The real maximum: 5,195,100x
Available at 12 mines and at 13, and nowhere else, because C(25,12) and C(25,13) are both 5,200,300. Those are also the odds: 1 in 5,200,300. A one-cent stake clearing that board returns about $51,951.00.
Mines has a closed form, so its payout table can be derived exactly rather than guessed at. The biggest win on record, 2,040,932.03x for $20,421.83, turns out to decompose to the cent, and it pins the house edge at 0.1%. Here is the whole schedule that follows.
Full-clear payout at every mine count, log scale. The ceiling sits at 12 and 13 mines, not at 24. Green marks the two settings that produce the record multiplier.
Big-win numbers are usually unfalsifiable marketing. This one is the opposite: it carries its own proof, and it tells you more about the game than any published RTP figure.
The largest win logged on Duel Mines is a multiplier of 2,040,932.03x, which paid $20,421.83. Divide the payout by the multiplier and the stake works out at a shade over one cent. That is a minimum-size bet, which is exactly what you would expect from someone hunting a full clear rather than a return.
Now look at the multiplier itself. In Mines there are only 24 possible full-clear payouts, one per mine count, and each is a binomial coefficient. 2,040,932.03 divided by 0.999 gives 2,042,975, and 2,042,975 is C(25,9) exactly, the number of ways to place 9 mines on 25 tiles.
Three things fall out of that, none of which appear on any other page covering this game:
That last point is what makes the whole table below computable. Once you know the edge, every multiplier in the game follows from the grid arithmetic, and there is nothing left to measure.
Mines needs no simulation and no long session to pin down, because it is a drawing-without-replacement problem with an exact answer. The grid holds 25 tiles. You choose how many carry mines, from 1 to 24, and the rest are gems.
Pick a tile at random with M mines on the board and it is safe with probability (25 − M) / 25. Pick again and the pool has shrunk by one, so the second pick is safe with probability (25 − M − 1) / (25 − 1). Chain those together and the chance of surviving K picks in a row is the product of (25 − M − i) / (25 − i) for i running from 0 to K − 1.
The payout is that probability inverted, scaled by whatever the operator keeps. At the 0.1% edge the record win establishes, cashing out after K safe picks pays 0.999 divided by the survival probability. Push K all the way to the last gem and the whole product collapses to a single binomial coefficient: a full clear pays 0.999 × C(25, M).
Every figure on this page comes out of those two lines. Nothing is rounded from a screenshot, nothing is averaged from a sample, and the tables are generated from the same code that produces the sentences around them, so the two cannot disagree.
The full-clear payout for all 24 mine counts, with the odds of getting there. Green is the ceiling, so it is also the answer to what the maximum win on Mines actually is.
| Mines | Gems to clear | Odds of clearing all | Payout | |
|---|---|---|---|---|
| 1 | 24 | 1 in 25 | 24.98x | |
| 2 | 23 | 1 in 300 | 299.70x | |
| 3 | 22 | 1 in 2,300 | 2,297.7x | |
| 4 | 21 | 1 in 12,650 | 12,637x | |
| 5 | 20 | 1 in 53,130 | 53,077x | |
| 6 | 19 | 1 in 177,100 | 176,923x | |
| 7 | 18 | 1 in 480,700 | 480,219x | |
| 8 | 17 | 1 in 1,081,575 | 1,080,493x | |
| 9 | 16 | 1 in 2,042,975 | 2,040,932x | the record win |
| 10 | 15 | 1 in 3,268,760 | 3,265,491x | |
| 11 | 14 | 1 in 4,457,400 | 4,452,943x | |
| 12 | 13 | 1 in 5,200,300 | 5,195,100x | the ceiling |
| 13 | 12 | 1 in 5,200,300 | 5,195,100x | the ceiling |
| 14 | 11 | 1 in 4,457,400 | 4,452,943x | |
| 15 | 10 | 1 in 3,268,760 | 3,265,491x | |
| 16 | 9 | 1 in 2,042,975 | 2,040,932x | the record win |
| 17 | 8 | 1 in 1,081,575 | 1,080,493x | |
| 18 | 7 | 1 in 480,700 | 480,219x | |
| 19 | 6 | 1 in 177,100 | 176,923x | |
| 20 | 5 | 1 in 53,130 | 53,077x | |
| 21 | 4 | 1 in 12,650 | 12,637x | |
| 22 | 3 | 1 in 2,300 | 2,297.7x | |
| 23 | 2 | 1 in 300 | 299.70x | |
| 24 | 1 | 1 in 25 | 24.98x |
Available at 12 mines and at 13, and nowhere else, because C(25,12) and C(25,13) are both 5,200,300. Those are also the odds: 1 in 5,200,300. A one-cent stake clearing that board returns about $51,951.00.
