15 November 2026 · Article · 23 min read · Nerdminer Store Team
Best Difficulty Explained: What the Number on Your NerdMiner or Bitaxe Really Means
Best difficulty is the highest difficulty of any hash your miner has found. How NerdMiner, AxeOS and NerdOS count it, what value to expect on each device after a day or a year, why it is a record and not progress, and when a low value means something is wrong.
Best difficulty is the highest difficulty of any single hash your miner has produced. Every hash a miner computes is a 256-bit number, and the smaller that number, the higher its difficulty. The best difficulty is simply the record: the one hash that came out smallest so far. To find a Bitcoin block, a single hash has to reach the network difficulty, which stood at about 134.8 trillion, written 134.8T, on 15 November 2026.
So the short answer to "what does best difficulty mean" is this: it is a high score, not a progress bar. A NerdMiner showing 2.4K or a Bitaxe showing 1.2G has not come any closer to a block than it was yesterday, because every hash is a fresh, independent draw. The number is still useful. It tells you whether your miner really does the work it claims to do, and it shows in plain figures how the odds of solo mining work.
This guide goes further than the usual one-line definition. It shows how the NerdMiner firmware and AxeOS on the Bitaxe and NerdOS, its branch for the NerdAxe family, actually record the value (the NerdMiner does it quite differently), calculates the best difficulty you should expect after an hour, a day or a year on each device, and explains why the number resets, drops or differs from the one on your pool's website. The figures are our own calculations with every assumption stated. Every statement about firmware behaviour comes from the source code, linked at the end.
What difficulty means for a single hash
A Bitcoin miner hashes an 80-byte block header twice with SHA-256 and reads the result as a 256-bit number. The protocol sets a target, and a block is valid only if its hash is at or below that target. Difficulty is a more readable way to express a target: it says how many times harder the target is than the easiest target Bitcoin ever had, the so-called difficulty-1 target.
The same scale works for any individual hash. Divide the difficulty-1 target by the hash value and you get the difficulty of that hash. Both firmware families do exactly this: the NerdMiner divides a constant named truediffone by the hash, and AxeOS and NerdOS run the same calculation in a function called test_nonce_value. A hash of difficulty 1 starts with about 32 zero bits. Every further zero bit doubles the difficulty, so the difficulty of a hash is roughly 2 to the power of its leading zero bits minus 32.
Hash values are uniformly random, so the chance that one hash reaches difficulty D or more is about 1 in D × 232, which is 1 in D × 4.29 billion. Everything else in this article follows from that one rule. The table translates it into time for the smallest and the most common home miner.
| Difficulty | Leading zero bits | Hashes needed on average | NerdMiner V4, 250 KH/s | Bitaxe Gamma, 1.3 TH/s |
|---|---|---|---|---|
| 1 | 32 | 4.3 billion | 4.8 hours | 3.3 milliseconds |
| 256 | 40 | 1.1 trillion | 51 days | 0.85 seconds |
| 65.5K | 48 | 281 trillion | 36 years | 3.6 minutes |
| 16.8M | 56 | 7.2 × 1016 | 9,100 years | 15 hours |
| 4.29G | 64 | 1.8 × 1019 | 2.3 million years | 164 days |
| 1.10T | 72 | 4.7 × 1021 | 600 million years | 115 years |
| 134.8T (network, 15 November 2026) | about 79 | 5.8 × 1023 | 73 billion years | 14,100 years |
The last row is the job a solo miner signs up for. The network difficulty adjusts every 2,016 blocks, roughly every two weeks, so check the current figure on a block explorer such as mempool.space when you read this. Our numbers use the value at block 975,943 on 15 November 2026.
Shares: the part of your work the pool sees
Your miner does not send every hash to the pool. On a Bitaxe Gamma that would mean more than a trillion messages per second. Instead the pool sets a share difficulty with the Stratum message mining.set_difficulty, and the miner submits only hashes at or above it. Each accepted share proves a known amount of work: a share of difficulty d stands for d × 232 hashes on average. That is how a pool estimates your hashrate without having to trust the number on your screen.
The share difficulty has to suit the miner. A Bitaxe Gamma at 1.3 TH/s finds about 300 hashes of difficulty 1 every second, so pools set its share difficulty far above 1 to keep the traffic sensible. A NerdMiner at 250 KH/s needs 4.8 hours on average for a single hash of difficulty 1. The NerdMiner firmware therefore asks the pool for a share difficulty of 0.00015 with the message mining.suggest_difficulty (the constant DEFAULT_DIFFICULTY in its source). At that level the miner finds a qualifying hash about every 2.6 seconds.
