You lose a game you thought you were winning. You open the engine review — the same one you’ve opened a hundred times — and realize it still doesn’t tell you what to do differently in the next game.

That’s not a you problem. It’s an engine-review problem.

What the engine sees

Game Review shows you the best move in every position of one game. It does this very well. But it never connects game 47 to game 46 to game 45 to see that you keep releasing central tension two moves before your opponent does — and that’s why you keep ending up in the same lost endgame.

One game is a story. Twenty games is a pattern.

What a pattern looks like

A pattern is a specific kind of mistake that shows up across many of your games. It has a shape:

The trigger — the position type where it tends to happen (castled king, IQP, opposite-side castling).
The move you play — usually a move that “feels” safe at the time.
The principle you missed — the underlying rule that would have stopped the leak.

For me, the pattern was releasing central tension too early. I’d play d4 or e4 exchanges on autopilot, end up with symmetrical pawns, and wonder why the resulting endgame felt dead. The engine flagged it in every loss. Game Review never once said “you do this every game.”

What to do about it

Three things changed the curve for me.

1. Stop watching more videos

I was consuming more content than I was practicing. Every new YouTube course felt productive but didn’t move my rating. The material was never wrong — it just wasn’t my advice.

2. Start clustering my own losses

I pulled the last fifty games I lost and sorted them by who was winning at move 25. The clusters were obvious once I wrote them down:

13 of them I’d been winning at move 25 and lost anyway.
8 of them I’d missed a clean tactic at move 12-18.
6 of them I’d hung a piece in the opening.

Those are three different patterns. Each one needs a different drill.

3. Make each leak a mission, not a lecture

When the pattern is “I keep missing the same knight fork”, watching a 40-minute course on tactics doesn’t help. Playing the fork from my position, with the pieces I had, until I see it — that helps.

That’s the bet behind GMSomeday: the work is already in your games. You just need someone to point at the right pattern.

If you’ve been plateauing and the engine review isn’t moving you, try this exercise: open the last 20 losses, sort them by who was winning at move 20, and see if the same trigger shows up three or more times. If it does, that’s your pattern. The rest is just drilling it.