Both top out at 24.975x, and both are 1 in 25 shots. Clearing 24 gems while avoiding one mine and finding the single gem among 24 mines are, arithmetically, the same bet.
Every setting M pays exactly what setting 25 − M pays, because choosing where the mines go and choosing where the gems go are the same problem. 3 mines and 22 mines both cap at 2,297.70x.
What your stake is worth after each safe pick, at every mine count. This is the table none of the sites covering this game publish correctly; the ones that try are out by a factor of three in places.
| Setting | Gems | 1 pick | 3 picks | 5 picks | 10 picks | Full clear | Reach 5 picks |
|---|---|---|---|---|---|---|---|
| 1 mine | 24 | 1.041x | 1.135x | 1.249x | 1.665x | 24.98x | 80.00% |
| 2 mines | 23 | 1.086x | 1.297x | 1.577x | 2.854x | 299.70x | 63.33% |
| 3 mines | 22 | 1.135x | 1.492x | 2.016x | 5.050x | 2,297.7x | 49.57% |
| 4 mines | 21 | 1.189x | 1.728x | 2.608x | 9.258x | 12,637x | 38.30% |
| 5 mines | 20 | 1.249x | 2.016x | 3.423x | 17.67x | 53,077x | 29.18% |
| 6 mines | 19 | 1.314x | 2.371x | 4.565x | 35.35x | 176,923x | 21.89% |
| 7 mines | 18 | 1.388x | 2.816x | 6.195x | 74.63x | 480,219x | 16.13% |
| 8 mines | 17 | 1.469x | 3.379x | 8.577x | 167.91x | 1,080,493x | 11.65% |
| 9 mines | 16 | 1.561x | 4.103x | 12.15x | 407.78x | 2,040,932x | 8.22% |
| 10 mines | 15 | 1.665x | 5.050x | 17.67x | 1,087.4x | 3,265,491x | 5.65% |
| 11 mines | 14 | 1.784x | 6.312x | 26.51x | 3,262.2x | 4,452,943x | 3.77% |
| 12 mines | 13 | 1.921x | 8.034x | 41.24x | 11,418x | 5,195,100x | 2.42% |
| 13 mines | 12 | 2.081x | 10.44x | 67.02x | 49,477x | 5,195,100x | 1.49% |
| 14 mines | 11 | 2.270x | 13.93x | 114.89x | 296,863x | 4,452,943x | 0.870% |
| 15 mines | 10 | 2.498x | 19.15x | 210.62x | 3,265,491x | 3,265,491x | 0.474% |
| 16 mines | 9 | 2.775x | 27.35x | 421.25x | – | 2,040,932x | 0.237% |
| 17 mines | 8 | 3.122x | 41.03x | 947.80x | – | 1,080,493x | 0.105% |
| 18 mines | 7 | 3.568x | 65.65x | 2,527.5x | – | 480,219x | 0.040% |
| 19 mines | 6 | 4.163x | 114.89x | 8,846.1x | – | 176,923x | 0.011% |
| 20 mines | 5 | 4.995x | 229.77x | 53,077x | – | 53,077x | 0.00188% |
| 21 mines | 4 | 6.244x | 574.43x | – | – | 12,637x | – |
| 22 mines | 3 | 8.325x | 2,297.7x | – | – | 2,297.7x | – |
| 23 mines | 2 | 12.49x | – | – | – | 299.70x | – |
| 24 mines | 1 | 24.98x | – | – | – | 24.98x | – |
Cumulative multiplier on your stake after the stated number of safe picks. A dash means the depth does not exist at that mine count: at 20 mines there are only 5 gems on the board, so 5 picks is the full clear. The last column is the probability of getting five safe picks in a row. Computed from the hypergeometric survival product at the 0.1% edge the record win implies.
