Ethereum Historical Flow Data Revisions Affect Trading Signals

Coin Metrics has recalculated Ethereum's historical Standard Flow Metrics, which could impact how traders interpret exchange outflows as signals. The company's October 1 update notes that it recomputed all ETH flow metrics using the latest information, affecting daily and hourly data. This change highlights a key distinction for investment research: charts downloaded today may reflect knowledge acquired later, while backtests require data available at the time decisions were made.

The issue stems from how exchange wallets are identified. Coin Metrics' Standard metrics use all currently known exchange addresses, but past values can be restated if new addresses are discovered. In contrast, Point-in-Time (PIT) series use only addresses known during the historical interval, preventing later discoveries from altering earlier data. This means the same historical date can tell different stories depending on when the data was compiled.

For analysts, this revision underscores the importance of data vintage, the version of data used in tests. While the immediate effect is the need to track data versions, any impact on returns requires further measurement. The notice also warns that historical values may change as exchange wallets are discovered and validated, a challenge highlighted by CryptoQuant's documentation. Glassnode provided an example using Bitcoin, showing how PIT data can lead to different trading decisions compared to revised balances.

To measure the rebuild's effect, paired observations from the same provider and metric are needed, with matching exchange coverage, intervals, and dates. The comparison should distinguish changed input values from signals, trades, and returns. Revised history can describe supply movements with today's address knowledge, but claims about trading edges require reproducible inputs, publication timing, and decisions.