In Stockfish vs Magnus Carlsen, Stockfish wins. As of October 2026, Stockfish 17.1 on one CPU is rated 3771 on the CCRL Blitz list and Stockfish 19 on 8 CPUs is rated 3804, while Carlsen, the world's top-rated player, is 2823 on the FIDE classical list. Taken at face value, the Elo formula gives Carlsen an expected score of about 0.4%: roughly one point (one win or two draws) every 235 games. The two numbers come from separate rating pools, so treat that figure as an illustration of the gap, not a forecast.
What is Stockfish's Elo?
Stockfish has no single official Elo. Engine rating lists play engines against each other under fixed conditions and publish their own scale. The CCRL Blitz list, computed October 17, 2026, uses a time control equivalent to 2 minutes + 1 second on an Intel i7-4770K, with a 12-move opening book and 6-piece tablebases.
| Engine | Hardware | CCRL Elo |
|---|---|---|
| Reckless 0.9.0 | 8 CPU | 3810 ±29 |
| Stockfish 19 | 8 CPU | 3804 ±22 |
| Stockfish 19 | 1 CPU | 3793 ±11 |
| Stockfish 17.1 | 8 CPU | 3786 ±10 |
| Stockfish 17.1 | 1 CPU | 3771 ±9 |
| Lc0 0.29.0 (w753723) | RTX 2080 GPU | 3696 ±10 |
Two points stand out. First, the top of the list is crowded: Reckless 0.9.0 sits 6 points above Stockfish 19, well inside the error bars, so "strongest engine" is a statistical tie at that level. Second, a single CPU costs Stockfish only about 15 points here. Our analysis board runs Stockfish 17.1 lite, single-threaded, in your browser, which is a smaller network than the full build on that list, so it is weaker than 3771 but still far beyond any human.
The Stockfish blog dates the releases: Stockfish 17.1 on 2025-03-30, Stockfish 18 on 2026-01-31 and Stockfish 19 on 2026-09-05.
Stockfish vs Magnus Carlsen: the Elo math
Carlsen is rated 2823 on the September 2026 FIDE classical list, ranked number one (Chess.com report on the September 2026 list). That is the latest list we could confirm in October 2026; we could not load his FIDE profile directly.
The Elo expected score for a player rated R_A against an opponent rated R_B is:
E = 1 / (1 + 10^((R_B − R_A) / 400))
| Opponent | Rating gap | Carlsen expected score | About one point every |
|---|---|---|---|
| Stockfish 17.1, 1 CPU (3771) | 948 | 0.43% | 235 games |
| Stockfish 19, 8 CPU (3804) | 981 | 0.35% | 284 games |
Method: we plugged the CCRL rating and Carlsen's FIDE rating into the formula above in Node and rounded. Why it is only an illustration: CCRL and FIDE are separate pools. CCRL engines only play engines and FIDE players only play humans, so no games connect the two scales, and 3771 CCRL does not mean "3771 FIDE". CCRL also uses blitz conditions, not classical. The real gap could be larger or smaller; what no one disputes is its direction.
Can a human beat Stockfish?
Not in a fair, full-strength game today. The last well-documented top human vs Stockfish match was already a handicap match. In August 2014 Hikaru Nakamura, then rated 2798, played Stockfish 5 without its opening book or tablebases. In games 1–2 Nakamura had help from an older Rybka engine; in games 3–4 he got pawn odds and White. Stockfish won each half 1.5–0.5, 3–1 overall (Wikipedia: Stockfish). Stockfish has gained a great deal of strength since version 5.
Humans do beat Stockfish in two situations:
- Reduced strength. The "Skill Level" and Elo-limit settings make Stockfish play deliberately weaker moves. That is how most app bots work. Our Stockfish levels Elo chart measures what each of Lichess's eight levels plays at.
- Very short searches. At tiny time limits an engine can miss deep ideas, though even our browser build reaches depth 22–23 in one second in normal positions (see the measurement in Stockfish vs Leela).
What Stockfish's strength means for your analysis
Because the engine is so far ahead, its evaluation is the reference, not a suggestion. When the analysis board shows a move costing 2 pawns, a human at any level will not find a hidden reason it was fine. The useful question is why the engine's move is better, which is what how to read chess engine evaluations explains.
You can see that gap in your own games:
- Run Game Review. It labels each move Best, Good, Inaccuracy, Mistake or Blunder against Stockfish 17.1. See blunder, mistake and inaccuracy.
- Check your accuracy score. It measures how close you stayed to engine play; chess accuracy explains the formula.
- Ask the coach for the plan. The engine gives a move; a plan explains it.
- Learn what powers the number. Stockfish NNUE covers the neural network inside the engine.
Frequently asked questions
What is Stockfish's Elo?
On the CCRL Blitz list computed October 17, 2026, Stockfish 17.1 on one CPU is 3771 and Stockfish 19 on 8 CPUs is 3804. Those are engine-pool ratings, measured against other engines, and do not translate directly to FIDE.
Can a human beat Stockfish?
Not at full strength under fair conditions. Even in 2014, Hikaru Nakamura lost 3–1 to Stockfish 5 while receiving engine help or pawn odds. Humans beat Stockfish when it is set to a reduced skill level.
Has Magnus Carlsen ever played Stockfish?
We found no officially documented Carlsen vs Stockfish match that we could verify. The Elo comparison above is a calculation, not the result of real games.
Is Stockfish the strongest chess engine?
It is at the top, but not alone. On the October 2026 CCRL Blitz list Reckless 0.9.0 is 6 points ahead of Stockfish 19, inside the margin of error, and Stockfish won TCEC Season 29 in April 2026. Compare it with neural-network engines in Stockfish vs Leela.
Want to test it yourself? Play any position in the analysis board, then review a full game in Game Review. For the wider tool landscape, see chess AI tools.