David's board

Distilled 2026-07-17 from 10 engine-analyzed rapid games; updated 2026-07-27 after both games against James that day, a Caro-Kann as White lost in fourteen moves and a Dutch as Black lost in forty-three. Nine straight losses to James. The Dutch is the useful one: the first game in weeks to reach an endgame, and he held it level for fifteen moves before three king moves ended it.

Style & strengths

• The family's precision player once the position is built: ~89% average accuracy in the bootstrap set. Middlegames are well-constructed when he reaches them intact.
• Endgame understanding, no longer untested. Upgraded 2026-07-27. In the Dutch he took a position that was +1.1 for White at move 23 back to +0.0 and kept it there through move 39, a rook and bishop ending with pawns on both wings. The loss came after that, not from the technique.
• Dangerous with the initiative as White in sharp openings. The King's Gambit is the weapon of choice, and the French Exchange has reached +3.6 by move 9.
• Creative with the exchange sacrifice. 16.Rxd6 (07-24) gave rook for bishop to keep the attack alive on James's stranded king, holding +2.8, a genuine find.
• Openings are not the leak. The Caro-Kann Classical had him at +0.2 with everything developed and the king castled at move 11 (07-27). Both losses that day were decided by a single middlegame or endgame decision.
• Resourceful under pressure, sets problems rather than collapsing.

Leaks the engine keeps flagging

1. The capture nearest the hand. Still the most expensive item on file. Grabbing material without asking what the square gives away. Newest and clearest: 39...Kxg4 (07-27) took a pawn with the king from a dead level position, onto a square where Bd1+ and Rg1+ were both immediately available, +0.0 to +2.6 in six seconds, and the game was mate four moves later. Earlier: 16...Qxe5 (07-23) took a pawn under direct attack when Ng4 was the defence; 22...Nxc4 (07-22) grabbed a pawn while a bishop hung on e4. Before a capture, name the checks the new square gives them.
2. Attacking on the wing where his own king is the exposed one. Widened 2026-07-27 from "king safety as Black" to colour-independent. 11.h4 as White, one minute forty, opened a flank attack and left a2 loose beside his own castled king on c1, +0.2 to -1.9; Kb1 defends a2 with the king and keeps the same plan. As Black, 10...O-O-O (07-23) castled into White's queenside attack and was cracked open by 14.Bxb7+. Before attacking, check the squares next to your own king.
3. The queen trade, in both directions. The keep-queens-on note was written for attacking: when their king is stuck, the queens are his best pieces. The converse now has evidence. Defending on 07-27 he passed on Qb1 twice in three moves, at 13.h5 (-0.7 to -3.5) and 14.Qc2 (-3.5 to -5.8); both times Qb1 turned a king hunt into a one-pawn deficit. Attacking, keep them on. Defending, the trade is the escape.
4. The piece with no good square. Confirmed again 07-27: 21...Ng4 in the Dutch left f6, opened the bishop on c3 onto g7, forced 22...g6, and the knight returned to f6 on move 23 having achieved nothing, -0.2 to +1.0. Same as the 07-22 knight that wandered Nc5-Nd7-Nb6 and was lost, and the 07-21 Caro-Kann bishop shoved around four times in six moves. A piece with no future is one to trade or bring home.
5. The forcing move when winning. Over +2, the natural move instead of the one that ends it (missed Nxg6 twice, 07-24). Untested since; he has not been ahead. Over +2, "what's forcing?" beats "what's better?"
6. Passive defence when worse. 17...Qb8 (07-23) retreated when Ng4 fought; 28...Rh8 over the active Rg6 (07-20). Defending, activity beats safety.

Clock, restated 2026-07-27

Previously "clock inversion: the long thinks produce the loose moves." That still holds and got its cleanest instance in 1m40 on 11.h4, the longest think of that game, ending on a flank pawn push that dropped a pawn at home. Also 2m25 on 20...Rb8 (07-23) and 1m59 on 17...Qb6 (07-22).

But the Dutch cuts the other way and the item needs both halves. The three moves that lost that game took six seconds (39...Kxg4), two seconds (40...Kh5) and seven seconds (42...Ra7), at the exact point a level endgame turned. So the finding is not "long thinks are bad." It is that the time spent is not tracking how much the position has just changed. A capture that changes the pawn structure, a king move in an endgame, and a long think ending on a quiet move are all signals to stop and ask what the opponent gets.

Working on

• Before a capture, name the checks the new square gives them the item that lost 07-27 outright.
• Before attacking, check the squares next to your own king.
• Defending, look for the queen trade before the next improving move.
• Spend time where the position turns, not where it feels complicated.