Crypto order book imbalance trading strategy, stated as a ratio

A crypto order book imbalance trading strategy, stated as a ratio. Depth is the one input that says something about the next few seconds rather than the last few. Read it as a ratio against the other side, never as a raw number.

1. The rule

# BUY if TotalBidSize > TotalAskSize * 1.6 and BuyVol > SellVol: if (Ask1 - Bid1) / Price < 0.001 and DayAmount > 50000000: Buy()# SELL — imbalance decays fast, so the time exit is short if ProfitPct >= 0.6 or ProfitPct <= -0.4 or HoldTime >= 180: Sell()

2. Why a ratio and not a size

A raw size threshold means nothing. Two hundred BTC of resting bids is enormous on one venue and unremarkable on another, and the same number means different things at 03:00 and at 14:00. The ratio against the other side is comparable across symbols and across the day, which is the only way a single number in your code can hold up. The spread requirement and the 180-second time exit belong together. An imbalance edge is measured in seconds. Holding past it converts an order book read into a directional bet you did not intend to take, and the 0.6% target is small enough that a wide spread eats it outright.

3. When this works and when it does not

It works on liquid pairs during active hours, where the book is deep enough that the ratio reflects real intent. It fails in two specific ways. The first is spoofing: resting size that is cancelled the moment price approaches it. You cannot detect this from a snapshot, only from watching how often large resting orders actually get filled on your venue. The second is depth quality. Exchanges differ in how many levels they publish and how often they update — some send a full book on every change, some send throttled snapshots, and a throttled snapshot makes an imbalance look stable when it is not. That second problem is measurable, and we measured it: see the real cost of crypto trading for what quoted depth is actually worth across venues, and stale quotes in crypto market data for how often a feed stops updating while the book keeps moving.

4. Before you trade it

Record your own book first, then backtest against the recording — see how to backtest a crypto trading strategy. A strategy whose entire premise is depth cannot be validated on candles. Check one number before anything else: on your exchange, how often does TotalBidSize change between consecutive one-second rows? If the answer is rarely, your feed is throttled and this strategy is reading a stale book.

5. What the round trip asks of it

A 0.6% target is small, so cost matters proportionally more here than in any other crypto strategy in this library: the break-even moves from 40% to 50% at a 0.10% round trip. Ten points for a strategy holding three minutes. That is precisely why the spread condition (Ask1 - Bid1) / Price < 0.001 is not optional. At 10 basis points quoted, the round trip is 0.2% and the break-even goes past 60%. The gate is doing the same job the target cannot. See the real cost of crypto trading.