Exchange scorecard: what each lets us check

When each list was published

Exchange by exchange

What each column measures

Newest list, and oldest list in use

The date of the address lists the exchange itself publishes (its proof-of-reserves files, page or repository) that our readings start from, chain by chain. Third-party label lists and archive copies don't count. The oldest is the chain whose list has gone longest without an update.

Chains on its lists

Which of Bitcoin, Ethereum, Litecoin, Dogecoin and Bitcoin Cash, the chains we run nodes for, its lists cover. Dollar stablecoins are read at its Ethereum addresses.

Its own bitcoin count

Where an exchange states how much bitcoin it holds, the bitcoin our node sees at its addresses now as a share of that statement. Above 100% is not an excess and below is not a shortfall: a statement can cover other networks or a custodian, and it is dated, while our reading is from the latest hour.

Reserves we see

Every covered coin at the addresses it lists (on Bitcoin, also those our wallet clustering ties to them), valued in US dollars at the hour's prices, from our own nodes. The bar shows the mix of coins and the size against the largest exchange here.

Net flow, 7 days

What arrived at its addresses minus what left over the past seven days, as a share of the reserves before. Prices don't move it; deposits and withdrawals do.

Drain score now

How unusual the exchange's net outflow over the past 24 hours is against its own previous 90 days, every coin we cover counted together in US dollars: zero is a normal day, below -2 an unusual outflow, below -3 with at least 1% of reserves gone a strongly unusual one. Its history, exchange by exchange, is a chart with Plus.

Flags, 30 days

Patterns our exchange monitor flagged on chain: a possible withdrawal pause (no withdrawals for longer than the exchange's usual gaps and chance explain, counted once per start whatever the coins) or an unusual outflow (reserves falling far more than its last 90 days suggest). A flag is a pattern, not a statement from the exchange.

Open interest ÷ reserves

The value of open positions in its bitcoin and ether perpetuals, one side, from its public API every hour, divided by the reserves we see. Customers post margin in many coins, so a ratio above 1 is not a shortfall. Bitfinex and Crypto.com perpetuals are not collected.

$1M bitcoin order

What a market order of one million US dollars in its bitcoin spot book would cost against the mid price, the average of a buy and a sell over the past 24 hours, in basis points (hundredths of a percent). From our own live copy of the order book, walked level by level each second; we keep the books of Binance, OKX and Bybit.

What the scorecard can't tell you

• It doesn't grade or rank exchanges, and it can't say whether an exchange could pay everyone back. That needs what it owes, which no chain shows.
• Every reading starts from the exchange's own lists, so reserves are a lower bound: coins at addresses it doesn't list, on other networks or with a custodian are not counted.
• An exchange that publishes no list can't be read at all. Coinbase and Kraken publish none, and we found none we can read for Gemini, Bitget, Gate or MEXC, so they aren't here.
• Columns are measured at different times: lists when the exchange publishes them, reserves and flows every hour, the order book over the past day.

The scorecard as data: /api/v1/exchanges?view=scorecard (free, no key). Free to reuse under CC BY 4.0, crediting CryptoKeyhole with a link to this page.