Coin Metrics said on October 1 that it had rebuilt Ethereum's historical Standard Flow Metrics using its latest information. The change affects every ETH flow metric at daily and hourly intervals, from the network's beginning to the present. For readers comparing exchange outflow charts, the update raises a practical question: which version of the historical data are you looking at?
The completion notice says corrected history is available for backfilling, meaning replacing previously stored values with the recalculated series. Coin Metrics describes the work as part of its Ethereum Point-in-Time release. The notice does not quantify how much individual observations changed or establish an effect on trading returns.
Wallet identification explains why an old chart can move without a new transfer occurring on that old date. CryptoQuant's ETH exchange-flow documentation warns that its own historical values can change as exchange wallets are discovered, added and validated. That endpoint explicitly does not promise point-in-time accuracy. This is a separate provider's methodological warning, not independent verification of Coin Metrics' recalculation.
Consider a hypothetical wallet that received ETH in June but was only identified as belonging to an exchange in September. A revised June chart could include that wallet. A June trading decision could not have used the September identification. The blockchain transaction and the analyst's knowledge of who controlled the address have different timelines.
That difference matters when replaying a strategy through past data, a process called backtesting. A test using today's revised history can accidentally give an earlier decision information learned later. It can also produce different signals without proving that either data series predicts prices reliably.
A reproducible comparison should retain the original download, record the provider, retrieval date, metric and interval, then compare it with the revised series under unchanged rules and fees. Readers should also distinguish a revised historical chart from a dataset designed to represent what was known at each point. No such ETH return comparison was performed for this article.