Assumes a player’s performance is a normal distribution. The difference in rating predicts the expected score. For every 400 points of difference, the stronger player is expected to score 10x more wins. A 2800 human vs. a 3600 engine has a theoretical win probability of near zero (approx. 0.8%).
| Player Type | Approx Elo/Glicko | Key Trait | | :--- | :--- | :--- | | Club Player | 1200–1800 | Tactical blindness | | FIDE Master (FM) | 2200–2300 | Strategic planning | | Grandmaster (GM) | 2500–2600 | Deep intuition | | World Champion | 2800–2850 | Positional perfection | | | 3600+ | Zero tactical errors, perfect endgame | chess emis ge
Used by most online servers (Chess.com, Lichess). It adds a crucial variable: Rating Deviation (RD) . RD measures reliability. A new engine with few games has high RD; after hundreds of games, RD shrinks. This prevents wild rating swings and handles inactive players fairly. The Gap: Human vs. Machine (2800 vs. 3600) The gap is so vast that the metaphor "Man vs. Machine" is obsolete. It is now "Mortal vs. Oracle." Assumes a player’s performance is a normal distribution
The Elo/Glicko number is no longer just a score. For engines, it is a measure of absolute tactical truth. The human world champion (2800) plays at 99% accuracy; the top engine (3600) plays at 99.999% accuracy. In that decimal lies the end of human supremacy—and the beginning of a new era of learning. Want to see the live rating list? Visit the Computer Chess Rating Lists (CCRL) or TCEC archive. As of 2025, Stockfish, Leela, and Dragon hold the top spots, separated by just 10–20 Elo—a razor’s edge in silicon terms. A 2800 human vs
Oops! We could not locate your form.