Road to ELO 1500 — Game Review #1
First weekly game review. I exported the last 50 rapid games from chess.com and ran them through Stockfish (depth 12). The picture is clearer than I thought, and not in a flattering way.
The week in numbers
| 50 rapid (10 | 0) games over May 15–17. |
| W / L / D | 26 / 24 / 0 |
| Win rate | 52% |
| Rating arc | 759 → 771 (peak 821, low 749) |
| Avg ACPL | 65.7 |
| Blunders | 68 (1.36/game) |
| Mistakes | 154 |
| Inaccuracies | 178 |
| Lost on time | 2 |
The summary version: I beat people my level just about as often as they beat me, but neither side is playing clean chess. 1.36 blunders per game means most games are decided by who blunders second.
Where the leaks are: the endgame
Mistakes by phase (user moves only):
| Phase | Blunders | Mistakes | Inaccuracies |
|---|---|---|---|
| Opening (moves 1–10) | 9 | 48 | 54 |
| Middlegame (11–25) | 25 | 76 | 79 |
| Endgame (26+) | 34 | 30 | 45 |
Half of all my blunders happen after move 25 — once queens are off and I actually have to find a plan. The opening is fine; the middlegame is the standard rating-bracket mess; the endgame is the single biggest leak.
Opening repertoire — what’s working, what isn’t
Almost every game is 1.e4. As Black, I almost always answer 1…e5. The top positions after 6 plies:
| Position | Played | W | L |
|---|---|---|---|
| Italian, 3.Bc4 Bc5 (Giuoco) | 5 | 3 | 2 |
| Scotch, 3.d4 exd4 | 4 | 1 | 3 |
| Two Knights, 3.Nc3 Nf6 | 4 | 3 | 1 |
| Italian, 3.Bc4 h6 | 3 | 3 | 0 |
| Italian, 3.Bc4 d6 | 2 | 2 | 0 |
| Nc3 Bb4 (Three Knights) | 2 | 0 | 2 |
The Italian / Giuoco lines (anything with 3.Bc4) — 8W/2L combined. That’s clearly the repertoire I should commit to.
The Scotch (3.d4 exd4) is bleeding — 1W/3L. Either I drill those lines properly or I stop playing it and lock in 3.Bc4.
Three blunders that show the pattern
Game 15, move 24 — Re1?? (best Bxg7) +353cp → −71cp. Winning a clearly won game and I reach for a “developing” rook move instead of taking the bishop with check-followed-by-mate threats. I still won, because the opponent blundered worse. Lucky, not good.
Game 22, move 12 — a4?? (best f4) Italian setup, around move 12 I went pawn-grabbing on the queenside while the centre was screaming for f4. Lost 311 cp and the game. Classic “I had a plan but it wasn’t the position’s plan.”
Game 11, move 10 — d5?? (best Kh7) Scotch as Black. Played a thematic-looking pawn push that actually walked into a forced sequence. The engine wanted a quiet king move; I wanted a “thinking-player move.” The game was effectively lost on move 10.
Worst games (highest ACPL)
| # | Opponent | Result | ACPL | Blunders / Mistakes |
|---|---|---|---|---|
| 34 | IPlayChess2434 | won | 160 | 6 / 6 |
| 2 | Alienontour | lost | 143 | 4 / 9 |
| 15 | chessenjoyerguy | won | 124 | 5 / 1 |
| 47 | Tunderhino | won | 111 | 4 / 9 |
| 33 | Albee34 | won | 101 | 3 / 6 |
Three of my five sloppiest games this weekend were wins. That’s the rating bracket — both sides are throwing pieces and someone has to be left holding the king.
Historical context
I also ran the previous 50 games (a longer tail going back to late 2024). The patterns line up almost exactly:
- Win rate 42% (vs 52% recent) — slight recent uptick.
- Avg ACPL 65.6 — identical.
- Endgame blunders proportionally the same.
- Same opening repertoire, same Scotch problem.
Translation: the recent improvement is more “I’m playing more games” than “I’ve fixed anything.” The leaks haven’t moved.
What’s next (before Game Review #2)
Endgame is the leak. Italian is working. Plan for the week — about 90 minutes of study, plus four habit changes that cost zero extra time.
Learn (≈90 min total)
- K+P vs K — opposition, key squares, rule of the square. 30 min on the Lichess endgame trainer.
- Rook endings — Lucena and Philidor positions, plus “rook behind the passer.” 30 min: one Hanging Pawns video + 10 R+P vs R drills.
- Lock the Italian — drop the Scotch. Build a 3-move-deep response tree for 3.Bc4 lines using the Lichess opening explorer. 30 min, one-time.
Do (every game)
-
**Switch to 15 10** rapid. Increment fixes endgame clock panic, which is half the leak. - Before every move: name the opponent’s three best replies. If I can only think of one, I haven’t looked hard enough.
- When ahead in material: trade pieces, push pawns. Three of my five sloppiest games this week were wins where I kept the position complicated instead of simplifying.
- Queens off the board → 10 seconds minimum per move. My endgame moves currently take the same clock time as my opening moves. Backwards.
Targets for next week: blunders/game under 1.1, endgame blunder share under 40%, ACPL under 60. Back with Game Review #2.
Generated with the chess-game-review skill — Stockfish 17, depth 12, python-chess. Move classification thresholds: inaccuracy ≥ 50cp, mistake ≥ 100cp, blunder ≥ 300cp, per-move loss capped at 1000cp.