Hashrate, and why it is an estimate rather than a record

A block header records the difficulty target it was mined against, the timestamp, and the nonce that produced a valid hash. It does not record how many attempts the miner made before finding that nonce, because the miner has no reason to publish the failures and no mechanism to prove them. The chain therefore contains no direct measurement of hash rate at all.

What the chain does contain is enough to infer it. Difficulty fixes the expected number of hashes required to find one block: the network's total hash rate multiplied by the expected time to find a block equals the work implied by the difficulty. Rearranged, the estimated hash rate is the difficulty-derived work per block divided by the observed average time between blocks. Every published figure is some version of that division.

The estimate is noisy because block discovery is a random process. Blocks do not arrive every ten minutes; they arrive at random intervals with a mean of ten minutes, and over a short window the observed mean can be far from the true one. An estimate computed over a single day can swing by tens of per cent without any change in the underlying hardware. Providers smooth this by averaging over longer windows — a week, a month, or a difficulty period — which is why two reputable sources can publish different current figures and both be defensible.

A second source of divergence is the treatment of the difficulty change itself. Difficulty is constant within a retarget period and steps at its boundary, so an estimate that assumes the current difficulty across a window that spans a retarget will be slightly wrong at the seam. Providers handle the seam differently, and the difference shows up as a small offset between their series.

Difficulty and hash rate are related by a constant, and the constant is the ten-minute target. Difficulty is a consensus parameter: it is a number written into every block header, agreed by every node, and changed only at a retarget boundary. Hash rate is a derived statistic: it is computed off-chain from difficulty and observed block times, and no two providers compute it identically.

The practical consequence is that difficulty is exact and hash rate is approximate. A reader can state the difficulty at a given block height and be certain of the answer, because it is in the header. A reader who states a hash rate must also state the window it was averaged over and the provider it came from, or the figure is not reproducible. This site follows that rule wherever it quotes a network figure.

The two move together over long horizons, because the difficulty adjustment exists precisely to keep block times near ten minutes as hash rate changes. A sustained rise in hash rate produces faster blocks, which produces a difficulty increase at the next retarget, which restores the interval. The difficulty adjustment page sets out that feedback loop, and the proof of work page explains why the work being counted is what secures the chain. The mining economics page covers what that capacity costs to run.