Whether the pool honours the request is up to the pool. In June 2026, public-pool.io, the pool the NerdMiner firmware has as its default, stopped accepting such low suggestions: the lowest share difficulty it now sends is 1. The change is documented in an open issue in the pool's repository, and on 14 June 2026 the NerdMiner maintainer confirmed that Public Pool no longer supports NerdMiners. For a 250 KH/s miner, difficulty 1 means one share every 4.8 hours on average, so the miner drops off the worker list on the pool website and its statistics stop moving.
The same arithmetic explains the strange hashrates pools show for small miners. A pool estimates your hashrate as the sum of your accepted share difficulties times 232, divided by the length of its time window. One share of difficulty 1 inside a ten-minute window therefore reads as 4.29 billion hashes in 600 seconds, or about 7.2 MH/s, nearly 30 times what a NerdMiner can do. The next window without a share reads as 0. When NerdMiner owners report "7 MH/s, but no worker shows up", this effect is the most likely explanation, not a faster miner. The estimate only becomes meaningful when many shares fall into each window, which is why a pool needs a low share difficulty for small miners in the first place.
The lottery ticket itself is not affected: a hash that wins a block is far above difficulty 1 and would still be submitted. What you lose is the feedback that the miner is working, and, as the next section shows, the NerdMiner's own best difficulty.
How NerdMiner, AxeOS and NerdOS record best difficulty
Both firmware families show a value called best difficulty, but they do not measure the same thing. That matters as soon as you compare your number with someone else's or with your pool.
AxeOS and NerdOS: every nonce the chips report
The ASIC chips in a Bitaxe, NerdAxe, NerdQAxe or NerdOctaxe report the nonces whose hash passes an internal filter. For every reported nonce on a Bitaxe, AxeOS calculates the difficulty and hands it to a function called SYSTEM_notify_found_nonce, whether or not the nonce is high enough to go to the pool. That function keeps two records. The best difficulty since the last start appears as bestSessionDiff in the API. The all-time best appears as bestDiff and is written to the miner's non-volatile memory, so it survives a restart. If a nonce reaches the network difficulty, the same function counts a found block.
NerdAxe, NerdQAxe and NerdOctaxe miners run NerdOS, a branch of this firmware for the NerdAxe family (version 1.1.2 at the time of writing). It handles the record the same way: every reported nonce goes through a check for a new best, whether or not it is submitted, with a session value and a stored all-time value. In other words, on all these miners the value describes the work of the chips themselves, independent of what the pool accepts. It is the cleaner of the two numbers, and it is the one the calculations below describe exactly.
NerdMiner: the best accepted share
The NerdMiner firmware calculates the difficulty of its hashes too, but it keeps only the best one of each batch of nonces and submits it if it beats the current pool difficulty. Its best difficulty is updated only when the pool replies that a share was accepted (the STRATUM_SUCCESS case in src/mining.cpp). So on a NerdMiner the number is strictly the best accepted share, not the best hash. With the low share difficulty the firmware asks for, the two are practically the same. With a pool that enforces difficulty 1, nothing below 1 ever counts, and rejected or lost shares never count at all.
Whether the value survives a restart depends on one checkbox on the setup page: "Save mining statistics to flash memory." It is off by default (DEFAULT_SAVESTATS is false in src/drivers/storage/storage.h). Without it, best difficulty, the share counter and the valid blocks counter start again from zero after every power cut. With it, the firmware stores them in flash, protected by a checksum, and resets them only if the checksum does not match.
Many NerdMiner screens show a second best difficulty: the one the pool reports for your Bitcoin address. In the firmware source, this figure is fetched from the public-pool.io API on most boards, whichever pool you mine on (src/monitor.cpp, discussed in issue #795). When that request fails, the firmware writes the letter P into the best difficulty field and E or Error into the hashrate field. Those letters are error markers, not the prefixes for peta and exa.
