Exchange Inflows & Outflows. Coins moving toward the place you sell them.
Exchange flows track coins moving between private wallets and exchange-controlled wallets. Inflows are coins arriving at an exchange; outflows are coins leaving it.
The conventional reading is that inflows suggest intent to sell, because you have to move coins to an exchange before selling them there, and outflows suggest accumulation and self-custody.
Why the simple reading is often wrong
The logic holds in aggregate and fails constantly in individual cases. Coins arrive at exchanges to be used as derivatives collateral, to be lent, to be moved between an exchange's own hot and cold wallets, or because a custodian reorganised. A large inflow can equally be an institution posting margin as a whale preparing to dump, and the transaction itself looks identical either way.
Exchange Inflows & Outflows in a strategy
A long stretch of net outflows suggests coins moving into storage rather than to market.
A transfer far outside the normal distribution is worth flagging even if the reason is unknown.
Total balance held on exchanges over time is a slower and steadier read than daily flow.
Flow data helps explain a move after the fact more reliably than it predicts one.
Use reserves as a slow regime filter
Flow data is too noisy to trigger trades. It is much better suited to deciding which direction a strategy is allowed to take.
- IFThe 30-day change in total exchange reserves for the asset is negative
- ANDThat net outflow has persisted for at least 3 consecutive weeks
- ANDThe reading uses reserve levels rather than daily flow, to smooth the noise
This is a filter on an existing strategy, not a strategy itself. It works at the timescale flow data is actually reliable at, which is weeks rather than hours, and it degrades gracefully: if the signal is wrong, the worst case is that your trend strategy took normal long trades in a market that did not trend. The three-week persistence requirement exists because single-week flow readings reverse constantly and are heavily distorted by internal exchange transfers.
Analytics providers maintain lists of known exchange addresses, and every flow metric is derived from those lists.
Total holdings per exchange over time, which is the slower and more reliable version of daily flow data.
Some exchanges publish their own wallet addresses, which is the only version of this data that is not inferred.
Treat it as slow context, not a trigger
Flow data is noisy at daily resolution, heavily dependent on wallet labelling, and routinely distorted by internal exchange transfers. It is reasonable as one slow-moving input among several. A strategy that trades directly off single flow events is trading someone else's labelling errors.
See the coins move before the price does.
Your keys, your account, your rules. We only send the orders.
- Sustained outflow trends
- Unusual single transfers
- Exchange reserve levels