Elo turns a rating gap into an expected win probability, then nudges both players' ratings by K × (actual − expected). FIDE uses K=40 for new players, K=20 for established sub-2400s, and K=10 for anyone rated 2400+; the USCF and online sites use their own K schedules. Enter a single game or a batch of results and see where each rating moves.
| Player | Old | Δ | New | K |
|---|
The blue dot shows the current result. Curve = K × (S − Eₐ) for S ∈ {0, ½, 1}.
Arpad Elo devised the system in the 1950s to replace the USCF's earlier Harkness rating, and FIDE adopted it in 1970. The expected score is a logistic function of the rating gap: EA = 1 / (1 + 10(RB − RA) / 400), which puts a 400-point favourite at ≈ 91% expected. After the game each player moves toward their actual score by a factor of K — the "responsiveness" knob. Big K reacts quickly (helpful for new players still finding their level) but jitters an established rating; small K is stable but slow to catch a genuine improvement. FIDE's current schedule: K = 40 for players with fewer than 30 rated games and under age 18, K = 10 once you hit rating 2400+ at any point, K = 20 for everyone else. Modern derivatives (Glicko-2, TrueSkill) add rating uncertainty and volatility, but classical Elo remains the lingua franca of two-player skill rating.