Week 35: how to get better at chess tactics, starting with my own misses
Week 35 ended exactly where it started: 945 → 945 after 8 rapid games, 4 wins and 4 losses, from September 2 to 6, 2024. The engine counted 26 blunders of mine in those eight games, more than three per game, and when I looked at the three biggest ones they were all about the same thing: a move that attacks two targets at once. If you are wondering how to get better at chess tactics, my games this week are a good map of what a beginner misses.
The week in numbers
| What | Number |
|---|---|
| Games | 8 (#836–#843) |
| Wins / losses / draws | 4 / 4 / 0 |
| Average accuracy | 76.7 % |
| Blunders (my moves) | 26 (3.2 per game) |
| Rating | 945 → 945 (+0) |
| Most played openings | Colle System Anti Colle Variation (1), Dutch Defense (1), Giuoco Piano Game Main Line (1) |
The colours split oddly: with Black I won three of four, with White I lost three of four. Only two of the four losses were recorded as "game abandoned", fewer than usual for me. One loss, game 840 against Thekamzanator (975), went on time: I had about 14 seconds left on my last move, played at 51.6 % accuracy, and the engine had given me an 80 % or better winning chance earlier in that game.
My blunders were spread across every phase: 12 in the middlegame, 7 in the opening and 7 in the endgame. Those seven endgame blunders all came in one game, game 842 against Jwh236Eue (936), which I still won by resignation. At this level both players blunder, and sometimes the last blunder decides the game, not the best move.
How to get better at chess tactics: three double attacks from one week
A double attack is one move that creates two threats. The most famous version is the fork, where a single piece attacks two enemy pieces and only one can be saved. All three moments below are double attacks: one I missed, one I allowed, and one quiet capture I skipped.
The knight fork I didn't play (game 836)
Against marcdev09 (935), with Black, the engine had me at -3.84. Two of my knights could take a white pawn: one on c2, one on d5.
I took on c2, which attacked one rook. The engine's choice was the other knight on d5, and position.py describes it plainly: it "attacks the bishop on the queen's side, which has no defender; attacks the queen on the king's side, which has no defender; double attack". After my capture on c2, the engine's best reply for White is Bxf6 and the position is about equal (+0.08). My opponent played 21.Rd1 instead, but one move later I blundered again with 21...cxd5 and the score flipped to +4.80. I lost the game by checkmate.
The quiet move that let the opponent off (game 837)
With White against kubusiem (931), a Dutch Defense, the position after Black's ...c5 offered me a pawn capture.
I supported my pawn with c3, a typical London System reflex. The engine wanted the capture: 6.dxc5 is "clearly better for white" (+1.99), while after c3 and Black's cxd4 it is only "slightly better for white". It is not a fork, but it belongs here because of how I missed it: I never even looked at the capture. I won the game by checkmate anyway, after 4 blunders.
The fork I invited (game 838)
The last one is the reverse. With Black against swanpa (919), I moved my knight to the edge to chase White's bishop.
Na5 does attack the bishop. But the engine's answer for White is Bxf7+: the bishop takes the pawn next to my king with check, and after my king takes back, the pawn move b4 "attacks the knight on the queen's side, which has no defender". The evaluation went from +0.32 to +2.30. My opponent played 5.d3 instead, and I won the game by checkmate. I got away with it; the engine didn't let me.
The lesson of the week
Tactics are not a separate subject from "normal" chess. Every one of these moments was a normal-looking position where I chose a move without listing the alternatives. The fix I'm adopting is a fixed order of questions before each move: what are all my checks, all my captures, and does any move attack two things at once? Then the same three questions for my opponent after my move. In game 836 the first question finds Nfxd5. In game 837 the second finds dxc5. In game 838 asking it for my opponent finds Bxf7+ and b4. That checklist is how I'm trying to get better at chess tactics: not by solving more puzzles, but by asking the puzzle questions in my own games.
All games this week
| Game # | Date | Colour | Opponent (rating) | Result | Opening | Accuracy | Link |
|---|---|---|---|---|---|---|---|
| 836 | Sep 2 | Black | marcdev09 (935) | Loss | Colle System Anti Colle Variation | 60.6 % | game |
| 837 | Sep 3 | White | kubusiem (931) | Win | Dutch Defense | 89.4 % | game |
| 838 | Sep 3 | Black | swanpa (919) | Win | Giuoco Piano Game Main Line | 92.5 % | game |
| 839 | Sep 3 | Black | dury_dhan (964) | Win | Kings Pawn Opening | 94.6 % | game |
| 840 | Sep 4 | White | Thekamzanator (975) | Loss | Queens Pawn Opening | 51.6 % | game |
| 841 | Sep 5 | White | starikovm (957) | Loss | Queens Pawn Opening Zukertort Chigorin Variation | 76.2 % | game |
| 842 | Sep 5 | Black | Jwh236Eue (936) | Win | Giuoco Piano Game Giuoco Pianissimo Variation | 70.2 % | game |
| 843 | Sep 6 | White | qweenshope (969) | Loss | Queens Pawn Opening Horwitz Defense | 78.6 % | game |
Previous: week 34 · Next: week 36 · All weeks of the journey · Review your own games like this
Frequently asked questions
How do I get better at chess tactics as a beginner?
In my week 35 games the missing skill was spotting double attacks: one move that hits two targets. Looking for every check, capture and double attack before moving would have caught all three key moments.
What is a fork in chess?
A fork is a single move that attacks two pieces at once, so the opponent can only save one. In game 836 my knight could have taken on d5 and attacked White's bishop and queen at the same time.
Can you win with lots of blunders at 945?
Yes, sometimes. I won game 842 despite 9 blunders, 7 of them in the endgame, because my opponent made mistakes too. That is a warning, not a strategy.