How Minerstat calculates profitability

Every figure on Minerstat answers one question: what would this hardware earn over the next 24 hours if network conditions and prices stayed exactly where they are right now. It is an estimate of a moving target, not a promise, and it is only as good as the inputs listed on this page.

The daily profit of a device on a coin is:

coins/day   = hashrate × 86400 × blockReward / (difficulty × 2^32)   (network path)
revenue/day = coins/day × price
pool fee    = revenue/day × poolFeePercent
electricity = power(W) / 1000 × 24 × pricePerKWh × units
profit/day  = revenue/day − pool fee − electricity

The exact hashes-per-block relation depends on the algorithm family; for networks that publish a block time instead of a difficulty in that form, we use the network hashrate and the observed cadence of recent blocks rather than the nominal one, because a stalled chain otherwise inflates the reward.

When we have live difficulty and block reward for a chain, we take the network path above. It is the one we trust and the one that drives the default ranking.

Multipools and hashrate marketplaces do not have a difficulty of their own: they publish an expected reward per unit of hashrate. Those rows take the reward path, which is capped against what the same algorithm actually pays on verifiable networks. A reward figure that claims several times the best network return is treated as a stale listing, not as an opportunity.

Pages are served with a reference electricity price of $0.080/kWh, stated next to every ranking. Setting your own price anywhere on the site re-derives every profit figure you see, and the preference is kept in your browser rather than in an account.

Pool fees are applied when we know them for the pool a row refers to. Where a pool publishes several fee schemes, the cheapest scheme applicable to a solo miner is used. Payout thresholds, withdrawal fees and network transaction fees are not deducted: they depend on how often you cash out.

A device is described by its hashrate on an algorithm and the power it draws while doing it. Benchmarks come from several places and disagree more often than you would expect, so each one carries its own confidence and the ranking uses the best-supported figure rather than the highest one.

Benchmarks that are far out of line with the rest of the fleet on the same algorithm are demoted rather than deleted, and a device with no verified benchmark for an algorithm is shown as having no benchmark. We never render an unknown value as zero.

Every metric carries the time its source last changed, and pages state it. These are the targets we hold ourselves to; a source that genuinely updates less often is judged against its own cadence, not against this table.

Stale data lowers confidence, is labelled where it matters, and does not take the top of a default ranking when a fresher alternative exists.

Confidence answers a different question from profit: how much should you trust this estimate? It is scored out of 100 from the signals we can observe, and every entity page can show the breakdown behind its own number.

• Freshness: how recently the price and network data behind the estimate changed.
• Market data: whether the coin has a usable price and enough traded volume for that price to mean something.
• Network data: whether difficulty, hashrate and block reward are known rather than inferred.
• Pool coverage: how many reachable pools actually mine the coin. One pool is a dependency, not a market.
• Source coverage: how many independent collectors agree on the network state.
• Identity: whether the project has a reachable site and a working explorer.

Hashrate marketplaces and multipools are scored on what applies to them instead: they have no price of their own (they pay in someone else's coin), no listed pools because they are one, and no explorer. What counts there is whether the payout currency is liquid, how fresh the quoted reward is, and above all how much hashrate the venue's own market holds. A market holding a rounding error of an algorithm's hashrate is quoting a top bid, not a payout anyone can be paid at volume, so that alone caps its confidence at experimental.

75 and above is reported as high confidence, 45 to 74 as medium, and anything below 45 as experimental. Experimental rows are kept out of default rankings and shown only when you ask for them.

The opportunity score is the economic half of the picture: profit margin relative to comparable networks, liquidity, network health, how concentrated the hashrate is across pools, data confidence and short-term stability, blended into a single 0 to 100 figure.

It exists so that a thin, volatile network cannot look identical to an established one at the same headline profit. We keep it deliberately separate from confidence: a real opportunity you should not trust yet, and a trustworthy opportunity that is not worth much, are different situations and deserve different answers.

The same vocabulary is used everywhere on the site:

• Low liquidity: the coin trades too thinly for its price to survive being sold into.
• Low volume: daily traded volume is below what we consider meaningful for a mining decision.
• Single pool dominance: one pool holds more than half of the network hashrate.
• Volatile difficulty: difficulty is swinging enough that today's reward is a poor guide to tomorrow's.
• Price volatility: the price moved sharply over the window used for the estimate.
• New network: launched recently enough that its economics have not settled.
• Low data confidence: the inputs behind the number are incomplete or ageing.

Automated checks flag improbable records: a device claiming an implausible share of a network, a profit that jumps by an order of magnitude, a price that two sources disagree about, a coin with no reachable pool, a chain that has not produced a block in hours.

Flagged records are not deleted. They are withheld from default rankings and queued for human review, because a genuine anomaly and a broken feed look identical for the first few minutes, and deleting the evidence makes both impossible to diagnose.

• Profitability is a snapshot. Difficulty rises, prices move, and a number that was right this morning can be wrong tonight.
• We model a device, not a farm: no PSU inefficiency, no cooling overhead, no downtime, no hosting fee, no hardware amortisation.
• Pool luck is ignored. Expected reward is a long-run average, and small miners on small pools will see it vary considerably.
• Prices assume you can actually sell what you mine at the quoted price, which is exactly what the liquidity signals exist to warn you about.
• Historical series start when we began recording them; where a chart is short, it is short because the archive is, not because the network is new.