Definition. How hard the orderbook is being rewritten, compared with how hard it is usually rewritten. Each change to a price level on the venue's depth feed is one update; this metric counts those updates in the interval and scores the count against the same instrument's own trailing 30 days at the same hour of day. Changes to one price inside a single feed message collapse into one update, and counts are specific to that venue's feed. Two values are returned: zscore, the standardised score against the same hour of day, and ratio, the raw count divided by the 30-day average count across all hours.

Interpretation. Market makers rewrite quotes far more often when repricing risk, so update activity rises with volatility rather than lagging it, and stays elevated into the following hour. A zscore above 2 picks out roughly one interval in twenty-three and above 3 roughly one in a hundred, in each case the busiest for that hour of day rather than the busiest outright. On Binance BTCUSDT and ETHUSDT since 2020, intervals crossing 2 were followed about 63% of the time by more realised volatility over the next hour than intervals matched on the same trailing volatility measured up to and including the interval itself, the same hour of day and the same year. A trailing z-score of the price range, matched on the same number of alerts, performs the same, so read this as a second view of the regime rather than a better volatility signal: its distinct value is that it measures book activity directly, and about half of its readings arrive when the price range is unremarkable. Scoring against each instrument's own history keeps the value readable across quiet and busy regimes; comparability across instruments is not yet validated beyond the two published here. Read ratio when the size of the burst matters rather than its rarity: a ratio of 5 says five times the usual traffic.

Technical. zscore = (ln(1 + updates) - mean) / standard deviation, where mean and standard deviation are taken over the same instrument's intervals at the same hour of day in the preceding 30 days, ending strictly before the interval being scored, so no value uses information from its own interval or later. ratio = updates / the average update count of the intervals in the preceding 30 days, across all hours, that had any updates. zscore is null until 14 days of history are available, and for intervals too sparse to score.

Note: Second intervals and tick-level data can be made available as a bespoke request. Please reach out via Contact Us to learn more.