Currie vs Guzman: Space, Structure, and Safe Simplification
A Queen City Chess Club contribution becomes an interactive lesson about central tension, queenside structure, and choosing a safe route into a balanced ending.
Reproducible analysis run complete
6 saved models on NVIDIA GeForce RTX 5080 · Stockfish 18 · 250,000 nodes per checked line.
Why this game?
Small choices shape a balanced middlegame
Queen City Chess Club helped turn one of its 2026 championship games into this learning page. Hudson Currie and Sophia Guzman reach a balanced middlegame where the important choices are not combinations, but changes to pawn structure and piece activity. The three positions show how a quiet move, the correct capture, or a safe simplification can matter more than a forcing-looking idea.
- White
- Hudson Currie
- Black
- Sophia Guzman
- Event
- 2026 South Saskatchewan Championship, round 5
- Result
- 1/2-1/2
Interactive comparison
Choice one: gain space or resolve the centre
White to move on move 13. The six evaluators split between 13.cxd5 and 13.Nxd5. Currie played 13.c5, while Stockfish slightly prefers the quiet 13.b3.
r2q1rk1/ppn1bppp/2p2n2/3p4/2PP2b1/2N1BNP1/PP3PBP/R2Q1RK1 w - - 0 13Model candidates
Choose a viewpoint
V6 Active: cxd5
Measured GPU output. The engine gate accepts this central capture, although the played c-pawn advance keeps the same balanced character.
Model top three: cxd5 · Nxd5 · Re1
Stockfish 18 safety gate
CompleteStockfish gives Black only a 0.22-pawn edge after 13.b3. The played 13.c5 is also balanced, and the model choice 13.cxd5 passes the gate. Only 13.Nxd5 is clearly unsafe.
- Engine move
- b3
- Depth
- 15
- Node budget
- 250,291
- Evaluation
- -0.22
Verification line
13.b3 Qc8 14.Ne2 Re8 15.Ne5 Bh3 16.Nf4 Bxg2 17.Kxg2 Bd6 18.Nfd3 dxc4 19.bxc4 c5
Position lesson
The position is balanced, so pawn structure matters more than a quick material count. The advance c5 keeps queenside space. The central pawn capture is also safe, but the knight capture allows Black to rebuild the centre with tempo.
Interactive comparison
Choice two: choose the right queenside capture
White to move on move 20. The evaluators choose Bxd5, Rxb6, or Qxh7+. Currie instead played 20.cxb6, the same move preferred by Stockfish.
1r3rk1/p1q1bppp/1pp1n3/2Pp4/3P4/2P1BBP1/PRQ2P1P/1R4K1 w - - 2 20Model candidates
Choose a viewpoint
V6 Active: Bxd5
Measured GPU output. The capture changes the queenside in White's favour, but only the played pawn capture passes the engine gate.
Model top three: Bxd5 · cxb6 · Rxb6
Stockfish 18 safety gate
CompleteThe fixed-node gate confirms the played 20.cxb6 as best at about +1.31 for White. Every distinct model top choice falls more than four pawns behind in the verified lines.
- Engine move
- cxb6
- Depth
- 15
- Node budget
- 250,857
- Evaluation
- +1.31
Verification line
20.cxb6
Position lesson
Capturing with the c-pawn removes Black's a-pawn and leaves a clear target on b6. Rxb6 exchanges too many active rooks, Bxd5 lets Black repair the centre, and Qxh7+ gives away the queen for no attack.
Interactive comparison
Choice three: simplify without losing activity
White to move on move 25. Three models choose Qxh7+, two choose Rxb6, and one finds the played 25.dxc5. Stockfish prefers 25.Qe2 but accepts the game move inside the safety gate.
1r3rk1/6pp/1p1qp3/2pp4/3P4/2P3P1/PRQ2P1P/1R4K1 w - - 0 25Model candidates
Choose a viewpoint
V6 Active: Rxb6
Measured GPU output. The rook capture gives Black an active first-rank entry, so the fixed-node engine gate rejects it.
Model top three: Rxb6 · Qxh7+ · dxc5
Stockfish 18 safety gate
CompleteStockfish prefers 25.Qe2 at about +0.68, but the played 25.dxc5 trails by only 37 centipawns and passes the gate. The rook and queen sacrifices are both rejected.
- Engine move
- Qe2
- Depth
- 15
- Node budget
- 250,053
- Evaluation
- +0.68
Verification line
25.Qe2 Rfe8 26.a4 e5 27.dxc5 Qxc5 28.Qd2 d4 29.cxd4 Qxd4 30.Qxd4 exd4 31.a5 d3 32.Rxb6
Position lesson
The calm pawn exchange removes central tension and keeps both rooks active. Qxh7+ is not an attack because the king can take the queen. Rxb6 allows Black's rook to enter the first rank with tempo.
Published finish
The forcing line
This is the finish recorded on the contributed score sheet. The draw was agreed after 28.Qe2, with queens and pawns still on the board.
Review method
Source, models, then verification
Start with a contributed tournament game
The complete legal score supplies each exact position and played move, while the club page confirms the public event details.
Run six saved models on GPU
Every legal move is ranked by each scalar evaluator with the same 30% hybrid scoring formula.
Apply the Stockfish gate
Every distinct model choice and the played move receive an independent fixed-node verification.
What this review can and cannot show
Clear boundaries keep a useful experiment from becoming a bigger claim than the evidence supports.
- The public event page lists the tournament and players but does not publish this game score. The complete score came from the club contribution.
- The contributed sheet records a draw after 28.Qe2. It does not claim that the final board position was forced to be a draw.
- These checkpoints are lightweight scalar position evaluators, not move-policy, search, or explanation models.
- Their inputs encode piece placement only; they do not receive move history, clocks, repetition state, castling rights, or en-passant state.
- The profile labels describe intended training behavior, not proven human-like personalities or internal reasoning.
- Stockfish used a fixed 250,000-node budget per line with two threads; this is a reproducible snapshot, not a claim of perfect play.
- Two-thread engine search can produce small score, depth, node-count, and principal-variation differences between reruns.
- This review measures three positions from one game. It does not establish Elo, general playing strength, or unseen-data accuracy.
- The prose explains observable chess consequences, not hidden neural-model reasoning.
Continue learning
Put the position on your own board
Start from the exact FEN and test whether you can find and convert the idea against a human-style bot.