Crypto Keno odds & RTP

Not sure which to choose? See what we recommend for each type of player.

The game the odds are computed for

Every figure on this page is for the keno that crypto casinos run as an in-house Original: a board of 40 numbers, a player who marks between 1 and 10 of them, and a draw of 10. Stake's fairness documentation describes the draw this way: "Traditional Keno games require the selection of 10 possible game events in the form of hits on a board." It adds that "Once a hit has been placed, it cannot be chosen again, which changes the pool size of the possible outcomes" (1 Nov 2026). BC.Game's help centre gives the same board: "Keno Single is played by choosing up to 10 numbers from a set of numbers ranging from 1 to 40" (article dated 28 November 2025). Cloudbet says of its own version that it "uses a 40-number board: you pick 1 to 10 numbers, and the game draws 10" (13 Nov 2026).

Two settings are yours: how many numbers to mark, and on most of these games a risk level. Gamdom's help centre lists the usual four: "Players can choose a risk mode - Classic, Low, Medium, or High - each with custom volatility and payout tables to suit different playstyles" (13 Nov 2026). The risk level picks a paytable. It does not touch the draw, which is the same ten numbers whatever you marked and whichever table you chose.

One casino's pages do not agree with each other on the size of the draw. Shuffle's Keno game page says "Then 20 winning numbers are randomly drawn", while its fairness documentation says the game "requires the identification of 10 unique gem placements on the keno board" and prints code that produces ten numbers (both 13 Nov 2026). The documented algorithm is the one a draw can be checked against, and it draws ten.

Hit odds on a 40-number board

With 10 of 40 numbers drawn, the chance of matching exactly h of your p picks is hypergeometric, the same arithmetic as drawing cards without replacement:

P(h hits from p picks) = C(10, h) × C(30, p − h) ÷ C(40, p)

Arithmetic from the formula above; it applies to any keno that draws 10 of 40. The average is picks ÷ 4, because a quarter of the board is drawn.

Three things in that table explain most of what players report about the game. First, misses are the normal result: with one number marked you miss three times in four, and even with ten marked you get no hit at all in about one round in 28. Second, the typical round lands near the average, so a ten-pick card usually shows two or three hits. Third, the full match falls away fast. Marking a fourth number takes it from 1 in 82 to 1 in 435; the tenth takes it to 1 in 847,660,528. Our keno calculator gives the same probabilities for any single case.

None of this depends on which numbers you mark. Each of the 40 has the same 1 in 4 chance of being drawn in a round, rounds are independent, and no pattern of picks changes the table.

What a paytable returns: Stake's tables, worked through

Hit odds are fixed by the board. What the game returns is decided by the paytable, which assigns a multiplier to every number of hits. The return of a table is each multiplier times its chance, added up. Stake states a single figure for the game: its Keno page carries an RTP badge of 99.00% and describes "99.00% and a low house edge of 1.00%" (1 Nov 2026). The payout tables its game used on the same date make it possible to check that figure cell by cell. Take three numbers on Classic risk, which paid 1x for one hit, 3.1x for two and 10.4x for three:

Arithmetic. Payouts: Stake Keno, Classic risk, three picks, as supplied to the game on 1 Nov 2026.

Repeating that sum for every combination of risk level and pick count gives 40 returns. They cluster around the stated figure without all reaching it.

Arithmetic: return of each Stake Keno paytable (rows: risk level; columns: numbers marked), from the 40 tables supplied to the game on 1 Nov 2026 and the hit odds above.

The lowest cell is 98.65% (two numbers on Medium or High risk) and the highest 99.07% (nine numbers on Low). 17 of the 40 tables come out more than 0.05 of a percentage point under 99.00%, and none is more than 0.07 above it. In money, the gap between the best and worst cell is about $4 per $1,000 wagered: an expected cost of $9.31 per $1,000 on the best table against $13.46 on the worst. The stated 99.00% is a fair description of the game as a whole and a slightly generous one for some settings. One-number bets show why exact 99% is not always possible: Low risk pays 0.7x on a miss and 1.85x on a hit, which returns 0.75 × 0.7 + 0.25 × 1.85 = 98.75%.

Three other casinos reviewed here used the same 40 tables on the same date: Shuffle (game client, 1 Nov 2026), Winna and Rainbet (game data, 1 Nov 2026). The grid above is therefore their arithmetic too. Rainbet adds an Extreme risk level played on a larger board, which this table does not cover.

What the risk level actually changes

Since every table returns about the same, the risk level is a choice about how that return is delivered: often and in small amounts, or rarely and in large ones. The comparison below uses Stake's tables for five and ten numbers.

Arithmetic from Stake's Keno paytables (1 Nov 2026) and the hit odds above.

With five numbers, Classic pays something in 78.3% of rounds, because it returns a quarter of the stake on a single hit, while High pays in 8.9% of rounds and nothing below three hits. With ten numbers, Low risk pays in 79.6% of rounds, mostly 1.1x to 1.3x, and High in 19.7%. The top multipliers of 1,000x on ten numbers are real entries in the table and contribute almost nothing to the return: a 1 in 847,660,528 event multiplied by 1,000 is worth about one ten-thousandth of a percent. The return on those settings comes from the lower and middle rows of the table: two to five hits on Low risk, four to seven on High.

