (you’re the underdog).
- You win! Your actual score is 1.0.
- The surprise: 1.0 − 0.24 = 0.76 more than expected.
- Apply the K-factor (say 20): 20 × 0.76 ≈ +15 points. You jump to ~1515, and your opponent drops by the same ~15.
Notice the elegance: rating points are basically conserved. The points you gain come out of your opponent’s pile. The whole system is a giant zero-sum scoreboard where beating tough opponents is the fast lane up.
What is the Glicko rating system?
Elo has one blind spot: it treats a brand-new player and a 30-year veteran with the same K-factor, even though we’re way less sure about the newbie’s “real” strength. Mark Glickman fixed that with Glicko (and later Glicko-2), which is what powers Chess.com, Lichess, and most online platforms.
Glicko adds a second number for every player: the Rating Deviation (RD) — basically the system’s confidence interval. Think of it as an error bar.
- High RD (new account, or you haven’t played in months): the system isn’t confident, so your rating swings big to find your level fast.
- Low RD (you play constantly): the system trusts your number, so swings shrink.
Glicko-2 adds a third ingredient, volatility, which measures how erratic your recent results have been. A player on a wild streak gets treated as less predictable. This is why your Lichess rating sometimes jumps 30 points after one game when you’ve been away — your RD ballooned while you were gone, so the system is recalibrating quickly.
What is a good chess rating?
Honest answer: “good” depends entirely on which pool you’re in and how long you’ve played. But here’s a rough lay of the land for established online (Glicko-2) ratings:
| Rating range | Roughly means |
|---|---|
| Under 800 | Brand new — learning the rules and how pieces coordinate |
| 800–1200 | Advancing beginner — you see basic tactics and don’t hang pieces every move |
| 1200–1600 | Solid club-level amateur — real opening and endgame ideas |
| 1600–2000 | Strong club player — many tournament regulars live here |
| 2000–2200 | Expert — seriously good |
| 2200+ | Master territory (titles like FM, IM, GM start around here and up) |
Don’t anchor your self-worth to the number. The average rated player is, by definition, average — that’s not an insult, it’s just math. The only rating that matters is whether yours is trending up over months.
Why did my rating drop after a win (or jump after a loss)?
This trips everyone up, so let’s clear it. On Glicko systems, a single game’s points depend on the whole batch the system processes and your current RD. If your RD is high, a result can recalibrate you in surprising ways. And if you beat a much weaker player, you might gain almost nothing — the system already expected you to win, so there’s no surprise to reward.
The flip side, which stings: lose to someone far below you and you’ll shed a big chunk, because that result was genuinely unexpected. The system isn’t punishing you — it’s updating its estimate of your true strength based on new evidence. Cold, but fair.
A fun bit of history
Before Elo, chess used the clunky Harkness system, which produced wonky results that pros openly grumbled about. Arpad Elo — a Hungarian-American physics professor and strong amateur player — proposed his statistical model to the US Chess Federation in 1960, and FIDE adopted it worldwide in 1970. His system was so sound that it later spread far beyond chess: it now underpins ratings in Scrabble, table tennis, esports, and even early versions of Facebook’s “rate people” experiment that eventually inspired a certain social network. Not bad for a chess formula.
Common misunderstandings to dodge
- “Elo is the only ‘real’ rating.” Nope. Elo, Glicko, and Glicko-2 are all legitimate — they just make different trade-offs about confidence and speed.
- “My online rating equals my FIDE rating.” They’re separate pools that don’t talk to each other. (We compare them head-to-head in FIDE vs Chess.com rating.)
- “Grinding weak opponents pumps my rating.” It barely moves it — and it stalls your actual improvement. Play people near or above your level.
- “A rating drop means I got worse.” Usually it means variance, or you faced a tough slate. Judge trends over dozens of games, not one session.
Key takeaways
- A chess rating system converts results into one number that predicts performance via an expected score from the rating gap.
- Elo is the classic; gains depend on beating expectations times a K-factor.
- Glicko/Glicko-2 add a confidence measure (RD) and volatility so ratings adjust faster when the system is unsure.
- A ~400-point edge ≈ winning about 10 of 11 games.
- Track your trend over months, not the daily wiggle.
Frequently asked questions
Is Elo or Glicko better? Glicko-2 is generally more accurate because it tracks confidence and volatility, which is why most online platforms use it. Elo’s strength is its simplicity and decades of proven use, which is why FIDE keeps it.
Why does a 400-point gap mean the favorite wins ~91%? That’s baked into Elo’s expected-score formula: each ~200-point step roughly multiplies the favorite’s odds. At 400 points, the expected score is about 0.91 — so roughly 10 wins per 11 games.
Do draws affect my rating? Yes. A draw counts as 0.5. If you draw a much stronger player you’ll gain points (you overperformed); draw a much weaker one and you’ll lose some.
What’s a “provisional” rating? On Glicko systems it’s a rating with a high RD — usually your first 15–30 games. The system isn’t confident yet, so your number swings hard until it settles.
Can my rating go down even if I keep winning? Briefly, yes, on Glicko — if a batch recalculation or RD shift outweighs a low-value win over a weaker player. Over a real sample it’ll climb if you’re genuinely winning.
What rating is “master”? National master titles typically start around 2200, with FIDE titles (FM, IM, GM) layered above. The exact threshold varies by federation and platform.
Keep climbing
Now that the number makes sense, the real question is how to raise it. That’s where the practice habits come in.
Next up: How to practice chess effectively · Also useful: FIDE vs Chess.com rating and how to get out of a slump. Want the big picture? Visit the improvement hub.
Written and fact-checked by the Chess Lab Academy Editorial Team (AI-assisted, human-edited) · Rating math verified against FIDE Rating Regulations and the published Glicko-2 specification; positions checked with Stockfish 17 · Last verified 2026-06-25 · Sources · How we verify content.