| AxeOS (Bitaxe) and NerdOS (NerdAxe family) | NerdMiner | |
|---|---|---|
| What counts | Every nonce the chips report | The best hash of each batch, once the pool accepts it as a share |
| Hashes below the pool difficulty | Counted | Not counted |
| After a restart | All-time value kept, session value starts at 0 | Starts at 0 unless "Save mining statistics" is ticked |
| Where you see it | Dashboard and API (bestDiff, bestSessionDiff) | Mining screen; pool screens show the pool's own value |
| Letters after the number | K, M, G, T, P for thousand up to quadrillion | The same letters; a lone P or E in the pool figures means the request failed |
What best difficulty to expect: the math
Every hash has the same tiny chance of reaching a given difficulty, so the best difficulty after N hashes follows a simple distribution. The probability that none of N hashes reaches difficulty D is:
P(best < D) = e−N / (D × 232), where N = hashrate × time in seconds
Solve for the point where that probability is one half and you get the median best difficulty, the value a miner beats in half of all cases:
median best = N / (232 × ln 2) ≈ hashrate × seconds / 2.98 billion
Three consequences follow, and together they answer most of the questions people ask about the number:
- It grows in proportion to time. Run the miner twice as long and the typical record doubles. A record that took a year to reach needs another year, on average, to double.
- It grows in proportion to hashrate. A NerdOctaxe at 14 TH/s reaches in one day the record a Bitaxe Gamma needs about eleven days for.
- The spread is huge. In 80% of cases the record lies between 0.30 and 6.6 times the median. Two identical miners running side by side end up ten times apart or more in almost one case out of five, and neither is broken.
You will not find an average in the table below, and that is deliberate. The distribution has such a heavy tail that its mathematical mean is infinite: a single lucky hash can be thousands of times above the median. The median and the 10% and 90% marks describe what you will actually see.
| Miner, hashrate from our catalogue | 1 hour | 1 day | 1 week | 30 days | 1 year |
|---|---|---|---|---|---|
| NerdMiner V4, 250 KH/s | 0.30 0.09 to 2.0 | 7.3 2.2 to 48 | 51 15 to 334 | 218 66 to 1.4K | 2.65K 798 to 17.4K |
| Bitaxe Gamma, 1.3 TH/s | 1.6M 473K to 10.3M | 38M 11.4M to 248M | 264M 80M to 1.74G | 1.13G 341M to 7.45G | 13.8G 4.15G to 90.7G |
| NerdAxe Gaia, 2.6 TH/s | 3.1M 946K to 20.7M | 76M 23M to 496M | 528M 159M to 3.47G | 2.26G 681M to 14.9G | 27.6G 8.3G to 181G |
| NerdQAxe++, 4.8 TH/s | 5.8M 1.75M to 38M | 139M 42M to 916M | 975M 294M to 6.42G | 4.18G 1.26G to 27.5G | 50.9G 15.3G to 335G |
| NerdOctaxe, 14 TH/s | 16.9M 5.1M to 111M | 406M 122M to 2.67G | 2.84G 856M to 18.7G | 12.2G 3.67G to 80.2G | 148G 45G to 976G |
Each cell shows the median and, below it, the range that covers 80% of cases, from the 10% mark to the 90% mark. The values assume the miner runs at the stated hashrate the whole time and, for a NerdMiner, that the pool accepts its low share difficulty. Reading the table: a NerdMiner V4 that has run for a month typically shows a best difficulty of a couple of hundred, and anything from about 66 to 1,400 is normal. If yours shows around 900 after four months, that is exactly what the math predicts.
Working it out for your own miner
You do not need a calculator page for this. Two shortcuts cover every miner: the median best difficulty is about 1.21M × hashrate in TH/s × hours of mining, or, for an ESP32 miner, about 0.029 × hashrate in KH/s × days. The 10% mark is 0.30 times the median, the 90% mark 6.6 times, and the 1% mark 0.15 times.
A worked example: a Bitaxe Gamma at 1.3 TH/s has run for ten days without a restart, which is 240 hours. The median is 1.21M × 1.3 × 240, about 377M. Values between 113M and 2.5G are what 80% of such miners show, and only one miner in a hundred stays below 57M. If the dashboard shows a session best of 52M after those ten days, the miner is either slower than its display claims or has not really been running all that time; if it shows 900M, it has simply been a little lucky. For a NerdMiner V4 at 250 KH/s after the same ten days, the median is 0.029 × 250 × 10, about 73.
Your best difficulty and your block odds are the same calculation
A block is nothing more than a hash whose difficulty reaches the network difficulty. So the chance of finding a block in a given time is the chance that your best difficulty over that time reaches 134.8T: 1 − e−N / (134.8T × 232). The best difficulty is your lottery ticket, seen from the other side.
