# My chess improvement plan after week 31: castle first, then attack > Week 31: 985 → 985 in 6 rapid games (3 wins, 3 losses). Two losses where castling was the best move, and a plan drawn from 15 weeks of data. - URL: https://thechesscoach.netlify.app/journey/week-2024-w32/ - Updated: 2026-10-22 - Author: Ignacio Soler Week 31 (August 5–9, 2024) was tiny: **6 rapid games, 3 wins, 3 losses, and a rating that started and ended at 985**. Four of the games were on Monday, August 5. There isn't much week to describe, so I'm using this entry for something I'd been putting off: writing down a chess improvement plan based on my own numbers, not on general advice. Two of this week's three key moments point straight at the first item on it. ## The week in numbers | | | |---|---| | Games | 6, all 10-minute rapid | | Wins / losses / draws | 3 / 3 / 0 | | Rating | 985 → 985 (±0) | | Average accuracy | 81.2 % | | Blunders (Stockfish) | 12 (opponents: 11) | | Openings | Van 't Kruijs, Center Game, French, King's Knight, Zukertort Chigorin, Ruy Lopez (one each) | The good: game #777 against Fratkurd was a checkmate win in a French Defense with no mistakes and no blunders, only five inaccuracies, at 91.7 % accuracy. My three wins averaged 91.8 %. The bad: in two of the three losses (#775 and #779) I had at least an 80 % chance of winning. In #779 against fosenheimen my winning chance peaked at 85.6 % and I ended up checkmated after 51 moves. With White I was still mixing first moves: 1.e4 in two games, 1.d4 in one. ## Three key moments ### Not castling, and the queen check that followed Game #775, Black against spillz987 (1010), after 1.e3. I was slightly better, −0.61, with my bishop on b4 and my king still on e8. > **Move 6: my bishop on b4 has no defender and my king is on the e-file** > Position (FEN): `rnbqk2r/ppppnppp/8/3P4/1b6/2N5/PPP1QPPP/R1B1KBNR b KQkq - 0 6` > Played: d6 — Pawn from d7 to d6 > Better: O-O — Castles, taking the king to safety 6...d6 looks like normal development. It's a blunder, −0.61 to +5.88, because of 7.Qb5+: the queen gives check and attacks my bishop on b4, which has no defender. That's a fork with check, and I can't answer both threats. My opponent missed it and played 7.Bd2. If I castle first with 6...O-O, Qb5 is no longer check, so I'd have time to save the bishop; Stockfish rates that −0.63. The engine had me at 96.5 % at some point in this game and I still lost it; it's one of the two abandoned losses this week ([game on Chess.com](https://www.chess.com/game/live/108563376260)). ### A pawn move that wins time, or a knight move that doesn't Game #779, White against fosenheimen (1016), in my old Zukertort Chigorin setup. I was clearly winning, +4.63, and Black's rook on f5 was in front of my pawns. > **Move 22: Black's rook on f5 can be attacked by a pawn** > Position (FEN): `rn4k1/p6p/2p3p1/1p1p1r2/5P2/2P1P1PP/PP1N4/2KR3R w - - 0 22` > Played: Nf3 — Knight from d2 to f3 > Better: e4 — Pawn from e3 to e4, attacking the rook on f5 22.e4 attacks the rook with a pawn, so Black must move it, and my pawns gain space in the centre. I played 22.Nf3, a quiet knight move, and the evaluation dropped from +4.63 to +1.91. Still better, but I didn't keep it: I was checkmated at move 51 with 73 seconds left on my clock. Attacking a rook with a pawn is one of the cheapest ways to gain time in chess. I didn't look for it ([game on Chess.com](https://www.chess.com/game/live/108588621304)). ### A double attack before castling Game #780, White against nanaira61 (990), in a Ruy Lopez. Equal, +0.05, and my king was still on e1. > **Move 9: Black's knight on e4 has no defender, but my king is still on e1** > Position (FEN): `r1bqkb1r/2pp1pp1/p1n4p/1p6/3Pn3/1B3N2/PP3PPP/RNBQK2R w KQkq - 0 9` > Played: Bd5 — Bishop from b3 to d5, attacking the knight on e4 and the knight on c6 with the rook behind it > Better: O-O — Castles, taking the king to safety 9.Bd5 is a real double attack: it hits the undefended knight on e4 and the knight on c6, with the rook on a8 behind it. But Black has a check available: 9...Bb4+, and