We audited the house edge of 156 crypto casino originals

We read the payout tables of 156 crypto casino originals at six casinos and worked out the real return of each one, then set it against the RTP the casino states. Two findings stand out. First, every Plinko board we read, at all six casinos, uses the same multipliers, and Stake and Shuffle share one set of keno tables. Second, the single RTP figure casinos quote hides a real spread: the same game returns anywhere from 98.65% to 99.16% depending on the setting you pick, so the house edge on the worst setting is 60% bigger than on the best. One Plinko mode advertised at 98% cannot be checked at all.

Key findings

• 156 payout configurations recomputed across Stake, Shuffle, Thrill, Gamba, Winna and Duelbits: 71 Plinko boards, 80 keno pay tables, 3 dice settings, 1 limbo setting and 1 published crash formula.
• One set of tables. Every Plinko board we read at the six casinos had identical multipliers, and Stake and Shuffle keno use identical pay tables. Choosing between these casinos does not change your odds on these games.
• The headline RTP is an average, not a promise per bet. On these shared tables, Plinko returns 98.91% to 99.16% and keno 98.65% to 99.07%, depending on risk level, rows or picks.
• Best and worst. The best setup we found returns 99.16% (Plinko, high risk, 11 rows). The worst returns 98.65% (keno, 2 picks, medium or high risk). On $10,000 wagered that is about $84 expected cost against $135.
• One mode cannot be verified. Winna advertises 98% RTP for Extreme Plinko but says the mode uses "weighted bucket selection" with probabilities it does not publish. The pin board on screen does not show the real odds, so no player can check the 98%.

Why the stated RTP is not the full story

Casinos usually quote one RTP, such as 99%, for every setting of a game. But a Plinko board or keno pay table is a list of rounded multipliers, and each combination of risk level, rows or picks has its own exact return. Rounding a multiplier to one or two decimals moves the return a little up or down.

Where a casino states one RTP for the whole game, we treat every pay table as making that claim. That is how a player reads it.

Plinko results

All figures are the exact expected return of the multipliers shown in each game on 7 November 2026.

Winna publishes its own return for every low, medium and high board, and its figures match ours to six decimals. That is a useful cross-check: the tables we read from the screen are the tables the game uses.

Thrill's left and right side bias settings move the big multipliers to one edge of the board. We checked high risk with a left bias: the return stays at 98.98%, because every multiplier on the favoured side is doubled and the other side pays nothing.

Keno results

Crypto keno draws 10 numbers from 40. The chance of each hit count follows the hypergeometric formula, so the return of a pay table can be computed exactly.

Neither Stake nor Shuffle states a keno RTP on the game page, so the table is the only way to know what a ticket costs. The gap between settings is small in money terms, $3.46 more expected cost per $1,000 on a 2-pick medium ticket than on a 1-pick classic one, but it is invisible unless you do the maths.

Dice, limbo and crash

Dice and limbo returns can be read straight from the screen: win chance times payout.

Stake publishes its crash and limbo formula: the result is 0.99 divided by a random number, rounded down to two decimals. That makes the chance of reaching any target m exactly 0.99 divided by m, so the return is 99% at every cash-out point.

We did not include Mines. Most casinos only show the Mines multiplier during a live round, and we did not place bets for this study.

What we could not check

• Winna Extreme Plinko. Advertised at 98%. The 16-row board shows two buckets of 2,000x, two of 100x and thirteen skulls that pay nothing. If the ball fell as the pins suggest, the return would be about 11%. Winna says the mode does not work that way and uses configured bucket weights instead, but it does not publish them. The 98% may well be true. There is no way for a player to confirm it.
• Duelbits keno and most Duelbits Plinko settings. Without an account the Plinko risk and rows controls are locked, and the keno page did not open, so we scored only the default board and dice.
• Roobet and Gamdom. Roobet's Plinko did not load from our test location and Gamdom's pages could not be opened with our tools, so neither is scored.
• Stake, Shuffle, Thrill, Gamba and Duelbits Plinko and keno. We found no RTP figure on these game pages. Players rely on help articles, blog posts or third parties for the number.

How we did it

1. We opened each game in a browser, without placing a bet or logging in, and read every multiplier shown for each risk level and row or pick setting. Where the page exposed the numbers as text we copied them directly; where the game is drawn on a canvas we read them from enlarged screenshots.
2. We recorded the source page and the date (7 November 2026) for every table and every stated figure.
3. We computed the exact return in Python with fractions, not decimals: for Plinko the sum of C(n, k) divided by 2 to the power n times each multiplier, for keno the hypergeometric chance of each hit count on a 10 from 40 draw times its multiplier, and for dice and limbo the displayed win chance times the payout.
4. We compared each result with the RTP or house edge the casino states on the game page or its own blog. A gap of more than 0.05 points is flagged.

We did not estimate any table we did not see. Where a value was missing we left it out rather than fill it in.

Check a casino yourself

You do not need our spreadsheet to test a game. Two checks take a few minutes:

1. Price the pay table. For Plinko, multiply each multiplier by its chance (our guide to Stake Plinko odds lists the chance of every bucket). For keno, use the hit odds table in which keno game has the best odds. Add the results. That is the real RTP for that setting.
2. Check the outcome. Rotate your seed pair after a session, then paste the revealed server seed, your client seed and the nonce into our provably fair verifier. It recomputes dice, limbo, Plinko, keno and Mines results for Stake-style games and Thrill, and checks that the server seed matches the hash you were shown before you played.

The first check tells you what a game costs; the second, that the result was fixed before you bet.

What this means for players

A single session is decided by luck, not by a few tenths of a point. Over thousands of bets the edge decides. Because these casinos share the same tables, switching casino will not lower the cost of Plinko or keno. Switching settings will: on keno, a 1-pick classic ticket or a 9-pick low ticket returns more than a 2-pick medium one. If you play for volume, check the two or three tables you actually use.

Gamble only what you can afford to lose. The house edge means that over time the casino wins.

Data

The full dataset lists 161 rows: casino, game, setting, stated RTP and where we read it, computed RTP, the gap in points, our flag and the page each table came from. Ask us for the CSV at partners@provablylegit.com. You may republish the figures with a link to this page.