Chess diary
What InvisibleKnight72 learns from three-minute games on Chess.com: the mistakes that cost games, shown on a board that can be replayed and tested. Between 17 October 2026 and 25 October 2026 the rating went from 1272 to 788.
Each day of play has a page with its key positions. Each position shows the move that was played, what the engine says happened, and a better move. The patterns pages collect the same kind of mistake from every day, so progress shows up as fewer examples over time. Opponents are anonymous.
- 59−92Wins and losses in 156 games
- 37 of 92Losses that began from a winning position (+3.0 or better)
- 1.67Blunders per game
- 13 of 19Losses on time that were already lost on the board
Rating
Progress by day
| Day | Games | Won and lost | Rating, first to last game | Blunders per game | Mistakes per game | Lost from +3.0 | Seconds per move |
|---|---|---|---|---|---|---|---|
| 17 October 2026 (3+0) | 6 | 1−5 | 1272 to 1109 | 1.83 | 2.17 | 3 | 5.3 |
| 18 October 2026 (3+0) | 22 | 5−16 | 1067 to 935 | 1.50 | 1.77 | 5 | 4.4 |
| Morning (06:08 to 07:39 UTC) | 4 | 2−2 | 1067 to 1099 | 1.25 | 3.00 | 0 | 3.9 |
| Afternoon (15:53 to 16:01 UTC) | 3 | 2−1 | 1069 to 1120 | 2.00 | 1.00 | 0 | 3.5 |
| Evening (18:32 to 21:59 UTC) | 15 | 1−13 | 1143 to 935 | 1.47 | 1.60 | 5 | 4.7 |
| 19 October 2026 (3+0 and 3+2) | 21 | 8−12 | 958 to 946 | 1.76 | 2.48 | 5 | 4.7 |
| 3+0 games (14:28 to 22:04 UTC) | 10 | 3−6 | 958 to 946 | 1.50 | 2.40 | 3 | 4.0 |
| 3+2 games (19:49 to 21:03 UTC) | 11 | 5−6 | 896 to 895 | 2.00 | 2.55 | 2 | 5.3 |
| 20 October 2026 (3+2) | 15 | 9−6 | 888 to 918 | 1.07 | 2.00 | 3 | 5.6 |
| 21 October 2026 (3+2) | 12 | 5−7 | 910 to 902 | 2.42 | 2.17 | 5 | 5.0 |
| 22 October 2026 (3+2) | 6 | 2−2 | 910 to 902 | 2.00 | 3.17 | 1 | 4.8 |
| 23 October 2026 (3+2) | 29 | 11−18 | 894 to 850 | 0.93 | 2.00 | 6 | 5.2 |
| Morning (07:31 to 11:58 UTC) | 9 | 4−5 | 894 to 897 | 1.11 | 2.22 | 2 | 5.1 |
| Afternoon (13:38 to 15:51 UTC) | 8 | 3−5 | 889 to 880 | 0.88 | 2.25 | 2 | 5.5 |
| Evening (18:16 to 19:13 UTC) | 12 | 4−8 | 872 to 850 | 0.83 | 1.67 | 2 | 5.2 |
| 24 October 2026 (3+2) | 18 | 10−8 | 858 to 866 | 1.56 | 1.83 | 2 | 4.8 |
| 25 October 2026 (3+2) | 27 | 8−18 | 874 to 788 | 2.48 | 2.59 | 7 | 4.8 |
| Morning (08:15 to 09:47 UTC) | 11 | 1−9 | 874 to 804 | 3.55 | 4.00 | 4 | 4.5 |
| Afternoon (13:53 to 16:48 UTC) | 13 | 6−7 | 796 to 796 | 1.54 | 1.62 | 2 | 5.4 |
| Evening (18:01 to 18:13 UTC) | 3 | 1−2 | 788 to 788 | 2.67 | 1.67 | 1 | 4.7 |
Where to start
The pattern pages, ordered by how many featured positions each one has.
- Hanging pieces and moved defenders: 18 featured positions. A piece is left attacked and undefended, or it was the only defender of something else and it moved.
- Missed tactics: 16 featured positions. A capture, a check or a mate was available and was not played, often for several moves in a row.
- King safety: 8 featured positions. Pawn moves and queen moves near the king that open the way for an attack, or pieces that leave the king when it is under pressure.
- Forks and pins: 8 featured positions. A knight fork, or a piece placed on a line that runs to the king and can be pinned.
How the analysis was made
Every move of every game was checked with an engine, and the positions shown here were checked again at greater depth. Stockfish 17.1 for 17 to 23 October and Stockfish 19 from 24 October: depth 12 for every move of every game; the featured positions were checked at depth 18 with Stockfish 17.1 and Stockfish 19. A blunder is a move that lowers the engine evaluation by three pawns or more. The engine finds the moves; choosing which positions to feature and explaining them is done by hand. The seconds per move come from the clock readings stored in the games. Each position has a link that opens it in Lichess for a second opinion.