How BackPoW calculates solo mining odds

Most mining calculators tell you ‘coins per day’, as if mining paid out on a schedule. It doesn’t. Block discovery is a memoryless random process, and treating it as linear is the single most common mistake in solo mining. This page documents exactly what BackPoW computes, so you can check it, argue with it, or reimplement it yourself.

1. Why the odds are not linear

Every hash your machine computes is an independent trial with a tiny, constant probability of success. The network has no memory of your previous attempts: having mined for three years without a block does not make you ‘due’ for one. Formally, block discovery is a Poisson process, and the waiting time between blocks follows an exponential distribution.

The practical consequence is counter-intuitive, and it is the reason this site exists: if your rig has a 1-in-2 chance of finding a block in a year, it does not have a 1-in-1 chance in two years. It has about 75%.

2. Expected time to a block

The mean time T between blocks for your hashrate is T = network_hashrate ÷ your_hashrate × block_time. BackPoW uses this form because it holds on every chain regardless of how that chain defines difficulty. On Bitcoin-style chains it is equivalent to the familiar T = (D × 2³²) ÷ H, since network_hashrate = D × 2³² ÷ block_time.

3. Probability over a period

The chance of finding at least one block within a window t is P(X ≥ 1) = 1 − e^−t/T. This is what BackPoW reports for a day, a month and a year — not a linear fraction. For a target number of blocks k, the same Poisson model gives P(X ≥ k), which drives the multi-block scenarios on the calculator.

4. What the numbers are worth

A block is valued in native units and USD from live market pricing. Electricity uses a default of $0.069/kWh, adjustable on every calculator. BackPoW also publishes a Cost of Production (CoP) per coin — the implied cost to mine one unit — which is what makes it possible to say whether a coin is currently being mined at a loss.

5. Data sources and freshness

Network difficulty, hashrate, block time and block reward come from live network and pool data, refreshed daily. Hardware specifications (hashrate, power draw) and market prices are aggregated across multiple vendors. Every figure on the site is derived from these inputs — nothing is hand-entered per page.

6. Hardware figures: stock, tuned, and what CoP is not

Hardware rows come from aggregated pool telemetry, which mixes stock and undervolted rigs. They are not a stock-only benchmark, and the basis is not guaranteed to be identical from one row to the next: on Ravencoin today the RTX 3080 Ti sits at 100% of its board TDP and the RTX 3090 at 94%, while the RX 6700 XT sits at 52% and the RX 6600 at 61%. Rows measured at stock power and rows measured undervolted are ranked against each other in the same efficiency table.

On GPU algorithms this is not a rounding question. KawPow is the clearest case, because tuning moves power draw far more than it moves hashrate: miners report a stock RTX 3070 at 26 MH/s for 239 W (0.110 MH/W) against roughly 25 MH/s for 131 W (0.195 MH/W) once undervolted — slightly less hashrate, nearly double the efficiency. A per-device profitability verdict can therefore be wrong for a tuned rig by a wide margin.

The headline Cost of Production is a narrower figure than it is often read to be. It is the floor of the single most efficient machine we track at the stated electricity rate, with a per-class uplift for capital and facility overhead — not the cost borne by an average fleet, and not the cost borne by a tuned one. On Ravencoin that floor is set by a datacenter card at 0.270 MH/W, ahead of any tuned consumer GPU, so tuning barely moves the headline number even where it changes the per-device rows a lot. Read CoP as a floor and a trend line rather than as anyone’s electricity bill.

Labelling each row with its own power basis is work in progress. Until it lands, if you run a tuned rig, compare your measured hashrate-per-watt against the efficiency column rather than trusting the row’s profitability verdict. Measurements contributed with a stated protocol are welcome — that is the profile the dataset is thinnest on.

7. Honest limits

Difficulty changes over time; the odds shown assume the current difficulty holds.
Expected value is not a forecast. Over any single year the most likely outcome for a small rig is zero blocks, even when the expected reward looks favourable.
Pool luck, orphan/uncle rates and pool fees are not modelled in the base odds.
Prices are volatile; a block’s USD value is a snapshot, not a promise.

Use these numbers

All of this is available as a free public JSON API. Citation with a link back is all we ask.