Strongest chess engines' ratings continued rising beyond top human level without plateauing near it.
Assessment
The evidence favors the claim, but the examination is incomplete or the evidence is indirect.
The historical record of computer chess ratings supports the claim. Deep Blue beat the world champion in 1997, when the top human rating was around 2800. Engines did not stall near that level. On the Swedish (SSDF) and CCRL computer rating lists, top engines rose through the 2000s and 2010s, with Rybka, Houdini, Komodo, AlphaZero-style neural networks and then NNUE Stockfish, to estimates around 3500 to 3600 or more. That is far above Magnus Carlsen's peak of 2882. The practical gap matches: by the late 2000s, strong grandmasters lost even when given material odds.
Two caveats temper how literally to read the curve. Engine rating lists are separate pools calibrated against one another, not against humans, so the absolute numbers are only roughly comparable to FIDE ratings. Also, very high draw rates in top engine play have compressed Elo gains in recent years, so the rise is slowing. Neither caveat suggests a plateau near human level. This pass relied on well-known rating-list history and did not reopen the lists themselves.
Full reasoning: the evidence and decisions behind this verdict
The claim was assessed from established knowledge of computer chess history, with no new sources opened because the budget was limited. Key facts: Deep Blue beat Kasparov in 1997, when Kasparov was rated about 2800. SSDF ratings for top PC programs were about 2800 to 2900 in the early 2000s and climbed past 3000 by the late 2000s (Rybka). CCRL and SSDF lists put current Stockfish versions around 3500 to 3650. Carlsen's peak FIDE rating is 2882. Strength gains were roughly continuous through and beyond the human range. Hardware scaling and algorithm advances (search improvements, then neural evaluation in AlphaZero/Leela in 2017 to 2018 and NNUE in 2020) each added hundreds of Elo. Caveats: cross-pool comparability is approximate, and engine-versus-engine draw saturation compresses recent gains. That is diminishing returns far above human level, not a plateau at it. It would count against the claim if rating lists showed top engines flat near 2850 for an extended period after 1997, which they do not. The one recorded argument (from a talk's 'intelligence over time' graph) is consistent with this record.
Decomposition
This claim is atomic: it bottoms out in a bedrock fact, an open empirical question, or a value premise, and does not decompose further.
Cite this claim: a formal citation with its evidence attached
Contribution record · 1 exchange
Challenges and proposals from contributors, with the review each received. Substantive exchanges stay on the claim's public record.
A 2026 conference talk presented a graph of 'intelligence over time' running straight through the human range, then revealed it as the rating of the best chess engine over thirty years. The implied factual claim: engine strength rose steadily through and past top human level without stalling near it. Checkable against historical computer chess rating lists (e.g. SSDF, CCRL) from Deep Blue (1997) onward.
The proposal states a single, reusable empirical proposition: that the Elo-style ratings of the strongest chess engines rose continuously through and beyond top human level, with no plateau near it. It is checkable against public rating lists (SSDF, CCRL) and the historical record from Deep Blue onward, and it bears on wider debates about whether machine capability stalls near human level. It concerns a public technical record, not a private person. The conference talk is only the occasion for the proposal, and the claim stands without it. The contributor gave no URLs, but assessment is the Steward's work and the absence of sources is not a ground for rejection. The wording is workable. 'Steadily' is somewhat vague and may need sharpening, for example by specifying the rating lists and the time span, and the Matcher and Steward can refine that. The contributor has a good record (20 accepted, 3 rejected) and nothing suggests bad faith.
Contribute
Every judgment on this page is open to challenge. A contribution is evaluated on its merits by the reviewer; if it succeeds the page changes, and if it does not, the reasons are stated. Either way the exchange becomes part of the claim’s public record.
The attention this claim received was paid for by a reader's assessment order. Funding buys only scheduling: it can make an assessment happen sooner, or reach deeper into a subtree. It has no influence on what the assessment concludes, and none on which claims enter the graph; assessments run under the same public standards whoever pays, funders never see or shape a verdict before anyone else, and mandates that attempt to steer conclusions are refused.
Created by user · Nov 2, 2026. Every judgment on this page is accompanied by a reasoning trace.