Network Hashrate: Why It Is an Estimate, Not a Meter Reading

Hashrate is the combined speed at which all the miners on a network try out hash variants in search of a valid block; it is measured in hashes per second. The key point to grasp from the start: hashrate is not measured directly anywhere. There is no sensor counting the attempts. The network sees only the outcome — the blocks that someone has found and the difficulty at which they were searched for. Hashrate is derived from this by working backwards: if, at a given difficulty, blocks have started to arrive faster than the target interval, then computing power has been added. Any hashrate figure on any chart is a statistical estimate, not an observed fact.

How difficulty and hashrate hold each other in check

The protocol sets a target pace for block production and adjusts difficulty to match it — the threshold below which a hash must fall for the block to be accepted. The recalculation does not happen continuously but in windows: the network looks at how long it took to find the previous window of blocks and corrects the threshold.

This has an important consequence for reading the data: difficulty is a discrete and lagging quantity, while hashrate is a continuous and estimated one. When computing power leaves the network, difficulty stays high for some time, blocks arrive more slowly than the target pace, confirmations stretch out and fees in the mempool edge up. The opposite situation — an inflow of power — produces faster blocks until the next recalculation. Almost all the effects attributed to "hashrate" actually live in this gap between power and difficulty.

Why daily hashrate is noisy and what to do about it

Finding a block is a random process. Even with perfectly constant power, the intervals between blocks wander, and so the estimate wanders too. Over a short window the hashrate estimate is so noisy that it cannot be used to tell a real withdrawal of equipment from an ordinary streak of luck.

The practical conclusion is simple: look at smoothed values, and read short dips as noise until they are confirmed by other signs — a rise in average block time, a congested mempool, confirmation from the large pools. A sharp plunge on the chart is not an event in itself. The event is a plunge that lasts longer than the recalculation window.

What hashrate says about mining economics

A miner receives the block reward and fees, and pays for electricity and equipment. Revenue per unit of power — what is known as hashprice — falls as total hashrate rises: the reward is split among a larger number of working machines. That is why a rise in hashrate means worse conditions for the individual miner, even if the price of the coin has not changed.

This explains why the system is self-limiting. Power comes in as long as it pays for itself and leaves when it stops doing so. Shutdowns more often begin with older-generation machines and with regions where energy is expensive — the break-even line runs closer there. This is also where the seasonal waves come from: tariffs and the availability of energy depend on climate and local regulation more than on the network itself.

Hashrate as the cost of an attack, not a price forecast

High hashrate is above all the price of an attempt to rewrite the history of blocks. To impose your own version of the chain on the network and spend again coins that have already been sent, you need to control the larger part of the power — that is, either buy and power a comparable fleet of equipment, or rent it on the market. The higher the total hashrate, the more expensive such an attempt is and the lower the chance that it will pay off. In this sense hashrate is a security metric.

What it does not have is any predictive power over price. The common reasoning that "power is growing, so miners believe in a rally" turns causality upside down: equipment is ordered and brought online with a delay, so hashrate reacts to past economics rather than predicting future economics. Claims along the lines of "miner capitulation marks the bottom" are built on selected examples and do not stand up to testing against the full history.

Where hashrate does not exist at all

In proof-of-stake networks there is no mining — blocks are proposed by validators selected according to the size of the stake they have pledged. Comparing such a network with a mining network by hashrate is meaningless: the closest equivalent in meaning is the share of coins in staking and the cost of accumulating it. If an article or a dashboard shows you the "hashrate" of a proof-of-stake network, that is a data labelling error.

An honest caveat about the data

The TradeAlmanac database is built around the Russian market: instruments, financial reports, payouts, events. It contains no on-chain metrics — hashrate, difficulty, hashprice — so they cannot be inserted here with a directive, and we are not going to invent values. For the series themselves, go to blockchain explorers and pool aggregators, and check the smoothing methodology as you do: it differs from source to source, and a divergence between their estimates is normal, not a sign of an error.

What is available here, and how it is useful alongside this topic: explanations of terms in the glossary; issuers in the energy and IT sectors whose revenue is tied to data centres and tariffs — in the equity data and in the financial reports; regulatory news, which matters more for mining in Russia than any chart of computing power — in the market news; corporate dates — in the events calendar.