TrueElo scores a chess position by what players at your rating and clock actually convert it into, not by what perfect play is worth. Every true score is read against 2B+ rated Lichess games played December 2023 to November 2025, split across 9 rating brackets and 5 time controls, with win rate, draw rate and loss rate.
Engine eval assumes both sides play near-perfectly. At 1400 in blitz, nobody does. Where the two numbers disagree is where your rating is.Engine evaluation assumes both sides play near-perfectly. At 1400 in blitz, nobody does. Openings that are "unsound" score brilliantly because the refutation is hard to find with three minutes on the clock — and "correct" main lines lose because the plans are hard to play. TrueElo shows you both numbers, and where they disagree is where your rating is.
An engine tells you what a perfect player would get from a position. You are not that player, and neither is your opponent. We read the same position against billions of rated Lichess games at your level and your clock, and tell you what players like you actually score from it.
That is your True Score. The engine’s own answer — the Theory Score — sits beside it, never instead of it, because the two disagreeing is the whole point: a position the engine calls lost can be one you win most of the time.
Sign in to set a position, pick your level and your clock, and compare the moves with uncertainty taken into account. Star the lines you want and drill them until you play them without thinking.
The opening dataset counts rated games from the Lichess Open Database, published into the public domain. The counts below describe those games. True Score combines game outcomes with Stockfish; Confidence Bands show the uncertainty in a score.
Each figure also carries the range it sits in, 90% confident. A move that won 60% of fifty games and one that won 55% of five thousand are not the same claim, and the range is what keeps the first from outranking the second.