in Stockfish's line 10.Bd2 Nxd2 11.Nbxd2 Qe7+ 12.Kf1 my king gets chased and loses the right to castle. The evaluation went from +0.05 to −1.94. The engine's choice was simply 9.O-O, and only then the tactics. My opponent chose 9...Nf6 in the game, and I still abandoned it later ([game on Chess.com](https://www.chess.com/game/live/108632394592)). ## The lesson of the week, and the plan Two of three key moments this week had castling as Stockfish's best move. An uncastled king gives the opponent checks, and checks are what ruin my double attacks. That's the lesson: **castle before I start the tactics**, unless I've checked every enemy check first. Looking back at weeks 17 to 31 together (369 games, 168 wins, 190 losses, 11 draws, and a rating that went from 997 to 985), the plan almost writes itself from the numbers: 1. **Blunder check before every move.** Stockfish counted 796 blunders in those 369 games. Most of the ones I've written about were loose pieces, forks and pins on the very next move. 2. **Convert winning positions calmly.** In 96 games I had at least an 80 % chance of winning and didn't win, including four stalemates. When winning, checks and captures first, quick simple moves, and one look at whether the opponent still has a legal move. 3. **Finish my games.** 125 of my 190 losses ended with me abandoning. A game I leave is a game I can't save, and a position I never get to practise. 4. **Castle early, then attack.** This week's lesson, from games #775 and #780. ## All games this week | Game # | Date | Colour | Opponent (rating) | Result | Opening | Accuracy | Link | |---|---|---|---|---|---|---|---| | 775 | Aug 5 | Black | spillz987 (1010) | Loss | Van t Kruijs Opening | 61.4 | [game](https://www.chess.com/game/live/108563376260) | | 776 | Aug 5 | Black | CrazyWebhead (980) | Win | Center Game Accepted Paulsen Attack | 93.5 | [game](https://www.chess.com/game/live/108564635092) | | 777 | Aug 5 | White | Fratkurd (995) | Win | French Defense Knight Variation | 91.7 | [game](https://www.chess.com/game/live/108564779568) | | 778 | Aug 5 | Black | arung2000 (964) | Win | Kings Pawn Opening Kings Knight Variation | 90.3 | [game](https://www.chess.com/game/live/108564939476) | | 779 | Aug 6 | White | fosenheimen (1016) | Loss | Queens Pawn Opening Zukertort Chigorin Variation | 72.5 | [game](https://www.chess.com/game/live/108588621304) | | 780 | Aug 9 | White | nanaira61 (990) | Loss | Ruy Lopez Opening Morphy Defense Caro Variation | 77.8 | [game](https://www.chess.com/game/live/108632394592) | Previous: [week 30](/journey/week-2024-w31/) · Next: [week 32](/journey/week-2024-w33/) · Every week in one place: [the journey hub](/journey/) · Build your own plan from your games: [review your own games like this](/features/ai-chess-analysis/). ## FAQ ### How do you make a chess improvement plan from your own games? Count what keeps repeating. Over weeks 17 to 31 I played 369 rapid games; Stockfish found 796 blunders, 125 of my 190 losses ended with me abandoning the game, and in 96 games I had at least an 80 % winning chance and didn't win. Those three numbers are my plan. ### Why is castling so often the best move for beginners? Because an uncastled king gives the opponent checks. In games #775 and #780 this week Stockfish's best move was castling, and the moves I played instead allowed a queen check that forked a bishop and a bishop check that started a run of checks. ### What does a flat week mean for improvement? Not much on its own. This week was only six games and ended exactly where it began, at 985, with a 3–3 score. ## Related - [From 631 to 1385: my chess journey in 2,117 games](https://thechesscoach.netlify.app/journey.md) - [Week 30 (Aug 2024): switching to 1.e4 as White](https://thechesscoach.netlify.app/journey/week-2024-w31.md) - [Week 32 (Aug 2024): the Englund Gambit and a seven-game skid](https://thechesscoach.netlify.app/journey/week-2024-w33.md) - [Can Claude analyze chess games? Yes, with Stockfish](https://thechesscoach.netlify.app/answers/can-claude-analyze-chess-games.md)