Search for Mines multipliers and you will meet 5,148,297x repeatedly, usually attributed to the 24-mine setting with a single gem found, and described as a 1 in 25 shot that lands several times a day. Every part of that is wrong, and it is worth unpicking because the number itself is real.
The setting is wrong. At 24 mines there is exactly one gem among 25 tiles. Finding it is a 1 in 25 event and pays 24.975x. It is the joint-cheapest outcome in the entire game, tied with 1 mine. Crediting it with a seven-figure multiplier is out by a factor of roughly 206,138.
The edge is wrong. 5,148,297 is 5,200,300 multiplied by 0.99. In other words it is the 12-mine full clear priced with a 1% house edge, the industry-standard Mines ceiling used by operators charging ten times what Duel charges. On Duel's own 0.1% the same clear pays 5,195,100x.
The frequency is wrong. A 12-mine full clear is a 1 in 5,200,300 event, not 1 in 25. At one round per second, non-stop, you would expect to see it about twice a year.
The irony is that the honest figure is the more impressive one. Duel's ceiling is higher than the number being misquoted, because a thinner edge means a larger payout on the same improbable event. There is no need to invent anything.
Choosing the number of mines feels like choosing a difficulty setting, and it is not. Because the payout at every depth is calibrated to the survival probability at that depth, all 24 settings return the same fraction of your stake over the long run. Three mines and twenty mines cost you exactly the same per dollar wagered. What differs is the shape: low counts pay small and often, high counts pay enormously and almost never.
The default 3-mine board, depth by depth: what you cash out for, how often you get there, and the two multiplied together. That last column is the whole argument.
| Safe picks | Cash out for | Chance of getting here | Expected return |
|---|---|---|---|
| 1 | 1.135x | 88.00% | 99.9% |
| 2 | 1.297x | 77.00% | 99.9% |
| 3 | 1.492x | 66.96% | 99.9% |
| 4 | 1.728x | 57.83% | 99.9% |
| 5 | 2.016x | 49.57% | 99.9% |
| 6 | 2.371x | 42.13% | 99.9% |
| 7 | 2.816x | 35.48% | 99.9% |
| 8 | 3.379x | 29.57% | 99.9% |
| 9 | 4.103x | 24.35% | 99.9% |
| 10 | 5.050x | 19.78% | 99.9% |
| 11 | 6.312x | 15.83% | 99.9% |
| 12 | 8.034x | 12.43% | 99.9% |
| 14 | 13.93x | 7.17% | 99.9% |
| 16 | 27.35x | 3.65% | 99.9% |
| 18 | 65.65x | 1.52% | 99.9% |
| 20 | 229.77x | 0.435% | 99.9% |
| 22 | 2,297.7x | 0.043% | 99.9% |
The expected-return column is flat at 99.9% all the way down, which is what it means for a payout table to be calibrated. Cashing out at three picks and holding to twenty-two are equally good decisions in expectation and wildly different decisions in practice: one pays 1.492x half the time, the other pays 2,297.7x once in 2,300 attempts.
Duel prints a Zero Edge badge under the Mines board, and its own explainer makes a real argument for why 0% is not merely a small improvement on 1%: the edge is charged on every bet, so it compounds over volume, and only at exactly zero do a player's chances of a large win and a large loss become symmetric.
The record win is the cleanest available test of that claim, and it comes back at 0.1% rather than zero. A full clear at 9 mines on a genuinely edge-free table pays 2,042,975x; the logged win paid 2,040,932.03x. The gap is small, consistent, and real. It is also exactly the 99.9% native return that two of the third-party write-ups quote, which is a useful piece of corroboration: an independent record figure and a published RTP landing on the same number is not a coincidence.
None of which makes Mines a poor game to sit at. A 0.1% edge is between a fifth and a tenth of what Duel's own Plinko tables keep, a fraction of any slot in the lobby, and thinner than every table game in a land-based casino. Wager $10,000 on Mines and you expect to lose $10.00. That is a genuinely excellent price. It simply is not free, and the difference between 0.1% and zero is worth knowing before you believe a badge.
The failure mode Mines players actually worry about is a board that decides where the mines are after reading your click. The commitment scheme exists to rule that out, and it does so in a way you can check rather than trust.
