What I was looking for
Evidence that market makers pull depth from the book just before quotes move.
What I found
I built an order-flow study to find depth withdrawal before quote moves. Withdrawal before a move was +6.4% ± 6.6, against +5.0% ± 1.4 with no move: nothing. The hypothesis died.
The instrumentation built to test it produced the actual finding, a map of where makers get picked off. The value came from the measurement apparatus, not the hypothesis.
The move
I build the measurement to outlive the hypothesis, and I publish when the hypothesis dies.