How Chess Elo Ratings Work: FIDE Formula and Examples

FIDE calculates chess ratings using the Elo system: a mathematical formula that compares your actual result to your statistically expected result, then adjusts your rating up or down accordingly. The bigger the surprise (beating a much higher-rated player, or losing to a much lower one), the bigger the rating swing.

Understanding the math behind the Elo system reveals why an 18-year-old Indian prodigy can gain 300 rating points in three years, why Magnus Carlsen’s peak of 2882 is almost incomprehensibly high, and why the K-factor matters as much as the result itself.

For a broader overview of how the rating system works in practice, see our guide to the FIDE rating system explained.

Who Invented the Elo System

Arpad Elo was a Hungarian-American physics professor and chess master who designed the system in the 1960s. FIDE officially adopted it in 1970, replacing an older, less rigorous system.

Elo’s insight was simple: instead of tracking wins and losses, track performance relative to expectation. Beat someone you were supposed to beat and you gain a little. Beat someone rated 300 points above you and you gain a lot.

“The Elo system is beautiful in its simplicity,” noted chess statistician Jeff Sonas, founder of Chessmetrics. “It doesn’t care who you are, only how you perform.”

The system has remained largely unchanged for more than 50 years, though FIDE has adjusted K-factors and publication frequency over time.

What a Chess Rating Actually Means

A rating is not a score out of 3000. It is a position on a continuous scale that predicts your expected performance against any other rated player.

Here is the standard breakdown of rating ranges:

Magnus Carlsen’s all-time peak was 2882, set in May 2014. No player has ever crossed 2900. As of the March 2026 FIDE list, world champion Gukesh Dommaraju sits at 2748.

The Expected Score Formula

The core of the Elo system is the expected score formula. It tells you, given two players’ ratings, what score each should expect.

The formula is:

Ea = 1 / (1 + 10^((Rb - Ra) / 400))

Where:

• Ea = expected score for Player A
• Ra = rating of Player A
• Rb = rating of Player B

Let’s work through a concrete example. A 2500-rated player faces a 2300-rated player.

Expected score for the 2500 player:

Ea = 1 / (1 + 10^((2300 - 2500) / 400)) Ea = 1 / (1 + 10^(-200/400)) Ea = 1 / (1 + 10^(-0.5)) Ea = 1 / (1 + 0.316) Ea = 1 / 1.316 Ea = 0.76

The 2500-rated player is expected to score 0.76 points out of 1 (76% chance of winning, essentially). The 2300-rated player’s expected score is simply 1 - 0.76 = 0.24.

The 400-point denominator is a calibration constant. It means a 400-point rating gap gives the higher-rated player roughly a 10:1 odds advantage in any single game.

Photo: Petr Kadlec, CC BY-SA 3.0 via Wikimedia Commons

The Rating Change Formula

Once you have the expected score, calculating the actual rating change is straightforward.

New Rating = Old Rating + K x (Score - Expected Score)

Where:

• K = the K-factor (explained below)
• Score = actual result (1 for a win, 0.5 for a draw, 0 for a loss)
• Expected Score = what the formula predicted

This means:

• If you score exactly as expected, your rating does not change at all
• If you outperform expectations, your rating goes up
• If you underperform expectations, your rating goes down

The amount of change depends entirely on how surprising your result was, multiplied by K.

K-Factors: Why They Matter

The K-factor controls how sensitive your rating is to a single result. A high K-factor means faster, larger swings. A low K-factor means more stability.

FIDE uses three K-factor levels:

K = 40 applies to:

• New players (fewer than 30 rated games)
• Players under 18 years old

K = 20 applies to:

• Players rated below 2400

K = 10 applies to:

• Players rated 2400 or above
• Players who have previously achieved a rating of 2400 (even if they have since dropped below)

The logic is sound: new and young players’ ratings are still settling toward their true level, so larger swings make sense. Established players at the top have ratings that are already accurate measurements, so only large, sustained outperformance should move them significantly.

A Full Calculation, Step by Step

Let us walk through a complete example with a real scenario.

Setup: A 2400-rated player (K = 10) plays a 2200-rated player (K = 20) in a classical tournament game.

Step 1: Calculate expected scores.

For the 2400-rated player: Ea = 1 / (1 + 10^((2200 - 2400) / 400)) Ea = 1 / (1 + 10^(-0.5)) Ea = 1 / (1 + 0.316) Ea = 0.76

For the 2200-rated player: Eb = 1 - 0.76 = 0.24

Step 2: Calculate new rating if the 2400-rated player wins.

New rating = 2400 + 10 x (1 - 0.76) New rating = 2400 + 10 x 0.24 New rating = 2400 + 2.4 New rating = 2402

A win against a much lower-rated opponent only gains about 2 rating points. This is expected: beating someone you were supposed to beat is not surprising.

Step 3: Calculate new rating if the 2400-rated player loses.