That gives a neat rule of thumb. Take the median record your miner sets in one year and see how many times it falls short of the network difficulty: that factor is the median number of years to a block.
| Miner | Chance of a block in one year | Median time to a block | Median best after one year | Times below network difficulty |
|---|---|---|---|---|
| NerdMiner V4, 250 KH/s | 1 in 73 billion | 51 billion years | 2.65K | about 51 billion |
| Bitaxe Gamma, 1.3 TH/s | 1 in 14,100 | 9,800 years | 13.8G | about 9,800 |
| NerdAxe Gaia, 2.6 TH/s | 1 in 7,100 | 4,900 years | 27.6G | about 4,900 |
| NerdQAxe++, 4.8 TH/s | 1 in 3,800 | 2,650 years | 50.9G | about 2,650 |
| NerdOctaxe, 14 TH/s | 1 in 1,310 | 910 years | 148G | about 910 |
The median time is shorter than the "average" figure most solo calculators show (for the Bitaxe Gamma, 9,800 rather than 14,100 years), because blocks follow the same heavy-tailed logic as best difficulty. Both figures assume the network difficulty stays where it is; over the long run it has risen, which makes the odds worse over time. They also assume a pool that passes a winning share on to the network in time. Nothing else changes them: not the size of the pool, not your record so far, not the time since your last restart. Our guide to solo mining at home covers the practical side of that lottery.
Why a high best difficulty does not mean you were close
It is tempting to read a big record as a near miss. If your Bitaxe Gamma once found a 20G hash, was it one 6,700th of the way to a block? It was not, because there is no "way". Each hash is a new draw with exactly the same odds as the first one, and the hash that wins a block, if it ever comes, does not build on any earlier record. A record is like the highest number you have ever rolled with a die that has about 580 sextillion sides: it says something about how often you rolled, not about how close the next roll will be.
The same holds for streaks. A miner that has just set a record is not "warmed up", and one that has been unlucky for a month is not "due". Restarting the miner resets the session value on the screen and changes nothing else. The pool, the time of day and the share difficulty do not change the odds of a single hash either.
Where a high record does carry information is against the expectation. A Bitaxe Gamma that reaches 20G within one week is well above its 90% mark of 1.74G; the chance of that is about 0.9%. That is rare for one miner, but with thousands of miners running, such records turn up all the time. It is luck and nothing else, which is exactly what solo mining is.
What your best difficulty can tell you
Because the record grows so predictably with the number of hashes, it makes a useful plausibility check that does not depend on the hashrate your firmware displays. The displayed hashrate is a figure the firmware works out for itself; the best difficulty comes from the hashes themselves. In a 2026 discussion on the NerdMiner GitHub page about faster hashing code (issue #618), a contributor had to correct one of his own speed-ups: a board running it was, in his words, "very likely hashing into the void right now while reporting a perfectly normal rate". A speed figure cannot catch that. A best difficulty that stays far below the expected range is one of the few signs that show it from the outside.
As a rule of thumb, an all-time best below the 1% mark after a long, uninterrupted run is worth a closer look. For 30 days of continuous mining the marks are:
| Miner | 1% mark after 30 days | Median after 30 days |
|---|---|---|
| NerdMiner V4, 250 KH/s | 33 | 218 |
| Bitaxe Gamma, 1.3 TH/s | 170M | 1.13G |
| NerdAxe Gaia, 2.6 TH/s | 341M | 2.26G |
| NerdQAxe++, 4.8 TH/s | 629M | 4.18G |
| NerdOctaxe, 14 TH/s | 1.83G | 12.2G |
If your miner sits below that line, check in this order. Did it really run for 30 days without a restart? The uptime counter tells you. On a NerdMiner, is "Save mining statistics" ticked, and does the pool accept shares at all? On AxeOS and NerdOS, does the dashboard show rejected shares or chip errors, and does the hashrate your pool estimates from your shares match the one on the screen? Give the pool estimate at least a day of shares before you judge it.
The reverse check does not work. A high record proves that the miner produced at least one good hash; it does not prove that it runs at full speed.
Why your best difficulty reset, dropped or differs from the pool
It dropped after a restart
On AxeOS and NerdOS you are probably looking at the session value, which starts at 0 with every boot; the all-time value is still there. On a NerdMiner, the value restarts from zero after a power cut unless "Save mining statistics to flash memory" is ticked on the setup page. A fresh firmware flash that erases the memory resets it on both.
