The two-engine era ended quietly in TCEC Season 29
Stockfish won again — its 19th title, its 12th in a row. The result everyone expected. The part worth noticing is who it beat.
Stockfish took the Top Chess Engine Championship Season 29 Superfinal 59–41, its 19th title overall and its twelfth consecutive season. On the scoreboard that is the least surprising sentence in computer chess. The engine has not lost a season since 2020.
The unexpected part is the name on the other side of the table. For eleven straight seasons, the Superfinal was Stockfish against Leela Chess Zero — the handcrafted-search tradition against the neural-network one, a rivalry stable enough that the final two slots stopped being a question. In Season 29 that pairing did not happen. The challenger was Reckless, described by Chessdom as a relative newcomer that reached the final on an aggressive, unconventional style.
It did not win, and the margin was not close. But it was not passive either: 64 of the 100 games were decisive, with Reckless taking 23 of them. A challenger that loses 41–23 while dragging two thirds of the match out of drawing territory is a different kind of opponent than one that holds draws and loses narrowly.
This is the structural problem every mature machine competition eventually hits. When the field converges on one dominant approach, the result stops carrying information — the title tells you which engine is best, but the games stop telling you why. Engine chess has been running against that ceiling for years: the Superfinal is a hundred games of near-perfect play, most of them drawn, decided by openings chosen for imbalance by the organisers rather than by the engines.
Which is roughly why the interesting experiments have moved sideways. The Kaggle Game Arena exhibition in August 2025 put general-purpose language models — not chess engines — across a board from each other, with Magnus Carlsen and Hikaru Nakamura on commentary, and OpenAI's o3 beat Grok 4 4–0 in the final. The chess was, by engine standards, poor. It was also watchable, because the players were unreliable in interesting ways.
Season 29 makes both halves of that visible at once. A field can be strong enough that its results are certain, and a field can be weak enough that its results are noisy. The thing worth building a season around sits between the two — and that is the gap this site is going to spend its time in.