Casefile
A formal report on one player's chess: what keeps going wrong, the evidence for it, and the single change worth making next.
Sample studyA study, not a person. a.study.not.a.person cannot exist on Chess.com or Lichess — the handle breaks both sites' username rules, so no one can ever own it. The positions are real chess; the subject is synthetic.
Across 412 rated blitz games this year you played 2,014 moves in under five seconds. That is not, by itself, a fault: blitz is a fast game and most fast moves are fine. The fault is where the fast moves cluster. They are not spread evenly through your games. They gather in the eight or nine moves immediately after your opponent leaves theory — the moment the position stops being remembered and starts being calculated.
In that window your accuracy falls from 84.1% to 71.6%, and it stays down until either the position simplifies or the game is already lost. Everything in this chapter is about those eight moves.
Table 1
| Time on move | Moves | Accuracy | Blunders per 100 |
|---|---|---|---|
| Under 2s | 861 | 68.4% | 6.1 |
| 2–5s | 1,153 | 74.2% | 4.0 |
| 5–10s | 1,606 | 86.9% | 1.7 |
| Over 10s | 742 | 89.3% | 1.1 |
The gap between the first row and the third is the entire chapter, stated as a number. Three extra seconds takes you from 68.4% accuracy to 86.9%, and from 6.1 blunders in every 100 moves to 1.7. Your opening knowledge is not the problem. You are losing rating to the three seconds you did not spend.
Exhibit 1
Ng5
played in 1.8 seconds, the first move out of book
Exhibit 2
Nxe4
played in 2.3 seconds; the knight is lost to Nxe4 d5, and the clock read 3:41
Exhibit 3
Ra8+
played in 0.9 seconds, with 2:14 on the clock and nothing forcing the pace
Three positions, three different openings, three different opponents, and the same signature: a move made in under two and a half seconds at the first moment the position required an original thought. In none of the three were you short of time. The clock was not making you play fast. Habit was.
Figure 1
The encouraging reading of that chart is that you have already done this once. Between March and August your blunder rate fell by a quarter and your rating rose 93 points, and nothing else about your game changed in that window — same openings, same time control, same volume. You slowed down slightly and it paid immediately.
The undertaking
For the next thirty rated games: when your opponent plays a move you have not seen before, you stop and count to three before touching a piece. Nothing else changes.
One opening in your repertoire has cost you 31 rating points this year — more than the other eleven combined — and it is not the one you would guess.
This chapter is written. It opens when your subscription does.
You reach rook endgames more often than 87% of players at your rating and convert them less often than 61%.
Waiting on your next 20 games before it can be written.
Your Casefile is a living document. Every game you play adds to it, and a chapter opens as soon as the evidence for it holds up.