Before the round the server generates a seed and publishes its hash. You supply a client seed of your own, and a nonce counts the round. Those three values are hashed together, and the digest is consumed in windows, and each window selects one tile for one mine and removes it from the pool, so no two mines land on the same square. The result is all 25 positions assigned before your first pick. Rotate the seed pair afterwards and the server reveals the plaintext seed it committed to: hash it, compare against the commitment you were shown, and re-derive the layout to confirm the mines sat exactly where the round claimed.
Two things this does not prove, both worth stating plainly. It does not make the game favourable: a verified layout with a 0.1% edge is still a 0.1% edge, because the edge lives in the payout table and not in the shuffle. And it says nothing about tile position: because the layout is uniform over all 25 squares, corners, edges and centre are identical, and any guide recommending diagonal openings is selling superstition. The full verification walkthrough covers the seed rotation flow as it applies to Mines, Dice, Crash and the rest of the originals.
A multiplier of 2,040,932.03x, which returned $20,421.83 from a stake of just over one cent. The number is not arbitrary: 2,040,932.03 is 0.999 multiplied by 2,042,975, and 2,042,975 is C(25,9) exactly. That identifies the round as a complete clear, every single gem revealed, on the 9-mine setting, where 16 gems sit among 25 tiles. The odds of that happening are 1 in 2,042,975.
5,195,100x, and it is only available at 12 or 13 mines. Both settings pay the same because C(25,12) and C(25,13) are both 5,200,300. Reaching it means clearing all 13 gems with 12 mines on the board, a 1 in 5,200,300 event. Several write-ups credit a figure of 5,148,297x to the 24-mine setting, which is wrong twice over: at 24 mines a single gem pays 24.975x, and 5,148,297 is the 12-mine ceiling priced at a 1% house edge rather than Duel's 0.1%.
0.1%, which the record win pins down precisely. If the game paid a true zero edge, a full clear at 9 mines would return 2,042,975x. The recorded win paid 2,040,932.03x, which is 0.999 of that. A tenth of a percent is thinner than any live table game and thinner than every slot in the same lobby, but it is not zero, and the Zero Edge badge under the board does not change the arithmetic.
It makes no difference to what the game costs you. Every mine count from 1 to 24 returns the same fraction of your stake over the long run, because the payout at each depth is calibrated to the survival probability at that depth. What changes is the shape of the outcome: 1 mine pays small and often, 20 mines pays enormously and almost never. Anyone telling you 5 to 7 mines is mathematically optimal is describing a preference, not a maths result.
Whenever you like, because every depth carries identical expected value. On the default 3-mine setting, cashing out after 3 safe picks pays 1.492x and gets there 66.96% of the time; holding to 10 picks pays 5.050x and gets there 19.78% of the time. Multiply payout by probability in either case and you get the same 99.9%. The choice is how much variance you want, not how much return you get.
No. The whole grid is committed before your first pick, and the mines are distributed uniformly across all 25 positions. Corners, edges, diagonals and centre are identical. Any pattern you notice across rounds is the ordinary clustering that random placement produces, and one of the third-party guides that recommends diagonal spreads contradicts itself a few paragraphs later by admitting position carries no information.
The opposite. You are drawing without replacement from a shrinking pool, so every gem you remove raises the share of mines among the tiles left. With 1 mine the first pick is 96% safe; after ten gems, fifteen tiles remain holding that one mine, so the next pick is 93.3% safe; after twenty-two gems only three tiles are left and the next pick is 66.7% safe. That falling curve is exactly why the multiplier climbs.
Yes, and it is the strongest part of the game. Before the round Duel hashes a server seed and shows you the hash, you supply a client seed, and a nonce counts the round. Those three are hashed together and the digest is consumed in windows, each selecting one tile for one mine without replacement, so the full 25-tile layout is fixed before you touch anything. Rotate the seed pair afterwards and the plaintext server seed is revealed: re-hash it, check it against the commitment you were shown, then re-derive the layout and confirm the mines were exactly where the round said they were.
18+ · Gamble responsibly · Payout schedule computed 11 August 2026 from the closed form, calibrated to the 2,040,932.03x record win · Game at duel.com/mines · Affiliate link disclosed