Week 84: how to get better at the middlegame in chess, starting with mine
Week 84 (September 1 to 7, 2025) was a good week on paper: 9 rapid games, 6 wins and 3 losses, and my rating went from 1205 to 1229. Under the result there's a split that answers, at least for me, how to get better at the middlegame in chess: first notice how much worse you play it. My openings averaged 88.4% accuracy this week. My middlegames averaged 67.2%. Of my 19 blunders, 15 came in the middlegame and 4 in the opening. The endgames, the few I reached, were fine.
The week in numbers
| Games | 9 |
| Wins / losses / draws | 6 / 3 / 0 |
| Rating | 1205 → 1229 (+24) |
| Average accuracy | 75.7% |
| Blunders | 19 |
| Most played openings | London System (2), Modern Defense with 1 e4 (2), Queen's Pawn Opening Zukertort Chigorin Variation (2) |
The week came in three runs: three wins (games 1659 to 1661), three losses (1662 to 1664), three wins again (1665 to 1667). As White I won four of five, all with my d4 and Bf4 setup, and three of those wins were above 91%. As Black I went 2–2. The three losses tell the story better than the average does. In each one my opening accuracy was high, 91.9%, 78.9% and 93.1%, and my middlegame accuracy fell to 42.5%, 42.1% and 65.2%. I get out of the opening with a decent position and then stop seeing my opponent's moves.
Three middlegame moments
This is the one that hurt. Against moldova727 (1264) I had White, a London System, and a winning position (+4.72). Black's queen was on g5, near my king. I pushed 30.g4. Stockfish's best reply for Black shows the problem: 30...Bxg4, the bishop takes the pawn and attacks my rook on d1, which had no defender. The engine's move for me was 30.Rc5, attacking the queen, so Black has to answer that first. My move swung the evaluation from +4.72 to -5.07. My opponent didn't find the bishop capture right away (30...h5 came instead), but I blundered again on the next move and got checkmated in the end. A pawn move near your own king is never just a pawn move: it changes which squares your pieces can still be defended on.
Against nelebarut (1217), in the Pirc's Geller System, I had just taken a pawn on e4 with my knight. White's queen on e2 was now attacking that knight along the e-file. I defended it with 5...Bf5. It looks natural, and Stockfish calls it a blunder (from -1.52 to +2.95). Its line for White starts with 6.g4: a pawn attacks my bishop. If the bishop takes it, 6...Bxg4, then 7.Qxe4 wins the knight, and my bishop on g4 is left with no defender. A defender that can be chased away isn't really a defender. The engine's 5...Nf6 simply brings the knight home with the extra pawn. My opponent played 6.Ng5 instead, I took that knight too, and I won by checkmate on move 52 with 56 seconds left.
The last one is technically still the opening, but it's the same kind of decision. Against yash_mates (1214) White sacrificed a bishop on f7 on move 4, my king went to e8, and I was a full piece up (-5.89). With 6...Bh6 I attacked the knight on g5 and offered to trade my bishop for it. Trading when you're a piece up sounds sensible, and that was my logic. Stockfish disagreed: its move was 6...c5, and after my bishop move the advantage dropped to -2.74. Still winning, and I did win, but I gave away half the margin with a move I thought was "safe". In the engine's line after 6...Bh6 (7.d4 d5 8.exd5 Bxg5 9.Bxg5) the trade happens and the evaluation settles around -3.
The lesson of the week
The middlegame needs a different question than the opening. In the opening I ask "where does this piece belong?", and my London and g6 setups answer that for me. In the middlegame the question has to be "what is my opponent's most forcing reply?" In all three positions above, the engine's refutation started with a direct threat: a bishop capture that hit a loose rook, a pawn chasing my only defender, and a trade I offered when nobody asked me to. So my rule for the next weeks is concrete: before each middlegame move, find the opponent's checks, captures and pawn attacks against my pieces, then decide.
All games this week
| # | Date | Colour | Opponent | Result | Opening | Accuracy | |
|---|---|---|---|---|---|---|---|
| 1659 | 2025-09-01 | White | chesschecknmate (1184) | Win | London System | 96.0% | game |
| 1660 | 2025-09-02 | Black | yash_mates (1214) | Win | Modern Defense with 1 e4 | 76.5% | game |
| 1661 | 2025-09-02 | Black | nelebarut (1217) | Win | Pirc Defense Modern Defense Geller System | 68.6% | game |
| 1662 | 2025-09-03 | White | moldova727 (1264) | Loss | London System | 48.0% | game |
| 1663 | 2025-09-03 | Black | VLAD6IXX (1243) | Loss | Nimzowitsch Defense Declined Williams Variation | 50.1% | game |
| 1664 | 2025-09-04 | Black | moner12365 (1268) | Loss | Modern Defense with 1 e4 | 77.2% | game |
| 1665 | 2025-09-06 | White | thingamajig07 (1213) | Win | Queen's Pawn Opening Zukertort Variation | 92.2% | game |
| 1666 | 2025-09-06 | White | mike_tayso (1196) | Win | Queen's Pawn Opening Zukertort Chigorin Variation | 91.0% | game |
| 1667 | 2025-09-07 | White | Swwfiane (1209) | Win | Queen's Pawn Opening Zukertort Chigorin Variation | 81.3% | game |
Previous week: week 83. Next week: week 85. All the weeks are on the journey page. The phase-by-phase accuracy above comes from the same Stockfish analysis Chess Coach runs when you review your own games like this.
Frequently asked questions
How do you get better at the middlegame in chess?
What helped me read my own games was splitting accuracy by phase. This week my openings averaged 88.4% and my middlegames 67.2%, and 15 of my 19 blunders were middlegame moves. The fix I'm working on: before each middlegame move, look for the opponent's most forcing reply, like a pawn that can attack one of my pieces.
Is it normal to play the opening well and then collapse?
It's my pattern. In my three losses this week, opening accuracy was 91.9%, 78.9% and 93.1%, and middlegame accuracy was 42.5%, 42.1% and 65.2%.
Should you trade pieces when you are a piece up?
Not automatically. In game 1660 I was a bishop up and offered a trade with 6...Bh6; Stockfish cut my advantage from -5.89 to -2.74 and preferred 6...c5. I still won.