What Is a Mining Pool?

What is a share? The pool's easier bar

A fair split needs receipts. The network's bar is brutally high — at today's difficulty a single pull clears it about once in 5.7 × 10²³ tries (difficulty × 2³²; the live figure feeds the tile below) — so a pool that only counted actual finds would learn nothing about who did the work between them. The fix: the pool sets a second, far lower, private bar and accepts every pull that clears that one as a share: a receipt that you were pulling.

The definition is older than almost everything else in pooled mining — Meni Rosenfeld's 2011 analysis, still the reference text: one share = a hash that would have made a block if the difficulty were 1. Any single pull becomes a share with probability 1 in 2³²; every share then has probability 1 in D of also clearing the real bar, where D is the network difficulty. That known exchange rate between receipts and expected blocks is the whole trick — it is what makes any split checkable from the outside. A share is not a partial block: it contributes nothing toward finding anything. It only proves you were pulling.

Raw difficulty-1 shares would bury the pool in receipts: an S21-class machine at its 234 TH/s spec-sheet rating (manufacturer spec, not a consensus number) clears that bar roughly 54,000 times every second (2.34 × 10¹⁴ ÷ 2³²). So pools run variable share difficulty — vardiff — sliding each machine's private bar until its receipts land every few seconds; each pool tunes the exact cadence. A desk-sized Bitaxe gets a low bar, a rack S21 a high one, and both report on the same clock. The pool reads your hashrate as cadence × bar height, and the same bar-height bookkeeping makes a big machine's shares count proportionally more. The two cadences, side by side:

Every payout acronym answers one question

Between finds — minutes to hours apart for a big pool, days for a small one — somebody is holding the risk that the next one comes late. Every honest payout scheme is a position on that one axis and nothing else. PPS and FPPS put the dice in the operator's hands: your shares are bought at a fixed price the moment they land, found block or not. PPLNS leaves the dice in yours: nothing is owed until the party actually finds. Which position you should want is a decision, and the decision rule — plus the how-dry-is-dry arithmetic behind it — lives on /btc/pools. This lesson is the machinery underneath, so that rule reads as arithmetic instead of vibes.

PPS: the pool buys your receipts on the spot

Pay-per-share (PPS) is a standing buy order for receipts. Every share is paid its exact expected value the instant it lands:

The unit price looks absurd because it is quoted per difficulty-1 share — your vardiff shares are worth bar-height times more, and they arrive every few seconds. Blocks found or not, the drip is constant: you could mine a year on a PPS pool, watch it find nothing, and still be paid in full for every receipt. (3.125 BTC is the subsidy until the next halving, around April 2028 — next lesson.)

Someone still holds the dice, and under PPS it is the operator — who has effectively become a solo miner backed by the whole pool's hashrate, paying fixed salaries through every dry spell. Rosenfeld's paper calls PPS the riskiest scheme for the operator, derives the cash reserve an operator must hold to survive the swings, and shows an under-reserved PPS pool has a good chance of eventually going bankrupt. The insurance is real, so it is priced: PPS-family fees run consistently higher than variance-carrying ones. And one detail the search results reliably get wrong: classic PPS prices the subsidy only. Transaction fees are not in the formula above — folding them in is a different scheme with its own name.

FPPS: the same fixed price, with fees folded in

Full-pay-per-share (FPPS) keeps the insurance and fixes the missing-fees hole: each share is priced at subsidy plus a pro-rata estimate of transaction fees, taken from a trailing window of real blocks. Luxor, which documents its method, recomputes the rate per block height with a 144-block lookback — about a day of chain:

The trailing mean is the honesty mechanism: you are paid an estimate of fees smoothed over the window, not the fees of whatever blocks your pool happened to find. As of 2026, FPPS is the dominant scheme among the large pools — dated deliberately, because dominance is a market fact, not a law. The lineage, roughly: proportional rounds (2010) → PPLNS (~2011) → PPS as a priced product (~2012) → FPPS (~2014–2015). One sibling in one line: PPS+ pays the subsidy PPS-style but shares fees only from blocks actually found — the halfway house.

PPLNS: you keep the dice, and the premium

Pay-per-last-N-shares (PPLNS) owes you nothing until the party finds. When a block lands, its reward is split over the last N shares submitted — in the simplest variant, each of those N shares is paid (1 − fee) × reward ÷ N. The expected value per share works out identical to the PPS price; what changes is the shape. Payouts arrive when blocks do: a lucky week pays over the odds, a dry stretch pays zero. How dry is dry for a pool of a given size is Poisson arithmetic — worked for a 1% and a 0.1% pool on /btc/pools, and not re-derived here.

Two receipts from Rosenfeld's analysis: splitting over a window of N shares cuts a miner's reward variance by roughly a factor of N compared with going solo — and even with N set to a full difficulty's worth of shares, 36.79% of shares still receive zero payment. The dice stay real. The pool, meanwhile, carries nothing: it pays out exactly what it finds, minus the fee. No insurance sold, no reserve to defend, no bankruptcy math — which is precisely why PPLNS fee schedules sit lower.

