Anatomy of Block 957,382: What Actually Happened

On 9 July 2026, at roughly 03:30 UTC, a Bitaxe Gamma averaging 995.2 GH/s solved Bitcoin block 957,382 and took the full 3.1382 BTC. Two things everyone repeated about it are wrong. The winning share was below median, not impressive. And the eight-hour runtime was a coincidence, not a fact about the device.

The story travelled fast, and the numbers travelled with it: a ~$150 device, a ~$200,000 reward, a 294 trillion share against a 134 trillion target. Every figure is accurate. The interpretation stapled to them is not. This is what the arithmetic actually says.

Key takeaways

• The odds were roughly 1 in 6.7 million per day. 995.2 GH/s against a network difficulty of 133.87 trillion gives an expectation of one block every 18,308 years.
• The winning share was below median. At 2.20 times the network target, roughly 45.5% of all Bitcoin blocks ever mined beat their target by more.
• The reward was 3.1382 BTC — the 3.125 BTC subsidy plus about 0.0132 BTC in fees. Fees were just 0.42% of the total.
• Network hashrate was near a yearly low, around 908 EH/s on 11 July, about 21% below the 1,154 EH/s peak of October 2025.
• Difficulty fell 5% two days later, to 127.17 trillion at block 957,600 — the fourteenth adjustment of 2026 and its third lowest reading of the year.
• Expected value at that hashrate is about 3 US cents a day. The block was a lottery outcome, not a return.

What actually happened

The facts that hold up across independent reporting and the on-chain record:

Sources: mempool.space block record; contemporaneous reporting from Bitcoin.com, Crypto Briefing, Cointelegraph and Cointribune, 9–14 July 2026. Difficulty figures from the 11 July retarget at block 957,600.

The device was the only worker on the address, and the pool charges no fee on solo work, so the entire reward reached the miner. That is the whole mechanical story. Everything interesting is in what the numbers mean.

Why “more than double the target” is not impressive

This is the claim that travelled furthest. A widely shared post noted the winning share hit 294 trillion difficulty against the 134 trillion needed, and called it impressive. Nearly every outlet repeated the framing.

It is backwards. Here is why.

A hash output is uniform across its range. A share is valid when the hash falls below the target. So conditional on a hash being small enough to win at all, the distribution of how much smaller is fixed by simple geometry:

P(share difficulty exceeds k times network difficulty, given a block was found) = 1 / k

That is a Pareto distribution with shape parameter one, and it produces a table every solo miner should memorise:

The median winning share beats the target by exactly two. This one beat it by 2.20. That places it slightly above the middle of the pack, which means roughly 45 out of every 100 blocks in Bitcoin’s history overshot by more.

Two consequences follow, and both matter more than the trivia.

An unusually large overshoot buys you nothing. The block is valid or it is not. A share at 100 times the target and a share at 1.001 times pay identically. Overshoot is a measurement artefact, not an achievement.

The average overshoot does not exist. For a Pareto distribution with shape one, the mean diverges. Any article quoting an average winning-share difficulty is quoting an artefact of its sample size. This is the same statistical shape that makes solo mining feel unfair: rare events dominate, and averages mislead. Our Poisson maths explainer covers why the same reasoning applies to waiting times, and best share explained covers the related record your dashboard keeps.

Why “eight hours” tells you nothing

The second sticky claim was that the device found a block after running for only about eight hours.

What the dashboard actually showed was roughly eight hours of session uptime. That is when the worker connected, not when the owner bought the machine, and it says nothing about the hardware.

Mining is memoryless. Each hash is an independent trial, so a device that has hashed for three years is exactly as likely to solve the next block as one switched on this morning. There is no accumulated progress and no “due” state.

The correct reading of the eight hours is the reverse of the popular one. At 995.2 GH/s against 133.87 trillion difficulty, the probability of a find inside any specific eight-hour window is about one in 20 million. The short runtime does not make the story more impressive because the device did something well. It makes it a more extreme draw from the same distribution, which is a statement about luck and nothing else.

If a headline implies short runtime is evidence of a good machine, the causation runs the wrong way.

What the odds actually were

Odds in solo mining depend on two numbers and nothing else: your hashrate and the network difficulty. Everything else is noise.

