Week 113: three winning positions lost at 1200
Week 113 covers April 6–12, 2026: 9 rapid games, 1227 → 1219, four wins and five losses, with no draws. Three of those five losses included a position the engine assessed as winning for me. These three winning positions lost at the 1200 level give this entry its focus, with the actual moves and results below.
Three winning positions lost at 1200: the week in numbers
| Games | 9 |
| Wins / losses / draws | 4 / 5 / 0 |
| Rating | 1227 → 1219 (−8) |
| Average accuracy | 80.2% |
| My blunders | 18 |
| With White / with Black | 3 wins, 2 losses / 1 win, 3 losses |
| Openings | Modern Defense (three move orders), Pirc, Old Benoni, four others once each |
The results read almost like a metronome: win, loss, win, loss, win, then two losses, a win and a loss. My rating bounced between 1219 and 1235 the whole week and never left that band. The average accuracy of 80.2% hides a wide spread, from 96.3% in my best game to 51.3% in my worst.
The best one was game 2034 against Wiggins_Plays_Chess (1199). Seventy-six moves, no blunders at all, and I won on time with only 4 seconds left on my own clock. Even there the engine found something I missed: after my opponent's 17th move I had mate in two (queen to h5 with check, then queen to f7), and I played a quiet knight move instead. I still won, but it's the same pattern you'll see below: the forcing move was there and I didn't look at it first.
Three moments where it turned
Game 2035: the same pawn, taken with the wrong piece
Against gelagela19649 (1261), playing Black in the Modern Defense, I was completely winning on move 11. The engine gave −6.14, which means Black is up by roughly six pawns' worth. White's pawn on d4 was hanging, and there were two ways to take it.
I took with the knight, which attacks White's queen. It looks active. But White doesn't have to retreat: the queen simply takes my f7 pawn next to my king, with check, and then grabs my bishop on g7. After my move the evaluation was +6.99 for White. Taking with the queen instead gives check, and a check is a double attack here: it also hits the bishop on d3 and the undefended pawn on e5. The game was over by move 16 (game 2035).
Game 2036: a queen move that unguarded my knight
Against hanumateja (1237), again with Black in the Modern, the position was about equal on move 18 (+0.05). My knight on e5 was protected by my pawn on d6. I moved my queen to d7 to attack the white knight on b5.
The problem: with my queen on d7, the d6 pawn can no longer safely take anything on e5. If it does, the d-file opens and White's queen on d5 takes my queen. So White's rook could just take my knight, and the engine jumped to +6.59. Queen to c8 kept everything level. I was lucky: my opponent made five blunders in the game and I won by resignation on move 39 (game 2036).
Game 2039: the bishop that was guarding a knight
With White against Allianzem (1229), in an Old Indian Defense, the evaluation was +0.26 on move 16. My bishop on e2 was quietly guarding my knight on f3.
Bishop to c4 pins Black's f7 pawn to the king, which felt aggressive. But Black's bishop on c6 can now take my knight on f3, and I have to take back with the g-pawn, which opens the shelter in front of my own king. The engine's line after that was clearly better for Black (−2.40 in the game). I blundered twice more, and by move 29 I had 55 seconds left and lost (game 2039).
The lesson of the week
Checks first, in both directions. In game 2034 I missed a mate in two that started with a check. In game 2035 I missed a check that won more material, and my move allowed a check that won the game for my opponent. When the position is good, the extra ten seconds I spend should go to one question: what are the checks and captures, mine and theirs, after this move? Two of these three turning points would have disappeared with that habit, and the third (game 2039) is its cousin: before moving a piece, ask what it was protecting.
The other three losses don't need a board to explain. Game 2037 I lost with 96.2% accuracy: the record shows it was abandoned at move 13 while I was well ahead. In game 2042 I was winning early, made six blunders and got checkmated on move 37.
All games this week
| Game # | Date | Colour | Opponent (rating) | Result | Opening | Accuracy | Link |
|---|---|---|---|---|---|---|---|
| 2034 | Apr 6 | White | Wiggins_Plays_Chess (1199) | Win | Modern Defense with 1 d4 | 96.3% | game |
| 2035 | Apr 7 | Black | gelagela19649 (1261) | Loss | Modern Defense with 1 e4 | 51.3% | game |
| 2036 | Apr 9 | Black | hanumateja (1237) | Win | Modern Defense | 83.0% | game |
| 2037 | Apr 9 | Black | BM_MASTER_69 (1247) | Loss | Pirc Defense | 96.2% | game |
| 2038 | Apr 9 | White | giorgosntz (1208) | Win | Old Benoni Defense | 80.6% | game |
| 2039 | Apr 9 | White | Allianzem (1229) | Loss | Old Indian Defense | 79.4% | game |
| 2040 | Apr 12 | Black | Rom142000 (1244) | Loss | Nimzowitsch Defense Declined, Williams Variation | 73.9% | game |
| 2041 | Apr 12 | White | VictoriousD71 (1221) | Win | Queen's Pawn Opening, Horwitz Defense | 88.1% | game |
| 2042 | Apr 12 | White | Gonzlez1234 (1244) | Loss | Queen's Pawn Opening, Zukertort Chigorin Variation | 72.7% | game |
Previous: week 112 · Next: week 114 · All weeks: the journey hub. Every evaluation here comes from Stockfish; if you want to review your own games like this, see how Chess Coach's AI chess analysis works.
For a general method beyond this week's games, see the guide to reviewing lost winning positions.
Frequently asked questions
How did I lose three winning positions around 1200?
Three of my five losses in nine games included an earlier position the engine assessed as winning. In game 2035 a knight capture missed the stronger queen capture with check, and the following move reversed the assessment.
Can you lose a chess game with high accuracy?
Yes. In game 2037 my accuracy was 96.2% and I still lost: Chess.com recorded the game as abandoned at move 13, while the engine had me well ahead.
How many blunders does a 1200 player make per game?
In this week I made 18 blunders in 9 rapid games, two per game on average, with an average accuracy of 80.2%. One game had none and one had six.