What Is an Elo Rating? The Number That Measures Playing Strength — and How Strong Our Engines Really Are

Ask two chess players to size each other up and the first question is always the same: "What's your Elo?" A single four-digit number has become the honour score of board-game players everywhere. But what does it actually measure, why does a 100-point gap matter so much, and how strong is the program running in your browser compared with a world champion? This article explains it from the ground up.

What Elo is

Elo is a relative rating system devised by Arpad Elo (1903–1992), a Hungarian-American physics professor who was also a strong chess player. The US Chess Federation adopted it in 1960, FIDE followed in 1970, and it gradually became the common language of almost every head-to-head discipline: chess, xiangqi, parts of professional Go, esports, even football has its own Elo index.

The core idea is elegant: Elo does not measure absolute strength — it predicts the probability of winning. Two players with equal ratings are each expected to score 50%. If you out-rate your opponent by 200 points, your expected score is not "a sure win" but roughly 76% — and that gap between expectation and reality drives the update formula: beating a stronger player earns you many points, losing to a weaker one costs you many, and drawing a stronger player is a small profit. After every game, both ratings adjust according to how surprising the result was, scaled by a factor (the K factor) that controls volatility: new players move fast so they settle at their true level quickly, while established players' ratings are steadier.

One consequence matters a lot: an Elo rating only means something inside its own pool of players. 1800 on one chess website, 1800 in FIDE tournament play and 1800 at a local club are three different levels and cannot be converted one-to-one. Remembering that saves a great deal of pointless argument.

How to read an Elo gap

A few landmarks are enough to read any matchup:

• +100 points: expected score about 64% — a clear edge, but losses are routine.
• +200 points: expected score about 76% — the weaker player needs a very good day.
• +300 points: expected score about 85%.
• +400 points: expected score about 91% — a full class apart; winning one game in ten is already an achievement for the underdog.

And the mirror image: gaining 100 Elo on your own rating over a year is a major achievement — it completely changes your results against the same group of opponents.

The human strength ladder

For a sense of scale, here is the FIDE chess ladder (the game with the richest data):

• 400–800: knows how the pieces move; learning the basic principles.
• 800–1200: casual player; knows openings and simple tactics.
• 1200–1800: club player — where most serious hobbyists live.
• 1800–2000: strong regional player.
• 2000–2200: expert; around 2200 is the national-master threshold in many countries.
• 2400 / 2500: the International Master (IM) and Grandmaster (GM) title norms.
• 2700+: "super-grandmaster" — only a few dozen people on Earth at any time.
• The highest rating in history: Magnus Carlsen, 2882 — nobody has ever touched 2900.

Vietnam's own Lê Quang Liêm has crossed 2740 and won a world blitz title — meaning that at any given moment, only a few dozen humans alive play chess better than him.

So how strong are the machines?

This is the fun part — and the part most often exaggerated, so let us be precise.

At full strength, today's top chess programs — such as Stockfish, the engine inside SynerWorlds' Chess game — are estimated at above 3500 Elo on the engines' own rating lists, some 600–700 points above the strongest human champion. Run that through the formula above and the expected score is roughly 99%: bluntly, the best human on Earth would do well to win a single game in a hundred against a full-strength engine. The same holds across the family: in Xiangqi, the Fairy-Stockfish engine sits far beyond the strongest human masters, and in Go, KataGo — a descendant of the AlphaGo wave — plays at a level professional players now study rather than challenge.

But an opponent you can never beat teaches you nothing. That is why every SynerWorlds game has 12 computer levels: the lower ones are deliberately held back — capped search depth, capped skill, and in Chess's middle levels built-in strength targets around 1500, 1800, 2100 and 2400 — so beginners get fair fights and strong players get a ladder to climb. Only the top two levels unleash the engine's real power. In fairness: the Elo figures attached to engines are estimates calibrated on engine rating lists, meant to help you pick a level that fits — they are not official FIDE ratings, and as noted above, different pools do not convert rigidly.

Your own SynerWorlds rating starts at 800 and updates after your games on the same philosophy: climbing steadily through the computer levels is the most honest progress measure there is, because the machine never has an off day and never gets lucky.

Using Elo to actually improve

Three practical suggestions:

Pick opponents slightly stronger than you. A level where you win about 3–4 games in 10 is the fastest learning zone: hard enough to make you think, close enough to show you exactly where you went wrong.

Aim for 100 points, not 500. A 100-point gain is a genuine class change; spread it over a few months and celebrate the small milestones.

Don't let one bad evening define you. Ratings wobble with form in the short term; the trend over months is the truth. Given enough games, the mathematics pulls your rating back to your real level.

In short: Elo is a compass, not a verdict. It tells you where you stand, whom you should be playing, and how far you have come. The rest — one game a day, one new engine level a week — is up to you. See you on the leaderboard.