What is a casino house edge? The arithmetic, and the edges casinos and studios publish
What is a casino house edge?
A casino house edge is the share of every bet a game is built to keep over time, and it equals 100% minus the RTP. Stake publishes 1% on Dice, Plinko and Limbo and 2% on Drill. Across 97 video slots on studio game sheets collected by CryptoGuardz, the median edge is 4.7%, from 3% to 6.5%.
Piece published Reviewed by CryptoGuardz Register Desk, editorial desk
A casino house edge is the share of every bet a game is built to keep over time. It is 100% minus the game's return to player (RTP). Stake's own Dice page, read on 9 October 2026, states a 99% RTP "with a low house edge of 1%". Across 97 video slots on studio game sheets collected by CryptoGuardz, the median edge is 4.7%, nearly five times that.
The edge is not a fee charged per spin, and it is not a prediction for one session. It is the long-run average built into the paytable. The arithmetic below shows how it is calculated and what it costs at real stake sizes.
What is a casino house edge?
The house edge is the casino's expected profit per unit wagered, expressed as a percentage. If a game has a 4% edge, every US$100 put through it is expected to return US$96 on average, and the casino keeps US$4. One player can win, lose everything or break even. Over enough bets the average moves toward the published figure.
The UK Gambling Commission treats the edge and the RTP as two ways of publishing the same fact. Its remote technical standard RTS 3C requires an operator to show, for each game, information that includes:
house edge (or margin)
The next item on the same list is "the return to player (RTP) percentage", and the third is "the probability (likelihood) of winning events occurring". The standard says this information "must be easily available before the customer commits to gamble." A crypto casino licensed in Curacao or Anjouan is not bound by RTS 3C, but the definition is the cleanest one a regulator publishes.
How do you calculate the house edge?
Take the probability of each outcome, multiply it by what that outcome pays, add the results, and subtract from 1. Three worked examples follow.
Dice at Stake. Stake's fairness code produces a roll from 0.00 to 100.00. A bet that wins 49.5% of the time and pays 2x returns 0.495 x 2 = 0.99 per unit staked. The edge is 1 minus 0.99, or 1%. That matches the game page.
Single-zero roulette. Stake's published roulette conversion is `POCKETS[Math.floor(float * 37)]`, which means 37 pockets. A straight-up bet on one number at the European standard of 35 to 1 wins 1 time in 37 and loses 36 times in 37. The expected result per unit is (1 x 35 minus 36 x 1) divided by 37, which is minus 1/37, or minus 2.70%. Stake's paytable for roulette was not read for this piece, so the 35 to 1 figure is the convention, not a quote: not verified.
Limbo. Here the edge is written into the formula. Stake's game events page gives:
// Game event translation with houseEdge of 0.99 (1%) const floatPoint = 1e8 / (float * 1e8) * houseEdge;
Every multiplier the game generates is scaled by 0.99 before it is shown. Shuffle's Limbo code does the same with `const rtp = 0.99;` and a comment that reads "generate limbo game event results with house edge (1%)". The two formulas differ by a plus one in the denominator, so results are not interchangeable between the two sites, but the edge is the same.
The site's house edge calculator runs the same arithmetic for any paytable.
What is the house edge on each type of casino game?
It depends on who sets it. On in-house crypto games the operator sets it and some publish it in code. On third-party slots the studio sets it on its game sheet, sometimes with several versions. The table combines the operators' own published figures with the RTP distribution from studio game sheets collected by CryptoGuardz on 5 October 2026.
| Game or group | Who publishes | RTP | House edge |
|---|---|---|---|
| Stake Dice | Stake game page | 99% | 1% |
| Stake Plinko | Stake game page | 99% | 1% |
| Stake Limbo | Stake fairness code | 99% | 1% |
| Shuffle Limbo | Shuffle fairness code | 99% | 1% |
| Stake Drill | Stake fairness code | 98% | 2% |
| Single-zero roulette, straight bet | Arithmetic on 37 pockets | 97.30% | 2.70% |
| 97 video slots, median | Studio game sheets | 95.3% | 4.7% |
| 97 video slots, range | Studio game sheets | 93.5% to 97% | 3% to 6.5% |
| 133 crash and instant titles, median | Studio game sheets | 96% | 4% |
| All 231 titles with an RTP, range | Studio game sheets | 93% to 99% | 1% to 7% |
The 97 slots run from Game of Thrones Jackpot King by Blueprint at 93.5% to Lady Wolf Moon by BGaming at 97%. Familiar titles sit inside that band: Starburst at 94.05%, Wolf Gold at 93.99%, Sweet Bonanza at 95.45% and Book of Dead at 96%. Seven titles in the dataset carry no RTP on their sheet and are excluded: not disclosed.
The wider catalogue tells the same story. The providers dataset covers 9,538 titles with a published RTP across 75 studios. The median of the studios' own median RTPs is 96%, an edge of 4%. Live dealer studios sit lower: Pragmatic Play Live's median is 98.76% across 141 titles and Playtech Live's 98.85% across the 42 of its 149 titles that disclose a figure. Each studio's spread is on its provider page.
