Notes on market structure, order books and the small tools I build to study them.
Two resting orders at the same price are not equal. Most of the edge in passive execution comes from where you sit in the queue, and the queue moves with every cancel. A replay of a week of book updates makes this obvious in a way averages never do.
Liquidity thins out long before volume does. Measuring the depth within ten basis points of the mid every minute gives a cleaner picture of a market's mood than the headline spread.
Every experiment gets a date, a question and a one-line answer. Most answers are "no effect". The log is still the most useful file I have.