Why PPLNS exists: the pool-hopping story

The first pools paid proportionally per round: when a block was found, it was split across all shares since the previous find. It feels fair, and it was exploitable — because rounds are memoryless. The next find is never “due,” no matter how long a round has run, so a share submitted early in a round was worth more than one submitted late: fewer shares to split with so far, same chance the round ends right now. Rosenfeld quantified the exploit — in theory, worst case: hopping into rounds younger than ~43.5% of difficulty in submitted shares was profitable, and an always-on loyal miner could earn up to ~43% less than its fair share.

Slush's pool — announced 27 November 2010, the first publicly available pool — shipped the first scoring method deliberately built to resist hopping. Its first blocks came in mid-December 2010: web sources disagree on the exact height (97,384 vs 97,834), and our own node timestamps those blocks 2010-12-13 18:39 UTC and 2010-12-16 07:16 UTC, so mid-December holds either way — the height attribution itself is community record, not chain data. Pool hopping was finally killed by deleting the thing it fed on: PPLNS has no rounds at all. Shares earn across a sliding window of the last N, so there is no “early” left to exploit — and that same window is why joining a PPLNS pool ramps up from zero while your shares age in, and why leaving forfeits the aging shares you walk away from.

What does 97% pool luck mean?

Pools publish a luck figure, and it is the most misread number on any pool stats page. The usual convention: luck = expected shares per find ÷ the shares it actually took, × 100. Run the arithmetic once and the number stops being mysterious. Say the window's blocks should have taken 100 million shares at the pool's bar and actually took 103.1 million: 100 ÷ 103.1 = 97%. The pool spent about 3% more pulls than expected — mildly unlucky. Above 100% means the finds came early; 200% means blocks landed at half the expected shares. It never means “the pool found 97% of its blocks.”

One trap before you read any pool's page: a few stats sites print the inverse — actual ÷ expected, where above 100% is unlucky. Same data, flipped fraction. Check which way is up before comparing anything across sites.

And the hard part: luck is a scorecard, never a forecast. The dice have no memory — every hash is as likely as every other — so last month's 97% says nothing about next month, and lifetime luck drifts toward 100% for every honest pool. The distribution math behind why windows wobble at all is the previous lesson's; here is what the wobble looks like when the search party is the entire network:

Scheme vs fee: the price of smoothing

Now the fee schedule reads differently. Steady pay costs more because someone is selling you insurance — an FPPS operator holding reserves through dry spells is running a risk business, and the fee is its premium. Variance-carrying pays less because nobody is insuring anyone. In 2026, published schedules mostly sit around 0–4% for FPPS and 0–2% for PPLNS — ranges, hedged on purpose: schedules churn, so the only current number is the one on the pool's own page, and we print no pool-by-pool fee table. Long-run expected earnings are identical under every honest scheme. You are choosing the shape of arrival and the price of smoothing — nothing else.

One note for the scrypt seat: on Litecoin pools, the merged-mined side coins — your DOGE leg — settle PPLNS almost everywhere, even where the LTC leg is FPPS. So a slice of scrypt revenue rides pool luck whichever scheme you pick; the receipts are in the /ltc/pools payout table, and the one-hash-two-chains mechanics are the merged-mining lesson's.

FAQ

How do mining pools pay out — what do PPS, FPPS and PPLNS actually mean?

They are three answers to one question: who holds the luck between finds. PPS buys every share you submit at its fixed expected value (subsidy only), found block or not — the pool holds the luck. FPPS does the same but prices shares at subsidy plus a trailing estimate of transaction fees — still the pool's luck, and the dominant scheme among large pools today (2026). PPLNS pays only when the pool actually finds a block, split over the last N shares — you hold the luck, and the fee is lower because nobody is selling you insurance.

What is a share in mining?

A receipt that you were pulling drawers. The network's bar is brutally high; your pool sets a far lower private bar and accepts every pull that clears it as proof of work done, sliding each machine's bar (share difficulty) so receipts land every few seconds. Each share has a known small chance of also clearing the real bar, which is what makes the split checkable — shares are evidence of effort, not partial blocks.

What does 97% pool luck mean?

The pool spent about 3% more shares than expected to find its blocks over that window — mildly unlucky, by the usual convention of expected shares divided by actual shares. It is a record of the past, not a forecast: the dice have no memory, and luck trends back toward 100% over long windows. A few sites print the inverse convention, so check which way is up before comparing.

Why do PPLNS pools charge lower fees than FPPS pools?

Because you are carrying the variance instead of the pool. An FPPS pool pays a steady expected value through dry spells and prices that insurance into its fee; a PPLNS pool owes nothing until a block lands, so it has no risk to charge for. Long-run expected earnings are the same under every honest scheme — you are paying, or not paying, for the smoothing.

What is pool hopping?

The exploit that killed per-round payouts. Early pools split each block over all shares in the round, but rounds are memoryless, so early shares were worth more — hoppers mined only young rounds and loyal miners ate the loss (in theory, worst case, up to ~43%). Scoring systems and then PPLNS deleted the “early” there was to exploit by paying over a sliding window of recent shares — which is also why PPLNS ramps up from zero when you join and forfeits aging shares when you leave.