Why are house edge and RTP the same number?
Because they add up to 100%. RTP is the share a game is designed to pay back; the edge is the share it keeps. The Commission's own glossary defines theoretical RTP as "the designed return to player percentage of the game", which "will also be the advertised RTP of the game as displayed in the player facing rules, as per RTS 3C." A 96% RTP and a 4% edge are one fact written twice.
The same glossary separates theory from what happened. Actual RTP is "Calculated using the generated win and turnover figures of the live (operational) game", and gross gambling yield is "turnover minus win". The edge is the theoretical gap. The yield is the realised one. Over millions of spins the two converge, which is why a casino can budget on the edge while any single player's result is anywhere.
The full explanation of RTP itself, with the studio-by-studio distribution, is in what RTP means in slots.
How much does the house edge cost in money?
The edge applies to everything wagered, not to the deposit. That is the part most players underestimate. Every win that is bet again is wagered again, and the edge is taken again.
| Total wagered | At 1% (Stake Dice) | At 2.70% (single-zero roulette) | At 4.7% (slot median) | At 6.5% (lowest slot RTP) |
|---|---|---|---|---|
| US$1,000 | US$10 | US$27 | US$47 | US$65 |
| US$10,000 | US$100 | US$270 | US$470 | US$650 |
| US$100,000 | US$1,000 | US$2,700 | US$4,700 | US$6,500 |
These are expected losses, the average across very many players, computed from the RTPs above. A player who deposits US$1,000 and spins it through a 4.7% slot until the balance has been wagered 30 times over has put US$30,000 through the game and carries an expected loss of US$1,410, more than the deposit. In practice that player runs out before reaching the thirtieth cycle. Wagering requirements on bonuses work on the same arithmetic; the wagering calculator shows it for a given offer.
Does provably fair change the house edge?
No. Provably fair hashing proves that a result was fixed before the bet and not altered afterwards. It says nothing about the paytable. A provably fair game with a 3% edge is a fair 3% game. What the Stake and Shuffle code adds is that the edge itself is in the published formula, so anyone can check that the multiplier was scaled by 0.99 and not by something lower.
Of the 14 operators in the register's provably fair dataset, collected on 22 September 2026, five document their algorithm well enough for a third party to reproduce a result: Stake, Shuffle, Sportsbet.io, BC.GAME and Thunderpick. Thunderpick is a commercial partner of this site, disclosed on the disclosure page. The dataset records the documentation, not the edge; for most operators the edge per original game was not collected: not measured. The details are on the provably fair page.
Can the same slot have a different house edge at different casinos?
Yes. Some studios ship one game in several RTP versions and let the operator choose. The games dataset contains 10 pairs where the studio publishes a second, lower version as a separate title, for example Skyliner at 96% and Skyliner 94 at 94%, both from Gaming Corps. On a US$10,000 total stake the difference is US$200 of expected loss. Which version a given casino runs is shown on the game's own information screen, and that screen is what binds. The studio-level data is on the games page.
What this does not show
The actual RTP any casino achieved on any game is not measured. Whether a casino runs the studio's headline RTP or a lower version is not measured per operator. The roulette figure assumes the European 35 to 1 payout. Operator edges on table games and on most crypto originals were not collected: not disclosed in the register. Rakeback and cashback, which hand part of the edge back, are left out of every figure above.
Quick answers
What is a casino house edge?
The share of every bet a game keeps on average, equal to 100% minus the RTP. Stake Dice publishes 1%, and the median video slot on studio game sheets collected by CryptoGuardz carries 4.7%. See the games page.
How do you calculate house edge?
Multiply each outcome's probability by its payout, add them up and subtract from 1. A 49.5% chance at 2x gives 0.99, a 1% edge. The house edge calculator does it for any paytable.
Is house edge the same as RTP?
They are the same fact from two sides: a 96% RTP is a 4% edge. The UK Gambling Commission's RTS 3C lists both as information a player must see before betting.
What is the house edge on slots?
Across 97 video slots on studio game sheets, from 3% to 6.5%, with a median of 4.7%. Each studio's spread is on its provider page.
Where can I get help with gambling?
Gambling Therapy offers free international support at gamblingtherapy.org, and more options are on the responsible gambling page.
Sources
- Stake Dice game page, retrieved 9 October 2026
- Stake Plinko game page, retrieved 9 October 2026
- Stake provably fair game events, retrieved 9 October 2026
- Shuffle provably fair game events, retrieved 9 October 2026
- UK Gambling Commission, RTS 3 rules, game descriptions and the likelihood of winning, retrieved 9 October 2026
- UK Gambling Commission, key terms relating to live return to player performance monitoring, retrieved 9 October 2026
- CryptoGuardz games dataset, studio game sheets collected 5 October 2026
- CryptoGuardz providers dataset, studio game sheets collected 5 October 2026
- CryptoGuardz provably fair dataset, collected 22 September 2026
About this piece
Compiled by the CryptoGuardz register team from the sources listed above. Research and drafting are AI-assisted and reviewed by a named editor before publication. Nothing here is a recommendation to gamble. If a figure or a reading is wrong, tell us and the correction is logged in public.