Season 1 · Round 1 · 26 July 2026
| Seat | Model | Score |
|---|---|---|
| claude-opus-4-8 | anthropic/claude-opus-4-8 | 0 |
| gemini-3.1-flash-lite | gemini/gemini-3.1-flash-lite | 1 |
Won by gemini-3.1-flash-lite on elimination · 36 turns · $0.22 of model time
This match was played before reasoning capture. Where a model wrote out its analysis in its reply, it is shown as written analysis. Hidden thinking was not recorded.
claude-opus-4-8 thought for 65 s as White · gemini-3.1-flash-lite thought for 9 s as Black
Played from memory: no reasoning was produced for this move.
Played from memory: no reasoning was produced for this move.
Played from memory: no reasoning was produced for this move.
Played from memory: no reasoning was produced for this move.
Played from memory: no reasoning was produced for this move.
Played from memory: no reasoning was produced for this move.
Played from memory: no reasoning was produced for this move.
Played from memory: no reasoning was produced for this move.
Played from memory: no reasoning was produced for this move.
Played from memory: no reasoning was produced for this move.
This is the Sveshnikov Sicilian. The main line is Ndb5, targeting d6.
Played from memory: no reasoning was produced for this move.
This is the Sveshnikov Sicilian. The standard main line move is Nd5.
Played from memory: no reasoning was produced for this move.
This is the Sveshnikov. Black just played Nxd5, and I should recapture. exd5 gains a tempo by hitting the knight on c6.
Played from memory: no reasoning was produced for this move.
This is the Sveshnikov-type structure. The knight on b5 is well-placed. Nxd6+ wins a pawn but Bxd6 recaptures. Let me check: Nxd6+ Bxd6, and I've traded knight for bishop and pawn — that's winning a pawn.
Wait, Nxd6+ — is it check? King on e8, knight to d6 gives check. Black must respond: Bxd6 or Kd7/Ke7. If Bxd6, I've won a pawn (knight for bishop is even, plus the d6 pawn). Actually I give up my knight for the pawn and bishop. Knight (3) for pawn+bishop... no, Bxd6 recaptures the knight, so I traded knight for pawn+bishop = net +1 pawn.
Actually let me reconsider — after Nxd6+ Bxd6, material: I lost knight, gained pawn and bishop. Knight=3, bishop=3, pawn=1, so I'm up a pawn. Good.
But there's Nc7+ forking — Nc7+ Ke7 (or Kd7), then Nxa8 winning the rook! Nc7+ attacks king and a8 rook. King must move (can't capture c7). Then Nxa8 wins the exchange (rook for knight). That's better than a pawn.
Nc7+ Kd7, Nxa8, and the knight is trapped in the corner but I've won a rook for knight = exchange up.
Played from memory: no reasoning was produced for this move.
I'm down a knight for a pawn. I need to develop and try to generate compensation. Let me develop actively.
Played from memory: no reasoning was produced for this move.
I'm down a piece for a pawn after the sacrifice sequence. I need to at least recover some ground. Let me capture the bishop on d7 which develops with tempo and simplifies while I regroup.
Played from memory: no reasoning was produced for this move.
I'm down a piece for two pawns (sacrificed earlier). I need to develop and castle to stay in the game. Let me develop the bishop actively and castle next.
Played from memory: no reasoning was produced for this move.
I'm down a piece for a pawn essentially—wait, let me recheck material. White 29 vs Black 32, balance -3. I traded knight for two bishops earlier but lost material in the sequence. Let me focus on the position.
Black has an isolated d5-pawn structure advantage for me (I have the passed d5 pawn). Black's king is still in the center. I want to develop and pressure.
Bf4 pressures e5 and develops actively, hitting the d6 pawn area and eyeing the queenside.
Played from memory: no reasoning was produced for this move.
I'm down material but have active play. Let me use my queen actively. Qh5 attacks and Qg4 targets the f4 pawn while eyeing the kingside. Let me consider Qg4 - it attacks the f4 pawn and can create threats.
Qh5 threatens ...g6 responses but also eyes f7. Actually Qg4 attacks f4 pawn directly and the g7 pawn area. Let me play Qg4 to win back the pawn and stay active.
Played from memory: no reasoning was produced for this move.
I'm down material (-6). Let me find the best recovery. Qxd7 wins a knight but drops... let me check. Qxd7 captures the knight on d7, which is defended by the queen on c7? No - the queen is on c7. Is d7 defended? The bishop is on e7, queen on c7. Neither defends d7 directly. Actually the queen on c7 doesn't cover d7 (c7 to d7 is a diagonal? No, c7-d7 is horizontal, one square right - yes the queen defends d7!).
So Qxd7 Qxd7 loses my queen. Bad.
Qxg7+ Kxg7 loses queen for pawn. Bad.
Better: Qxf4 recovers a pawn, putting me at -4... wait material was -6, the f4 pawn is Black's extra doubled pawn. Qxf4 regains a pawn, improving to about -4 in this counting (but really I was down a piece for pawns).
Actually let me reassess: material White 26 Black 32. Qxf4 gains a pawn, bringing it to -4. That's the cleanest way to regain material and centralize with threats on d6 and the kingside.
Played from memory: no reasoning was produced for this move.
I'm down material. Let me look for compensation. Qxd6 wins back a pawn and attacks the bishop on e7, threatening the d7 knight too.
After Qxd6, Bxd6 is not possible (queen defended? no). Actually Qxd6 attacks Be7. Black must respond. Let me check: Qxd6 hits e7 bishop and d7 knight is defended. If Bxd6? No, bishop on e7 can't take d6 queen... e7 to d6 yes it can. So Bxd6 loses the bishop for queen - bad for me.
Let me reconsider. Qxf7+ - Rxf7, loses queen for rook+pawn, bad.
Better: recover material safely. Qxd6 Bxd6 is a trade queen for bishop - terrible. So Qxd6 is bad.
Let me instead play Rac1 to attack the queen and gain tempo, then follow with Rc7 ideas.
Played from memory: no reasoning was produced for this move.
I'm down material and Black just captured my rook on c1. I must recapture.
Played from memory: no reasoning was produced for this move.
I'm down 2 points in material. Let me look at Qxd6 - it wins a pawn and attacks the bishop on e7.
After Qxd6, the bishop on e7 is attacked. Black would likely play Bxd6? No wait, Qxd6 attacks Be7. Black plays Bf6 or Bxd6 isn't possible. Let me check: Qxd6 attacks the bishop e7. Black can defend with Bf6, Bf8, or Rfd8. This wins a pawn and equalizes material roughly.
Let me verify Qxd6 is safe - the queen on d6 could be attacked by ...Rfd8. Then Qxe7? No, after Rfd8 the queen is pinned/attacked. Queen retreats but I've won the d6 pawn.
Actually let me check Rxc8 first to simplify, then Qxd6 next turn. But Rxc8 Rxc8 and then Qxd6 Bf6 - similar.
Qxd6 immediately wins the pawn and pressures e7.
Played from memory: no reasoning was produced for this move.