The pool shows a different number
The pool only sees the shares you submit, and it keeps its own records. Its figure belongs to your Bitcoin address, so several miners on the same address share one pool record, which is then higher than any single miner's. Depending on the pool software, the record can also reset when a worker reconnects or after a period without shares. And after switching pools, the new pool starts from zero while your miner keeps its all-time value.
My NerdMiner shows P and E
Then the firmware could not fetch the pool statistics for your address, usually because you mine on a pool whose API the firmware does not query. Mining is not affected. Check your miner on your pool's website instead, and ignore the letters.
It stays at 0
In the first minutes after a start that is normal on a NerdMiner, which displays any value below 0.001 as 0 (min_diff in src/utils.cpp). If it stays at 0 for hours, the miner has not had a single share accepted: check the pool address and port, and whether your pool still accepts the low share difficulty a NerdMiner asks for. On a pool that enforces difficulty 1, a NerdMiner needs 4.8 hours on average for its first share. The NerdMiner setup guide walks through the screens and the common connection problems.
What the letters after the number mean
K stands for thousand, M for million, G for billion (giga), T for trillion, P for quadrillion and E for quintillion. A best difficulty of 1.2G is 1.2 billion, and the network difficulty of 134.8T is 134.8 trillion. On a NerdMiner's pool figures, a P or E on its own is the error marker described above, not a unit.
If you want a higher record
There is exactly one lever: more hashes. Time and hashrate enter the formula in the same way, so a miner with five times the hashrate reaches in a week what the smaller one needs five weeks for. Settings, pools and restarts do not change it. That is also why the record says little about value for money; what you buy with more hashrate is a better chance at a block, and the table above shows how far even a 14 TH/s miner is from a sure thing. If you are choosing a miner, the guide to the best Bitcoin miners for home and the miner comparison set hashrate, power draw and noise side by side, and what a Bitaxe can mine explains why every one of them works on SHA-256d.
A NerdMiner plays the same lottery with odds that are, frankly, symbolic. Its best difficulty is still the most honest number on its screen: proof that a small ESP32 is doing real Bitcoin work, one hash at a time.
Frequently asked questions
What is a good best difficulty for a NerdMiner?
What is a good best difficulty for a Bitaxe?
Does a higher best difficulty increase my chance of finding a block?
How high must the best difficulty be to find a block?
Why did my best difficulty reset?
Why is my best difficulty different from the one on the pool website?
What does P or E mean on my NerdMiner?
Why does my NerdMiner no longer appear on public-pool.io?
Should I restart my miner to reset its luck?
Sources
The network difficulty is that of block 975,943 on 15 November 2026, as shown by mempool.space. The hashrates of the miners are the values on our product pages. All medians, ranges, marks and odds are our own calculations with the formulas shown above. Statements about firmware behaviour come from the source code of the NerdMiner firmware (main branch, June 2026; current release 1.8.3) and of ESP-Miner (AxeOS) 2.15.4 with its NerdOS branch 1.1.2:
- Bitcoin Wiki: Difficulty, definition and the formula for the expected time to a block
- Bitcoin Wiki: Target, how the target and the difficulty-1 target are defined
- Bitcoin Wiki: Stratum mining protocol, the messages mining.set_difficulty and mining.suggest_difficulty
- NerdMiner_v2: src/mining.cpp, share submission and the update of best difficulty on accepted shares
- NerdMiner_v2: src/monitor.cpp, the pool statistics and the P and E markers
- NerdMiner_v2: src/utils.cpp, the difficulty calculation and the display format
- NerdMiner_v2: src/drivers/storage/storage.h, the default for saving statistics
- ESP-Miner 2.15.4: main/system.c, session best, all-time best and block detection
- ESP-Miner 2.15.4: main/tasks/asic_result_task.c, how every reported nonce is checked
- ESP-Miner for NerdQAxe 1.1.2: main/tasks/asic_result_task.cpp, the same check in the firmware of NerdAxe, NerdQAxe and NerdOctaxe (NerdOS)
- NerdMiner issue #618, the correction about hashing into the void
- public-pool issue #152 and NerdMiner issue #782, the June 2026 change of the share difficulty on public-pool.io
- NerdMiner issue #795, the hard-coded pool statistics
Updated on 16 November 2026
Products mentioned in this article
−17%
−44%