New rating = 2400 + 10 x (0 - 0.76) New rating = 2400 + 10 x (-0.76) New rating = 2400 - 7.6 New rating = 2392

A loss against a lower-rated player costs almost 8 rating points. The system is asymmetric by design: unexpected losses sting more than expected wins reward.

Step 4: Calculate the 2200-rated player’s change if they win (the upset).

New rating = 2200 + 20 x (1 - 0.24) New rating = 2200 + 20 x 0.76 New rating = 2200 + 15.2 New rating = 2215

The 2200-rated player, with K = 20, gains 15 points for the upset. This is how dramatic upsets move players’ ratings quickly.

FIDE Rating Lists: Publication and Floors

FIDE publishes official rating lists monthly. This has been the case since January 2012. Before that, lists came out quarterly (and before 2009, just twice a year).

Monthly publication matters because it means a strong tournament result shows up in the standings within weeks rather than months.

Key rules about FIDE ratings:

• Minimum to be rated: 5 games against FIDE-rated opponents, all in the same rating period
• Rating floor: 1000. No player’s rating can drop below 1000, no matter how many games they lose.
• Live ratings: FIDE and third-party sites like 2700chess.com track ratings in real time between official lists. Live ratings are not official but are widely used.

Photo: FIDE, Public Domain via Wikimedia Commons

The India Effect: Why K = 40 Explains a Generation

India’s chess explosion is not just about talent. The mathematics of the K-factor has accelerated it.

When a young Indian player like Gukesh Dommaraju or R Praggnanandhaa enters the rating system at 1600 and faces stronger opponents, every result counts double or more compared to an established 2500-rated player. K = 40 means a single upset against a 2200-rated player can yield 30 or more rating points in one game.

Gukesh’s rise from roughly 2400 in early 2020 to his peak of 2794 is extraordinary by any measure; his current March 2026 rating of 2748 reflects a difficult recent stretch at Tata Steel and Prague. But the K-factor math explains much of the early trajectory. Once he crossed 2400 and dropped to K = 10, gains slowed to their natural pace, requiring sustained excellence rather than individual upsets.

“Reaching 2700 was a validation that my results over many months were genuine,” Gukesh said after crossing the super-grandmaster threshold in 2022. “You cannot inflate your rating at that level. Every point has to be earned against the best.”

Praggnanandhaa’s trajectory was similarly explosive under K = 40, as was that of Arjun Erigaisi. You can read more about India’s chess generation in our profile of India’s chess golden generation.

See all current Indian player ratings on our players page.

How the 2026 Candidates Tournament Affects Ratings

The FIDE Candidates Tournament is the most rating-intensive event most elite players play in a given year. Follow the Candidates 2026 live standings as ratings shift in real time. With 14 rounds against opponents all rated 2700 or above, the expected scores are tightly clustered near 0.5 for every game.

This means nearly every result is a surprise in one direction or the other. A win earns 5 rating points (K = 10, unexpected score of +0.5 in many matchups). A loss costs nearly the same.

Over 14 rounds, a player who goes +3 (8.5/14) against a field averaging 2740 could gain 20-25 rating points on a single event. A collapse to -3 (5.5/14) against the same field would cost a similar amount.

Magnus Carlsen’s participation in the 2026 Candidates adds an element of volatility at the very top of the rating list. Playing 14 games against a 2830-rated opponent shifts expected scores significantly for everyone in the field.

For more on how the Candidates Tournament qualification works, see our guide to Candidates Tournament history and format.

Common Misconceptions

“A higher K-factor always benefits you.” False. K = 40 means faster gains if you outperform, but also faster drops if you underperform.

Young Indian players who had spectacular early results gained quickly. Those whose early results were mixed saw ratings stall or drop just as fast.

“You gain rating by playing many games.” False. You gain rating only by outperforming your expected score.

Playing 100 games against opponents you are expected to beat 50% of the time and going exactly 50-50 leaves your rating unchanged.

“Draw with a 2800-rated player means big gains.” It depends entirely on your own rating.

A 2750-rated player drawing Magnus Carlsen at 2830 expects roughly 0.39, so a draw (0.5) is a slight outperformance worth about 1 rating point with K = 10. A 2200-rated player drawing a 2800-rated player expects nearly 0 and gains enormous points.

The Limits of the Elo System

The Elo system assumes that a player’s strength is constant and normally distributed. Neither is entirely true.

Players improve, decline, have good and bad tournaments, and play differently against different opponents.

This is why FIDE and various researchers have proposed supplements: performance ratings (your Elo for a single event), expected score adjustments, and more sophisticated systems like Glicko that account for rating uncertainty.

For now, the Elo system remains the standard. After more than 50 years, its elegance has proven more durable than its limitations.

For a deeper look at how Viswanathan Anand navigated the rating system across five world championship cycles, see his full player profile.

Sources: FIDE Handbook, FIDE Rating Regulations (effective January 2022), Chessmetrics research by Jeff Sonas.