Didier Leclaire vs Jan Stornebrink: Build the Mating Net
Three measured decisions show how move order, development, and forcing checks turn an exposed king into a finished attack.
Reproducible analysis run complete
6 saved models on NVIDIA GeForce RTX 5080 · Stockfish 18 · 250,000 nodes per checked line.
Why this game?
Make every attacking piece count
This rapid game begins with a Vienna structure and quickly becomes a lesson in attack. Jan gives up a rook to pull the white queen away, brings the last bishop into play, and then chooses a forcing check over an easy capture. At each position, compare the tempting material move with the move that keeps the king under pressure.
- White
- Didier Leclaire
- Black
- Jan Stornebrink
- Event
- Casual rapid game, Chess Club Enschede Inclusief member collection
- Result
- 0-1
Interactive comparison
Choice one: change the move order
Black to move after 18.Be3. The bishop on e3 is available, but White's queen can recapture if Black takes it immediately.
r3kb2/ppp2p1p/4bpr1/4p3/4P3/1PPqBP2/4NQPP/R3K2R b KQq - 1 18Model candidates
Choose a viewpoint
V6 Active: Qxe3
Measured GPU choice, rejected by Stockfish. Taking the bishop first allows White's queen to recapture and turns the position in White's favour.
Model top three: Qxe3 · Qxe2+ · Rxg2
Stockfish 18 safety gate
CompleteStockfish prefers the played 18...Rxg2 and gives Black about +2.91 from the side-to-move view. The immediate 18...Qxe3 falls 647 centipawns behind in its independent check.
- Engine move
- Rxg2
- Depth
- 16
- Node budget
- 250,314
- Evaluation
- +2.91
Verification line
18...Rxg2 19.Qxg2 Qxe3 20.Qf2 Qxf2+ 21.Kxf2 Bc5+ 22.Kg3 Bxb3 23.Nc1 Bc4
Position lesson
The move order is the whole tactic. The rook first takes on g2 and draws White's queen away from e3. Only then can Black take the bishop while keeping the attack alive.
Interactive comparison
Choice two: bring the last bishop
Black to move after 22.Kxd1. The white king is exposed, while Black's bishop on f8 is still outside the attack.
4kb2/ppp2p1p/4bp2/4p3/1P2P3/2P1qP2/4N1QP/3K3R b - - 0 22Model candidates
Choose a viewpoint
V6 Active: Qxe2+
Measured GPU choice, rejected by Stockfish. White's queen on g2 can answer Qxe2+, so the tempting knight capture gives away the attack.
Model top three: Qxe2+ · Bxb4 · Qxe4
Stockfish 18 safety gate
CompleteStockfish prefers the played 22...Bh6 and gives Black about +4.94. The six-model choice 22...Qxe2+ loses 1,054 centipawns against the best line.
- Engine move
- Bh6
- Depth
- 17
- Node budget
- 250,952
- Evaluation
- +4.94
Verification line
22...Bh6 23.Ke1 Bc4 24.f4 Bxf4 25.Qg4 Ke7 26.Kf1 Qxe4 27.Rg1 Qd3
Position lesson
22...Bh6 develops with a direct purpose. The bishop can enter on c4, the queen stays active on e3, and White must solve threats from several directions. Grabbing the knight lets White trade queens and escape.
Interactive comparison
Choice three: check before material
Black to move after 25.Nc1. The bishop can take the knight, but the queen also has a forcing check on c3.
4k3/ppp2p1p/5p1b/4p3/1Pb1P3/2Pq1P2/5Q1P/2N1K2R b - - 6 25Model candidates
Choose a viewpoint
V6 Active: Bxc1
Measured GPU choice. Stockfish confirms that 25...Bxc1 still forces mate, so it passes the safety gate even though 25...Qxc3+ is much faster.
Model top three: Bxc1 · Qxe4+ · Qxf3
Stockfish 18 safety gate
CompleteStockfish finds mate in two with 25...Qxc3+. The recorded game continued 26.Kd1 Qxc1#. The model move 25...Bxc1 also forces mate, but in six moves against best defence.
- Engine move
- Qxc3+
- Depth
- 245
- Node budget
- 30,862
- Mate
- M2
Verification line
25...Qxc3+ 26.Qd2 Qxd2#
Position lesson
Checks come first. 25...Qxc3+ forces mate in two against best defence. The model choice 25...Bxc1 also wins, but it needs a longer mating route and gives White more chances to resist.
Published finish
The forcing line
This is the recorded finish from the public game score. Black uses a rook sacrifice, activates both bishops, and ends with a queen checkmate on c1.
Review method
Source, models, then verification
Start from the public score
Replay the complete legal game from the club's member page and select three connected positions from the final attack.
Measure six saved evaluators
Rank every legal move on the RTX 5080 with the fixed 30% hybrid positional weight. The saved checkpoints remain unchanged.
Verify every distinct choice
Stockfish 18 checks its best line, the shared model choice, and the played move independently with 250,000 nodes per line, two threads, and 256 MB hash.
What this review can and cannot show
Clear boundaries keep a useful experiment from becoming a bigger claim than the evidence supports.
- The source identifies the players by the handles didierleclaire and JanRobertS. The public page and Jan's message support the display names used in this review.
- The PGN header says casual rapid game and gives the date and result, but it does not name a platform, round, location, or time control. This review does not guess those details.
- Only the factual game score is reused. The teaching explanations are original and focus on observable board features.
- All displayed evaluations use the side-to-move perspective. Positive values favour the player whose turn it is in the reviewed position.
- The six neural checkpoints are scalar evaluators, not move-policy models or explanation models.
- Their 768 inputs encode piece placement only; side to move is applied after inference, while castling rights, en-passant state, repetition, move history, and clocks are not model inputs.
- Profile labels describe intended training or runtime behaviour and do not prove human-like personality or internal understanding.
- Stockfish results use a fixed node budget and configuration and are a verification snapshot, not a claim of perfect play.
- Stockfish used two search threads. Parallel search can produce small score, depth, node-count, or principal-variation differences on a fresh rerun.
- The analysis does not measure Elo, playing strength, or clean unseen test accuracy.
- The prose explains observable chess consequences; it does not claim to reveal internal 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.