Calculated as difficulty multiplied by 2³², divided by hashrate. Difficulty values are the epoch in force at block 957,382 and the epoch that began at block 957,600.

This table also settles a discrepancy readers noticed in July. Some outlets said the odds were one in 16,000 years, others said 18,000. Both are correct. They used different device hashrates: 1.2 TH/s nameplate versus the 995.2 GH/s actually measured. Always check which figure a source is using before comparing.

To put the scale in perspective, reaching an expectation of one block per century would take about 182 TH/s at that difficulty, which is roughly 150 Bitaxe Gammas running together. Our solo mining odds page has the full curve, and can a Bitaxe find a block covers the documented history.

The context nobody reported

July 2026 was a slightly better than usual month to get lucky, and this went unmentioned in the coverage.

Network hashrate had fallen hard. The seven-day average was about 908 EH/s on 11 July, down 7.9% from roughly 986 EH/s ten days earlier, and about 21% below the 1,154 EH/s peak of October 2025. Hashprice was near $31.10 per PH/s, still around 37% below its October 2025 high.

Difficulty follows hashrate with a lag, and 2026 has been a year of retreat. Eight of the fourteen adjustments in the year were negative. The 127.17 trillion set on 11 July was the third lowest reading of 2026.

Lower difficulty means better odds for everyone, including a single Bitaxe on a shelf. It is a modest effect, on the order of a fifth to a quarter compared with the peak, but it is real and it is the most plausible mechanical driver behind the rise in solo finds. It is also the reverse of the usual narrative: the door opened slightly wider because large operators pulled back, not because home hardware got better.

What 3.1382 BTC is actually made of

The breakdown is worth reading carefully because it is the part that affects how you set up your own miner.

• 3.125 BTC subsidy, fixed by the protocol until the next halving
• ~0.0132 BTC in transaction fees, which is 0.42% of the total

Fees were low that night. In busier fee environments the same block could have carried several times that, and in the highest-fee periods of past cycles fees have exceeded the subsidy entirely. Your payout depends on the block template your pool builds, which is a decision made off your device.

This is also where the payout model matters. The full amount reached the miner because the reward was paid directly in the coinbase transaction rather than routed through a pool balance. There is nothing to withdraw and nothing to claim. Our tracker of confirmed solo wins links the on-chain coinbase output for each documented block, and you found a block, now what covers the 100-block maturity wait that follows.

Where this sits in the record

Reported alongside the win were figures that have since been quoted widely:

• 24 solo blocks in the trailing twelve months, up 41% year on year
• 75.44 BTC paid out across those blocks
• Average interval of 15.2 days between finds
• 12 blocks found by hobby-level miners since the start of 2026

Treat these as directional. The tracker they come from states plainly that attribution is imperfect and incomplete, because identifying a solo miner from on-chain data alone is not always possible. A block mined solo through an unusual route may never be counted, and a block counted may have been mined by a small farm rather than a hobbyist.

The reported 2026 wins are worth reading side by side, because they show how differently the same outcome can be reached:

Note the spread in hashrate: 157 TH/s and 995.2 GH/s are separated by a factor of 158, and both produced a block within six weeks of each other. That is variance doing exactly what the maths predicts, not evidence that either setup was better. Our rented hashrate page covers the February route in detail.

What actually moves your odds

Three things, in order of effect.

1. Hashrate. Linear. Doubling your hashrate exactly doubles your chance per unit time. Nothing else in solo mining has this property.
2. Uptime. A device that is throttling, disconnected or rebooting is buying no tickets. This is the most commonly wasted edge, and it is free to fix.
3. Difficulty. Outside your control, but worth watching. The same device had roughly 5% better odds after the 11 July retarget than before it.

Everything else is either neutral or a distraction. Overshoot is meaningless. Runtime is meaningless. The pool you choose sets the fee you pay and decides whether the reward lands in your wallet or in a pool balance, but it does not change the probability of finding the block.

Solo mining is a lottery with honest, calculable tickets. Block 957,382 is a good reminder that the tickets are real, and an equally good reminder that expected value at 1 TH/s is about three cents a day. Both statements are true at once, and any article that gives you only one of them is selling something.