Crypto order flow without the mythology
A practical guide to reading who is moving, what liquidity did, and whether a visible cross-exchange edge survives realistic costs.
1. Price is the result; order flow is the path
A candle records open, high, low and close for an interval. It does not show which venue changed first, whether resting orders were pulled, or where aggressive market orders arrived. Order-flow analysis adds that path.
2. Keep exchanges separate before consolidating
BTC/USDT on Binance and BTC/USDT on OKX are related but distinct books. A useful MultiBook preserves each origin, then aggregates only fresh compatible levels. Mixing unlabeled orders can create false liquidity.
3. A wall is evidence, not intent
Large resting size can absorb trades, move away, split across prices or disappear. Compare the wall with executed volume, add/pull behavior and subsequent price response. “Spoof-like” behavior is a suspicion—not proof of manipulation.
4. Footprint and aggressor flow
Footprint candles group executed volume by price and side. Positive delta means more aggressive buying in the sample; it does not guarantee the next move. Pair delta with liquidity and cross-venue confirmation.
5. Gross spread is not net edge
Start with best executable bid minus best executable ask across venues. Then subtract both taker fees and consider slippage, funding, transfer constraints, latency and capital already positioned on each venue. A positive screen number is not a guaranteed fill.
6. Freshness before signal
A stale quote can look like an extraordinary arbitrage. Check publisher time, browser arrival time, sequence gaps and last-packet age. Exclude stale books before calculating leadership or dislocation.
7. Replay the decision, not just P&L
Use market replay and paper trading to review what information existed at the decision time. Paper results omit real queue position, rejections and market impact, so they are process tools rather than performance evidence.