Small pick counts on High risk are the opposite case. Three numbers on High pay only for a full match, 81.5x at 1 in 82, so the whole 98.99% return rides on one outcome and 98.8% of rounds lose the stake. That is the highest-variance way to play the game at close to the same price.

Stated keno returns at the casinos reviewed here

Keno has no industry-standard price. The figure each operator publishes for its own game, and what its tables return where they are public, are set out below.

The arithmetic column uses each game's own paytables and the hit odds for 10 of 40. "Not computed here": the paytable is shown in the game and no return is worked out on this page.

Two rows deserve a second look. BC.Game's game data lists 98 for Keno, its help article says "In Keno Single, the house edge is 1%", and the 40 tables its game used on 1 Nov 2026 return between 97.66% and 98.07%: the tables fit the 98 and not the 1%. Duel is the cheapest keno on the list by its own numbers. Each of its 40 tables returns 99.90%, a 0.1% base edge, which its Zero Edge rakeback then refunds inside a daily allowance; the mechanics and limits of that allowance are in our Duel review. Duel is a commercial partner of this site.

The spread across the table is the practical point. At 99% a $1,000 session of keno bets costs about $10 on average, at 98% about $20, and at 97% about $30, for the same board and the same hit odds. Where a casino states no figure and publishes no table, the in-game paytable is the only source, and Shuffle's own page gives the right habit: "Different Keno games can use different pay-tables, so the best habit is checking the paytable in the game panel before settling into a session." To see how the game sits against dice, crash and the table games, use the lowest house edge comparison.

Is keno rigged? What the draw commitment proves

A long run of misses is what the hit table predicts, so it is not evidence of anything. The question that can be answered with certainty is narrower: was the draw fixed before you marked your numbers? On a provably fair keno it was, and you can check it. The casino commits to a server seed by showing its hash, you supply a client seed, a counter numbers your bets, and the ten drawn numbers are computed from those three values alone. Neither your picks nor the risk level is an input.

The casinos that document the conversion do it in different ways, which matters when you verify:

• Stake turns each number from its HMAC-SHA256 stream into a square by multiplying it by the squares still available, 40 for the first hit, then 39, and so on: "The fisher-yates shuffle implementation is utilised to prevent duplicate possible hits being generated" (fairness documentation, 1 Nov 2026).
• Shuffle generates ten numbers with Math.ceil(val * (40 - index)) and then steps past positions already taken, and prints a worked example for server seed "def", client seed "abc" and nonce 1 (help centre, 13 Nov 2026).
• Duel: "Keno uses a reversed "Fisher-Yates" shuffling method to shuffle numbers from 0-39, each corresponding to a tile on the keno table", with the first ten entries forming the draw (help centre, 11 Nov 2026).
• BC.Game orders a fixed list of the 40 numbers by rotated SHA-256 strings, twice, and takes the first ten; the code is printed in its help article.
• Cloudbet describes a draw generated from its SHAKE256 number stream in the same explainer that states the 99% RTP; the explainer prints no paytable.

What this proves is specific. It proves the ten numbers were not chosen after seeing your card. It does not prove the paytable is generous: the return is a property of the multipliers, which is why the previous section matters as much as this one. A game can be perfectly verifiable and priced at 97%.

How to verify a draw

1. Before playing, copy the hashed server seed from the fairness panel, and set a client seed of your own.
2. Play. Each bet uses the next nonce; note the nonce of the round you want to check.
3. Rotate the seed pair. The casino now reveals the server seed that was in use.
4. Hash the revealed seed with SHA-256 and compare it with the hash you copied in step 1. If they differ, stop there: the commitment was not honoured.
5. Recompute the ten numbers from the server seed, client seed and nonce using the casino's documented method, and compare them with the round's draw.

Our keno verifier does steps 4 and 5 for Stake and Shuffle, whose methods are published in full, and reproduces Shuffle's own worked example. For other casinos it checks the commitment only, and the recomputation has to be done with the operator's own verifier. The general provably fair verifier covers the other Originals.

Limits that cap the top prizes

A 1,000x multiplier is only worth 1,000 times the stake if the casino will pay it. Published limits on in-house keno:

Shuffle: help centre (13 Nov 2026). Stake: help centre (13 Nov 2026). Thrill, Winna, BitStarz, BetFury and Wild.io: as documented in our reviews (11 Nov 2026). Last column: arithmetic, cap ÷ 1,000 (for Stake, profit cap ÷ 999).

Shuffle's help centre explains how its cap is applied: on most games a bet whose possible win exceeds the maximum is refused before it is placed. For keno that means the cap binds on the stake you are allowed to enter for a given pick count and risk level, not on the payout after the fact.

Related: best crypto casinos · provably fair casinos · dice strategy guide · dice verifier · mines verifier

Key takeaways

1. Hit odds are fixed by the board. Ten of 40 drawn means a quarter of your picks hit on average, and all ten of ten is 1 in 847,660,528.
2. The return is set by the paytable. Stake states 99.00%; its 40 tables return 98.65% to 99.07% by arithmetic.
3. Stated returns among the casinos reviewed here run from 97% to 99.9%. Read the figure, or the table, before playing.
4. Risk level changes the shape, not the price. High risk pays rarely and large; Low pays often and small.
5. The draw can be verified; the price cannot be changed by any pick. Check the commitment, then treat the game as entertainment with a known cost. Our responsible gambling page covers limits.

Frequently asked questions