Bitcoin hashrate: distinguish an estimate from difficulty
Bitcoin network hashrate is an estimate inferred from observed work and time. Difficulty is a consensus parameter used to determine valid proof of work. A hashrate chart can change while difficulty remains unchanged, because short windows contain randomness in block arrival rather than direct readings of every mining machine.
The Bitcoin Core getnetworkhashps reference explicitly describes an estimate over a chosen number of blocks and a selectable height. Record those parameters. The method is not a census of equipment, electricity consumption, or independent mining operators.
Specify the estimation window
Keep the start and end blocks, their identities, elapsed-time convention, cumulative-work change, node version and query parameters. Conceptually, estimated hashrate is work represented by a chosen block interval divided by the interval’s measured seconds. Use compatible units before converting hashes per second into exahashes per second.
Do not multiply every block in a long interval by the final difficulty if difficulty changed within that interval. A cumulative-work method accounts for the work represented by each included block. If reproducing the exact Core result matters, inspect the documented implementation and endpoint handling rather than assuming a simplified classroom formula is byte-for-byte equivalent.
Compare several prespecified window lengths. A short window responds quickly and is noisy; a longer window smooths variation and reacts slowly. Choosing whichever window produces the most dramatic movement after seeing the chart is a form of result selection, not stronger evidence that machines entered or left.
Hypothetical equal work, different elapsed time
Assume two fictional samples each represent an accumulated 6 × 10^24 hashes of expected work. The first spans 60,000 seconds and the second spans 50,000 seconds.
| Sample | Work | Elapsed seconds | Estimated rate |
|---|---|---|---|
| A | 6 × 10^24 hashes | 60,000 | 100 EH/s |
| B | 6 × 10^24 hashes | 50,000 | 120 EH/s |
The second estimate is 20% higher. The arithmetic is correct under the assumptions, but it does not prove a 20% equipment expansion. Block discovery is random, and the elapsed-time sample can move even when underlying machine capacity is stable. These figures describe no actual period.
Suppose a separate difficulty observation is unchanged across the two samples. That is not a contradiction. The estimate and the consensus parameter answer different questions and update on different schedules. Label both series, including whether the estimate uses a rolling block window or a calendar window.
Test a claim about capacity
Break the claim into observed estimates, inferred physical capacity and identified operators. Hardware disclosures, pool reports or electricity evidence can add context, but each has its own coverage. A pool’s share of observed blocks does not necessarily reveal ownership of the machines contributing to it.
Check whether provider methodology changed at the apparent break. Revised timestamps, different block coverage, smoothed estimates and altered window lengths can create a visual shift. Reproduce a narrow interval from blocks before treating a provider chart as an independent physical observation.
When reporting uncertainty, avoid inventing precise confidence bands from an unspecified model. A simplified Poisson arrival model can illustrate sampling variability only under its assumptions; real capacity can also change within the window. Report the model, effective observation period and sensitivity to windows if you use it.
The final note should state “estimated hashrate under this method” and distinguish that from difficulty, installed nameplate capacity and network control. Recheck when the selected chain changes, the measurement provider changes its method, or additional operational evidence materially challenges the estimate’s interpretation.
Avoid comparing an end-of-day estimate with a calendar-day average as if they were synchronized observations. In a rapidly changing period, their underlying blocks may overlap only partly. List the actual block span next to each value and distinguish a true recalculation using the same sample from an average of previously published estimates.
FAQ
Does a daily hashrate increase prove more miners switched on?
No. It may reflect capacity changes, random block timing, or measurement choices. Compare windows and seek separate operational evidence.
Is difficulty just another name for hashrate?
No. Difficulty is a consensus work threshold parameter; hashrate is an inferred rate of work. They are related but are not identical measurements.