Started recording depth myself and now I don't trust my old numbers
Two months of running the same mean reversion thing on sample data, and my fill assumptions were always a bit hand-wavy — I just assumed the book looked roughly like the bars said it did. Then that guide on recording your own crypto data got me set up capturing depth off the venue I actually trade, and after a couple of weeks the distribution of spreads looks nothing like what I'd been implicitly assuming. Wider tails, more of the day spent in states I'd have called outliers. So I'm wondering, for those of you who record your own: how long did you let it run before you trusted it enough to recalibrate off it? A week feels like one sample of a week, not a sample of the market.
