We measure realized RTP from our own crypto-casino bet feed and compare it against the provider’s declared RTP and each casino’s configured RTP build. Sample sizes, wagered volume and ±1σ bands on every number.
We computed 9,831 casino × game pairs from 1,150,120 observed bets ($77.8M wagered, feed window 2026-08-11 → 2026-08-18). Only pairs with at least 1,000 bets and $10.0K wagered are published; the other 9,767 cells are suppressed — below the threshold a realized RTP is variance, not signal. 23 pairs are triple-verified: declared, configured and realized all known.
Largest published cells by wagered volume (all-time window):
| Game | Casino | Declared | Configured | Realized ±1σ | Bets | Wagered |
|---|---|---|---|---|---|---|
| keno | Shuffle | — | — | 101.39% ±3.62 | 27,850 | $6.3M |
| blackjack | Shuffle | — | — | 94.42% ±1.09 | 8,495 | $6.0M |
| mines | Shuffle | — | — | 97.09% ±5.62 | 5,381 | $3.5M |
| blackjack | Rainbet | — | — | 158.13% ±0.42 | 46,844 | $3.2M |
| Duck Hunters | Shuffle | 96.05% | 96.05% | 96.68% ±11.08 | 3,001 | $2.5M |
| plinko | Shuffle | — | — | 104.81% ±10.83 | 9,680 | $1.5M |
| dice | Shuffle | — | — | 98.74% ±8.15 | 12,848 | $1.3M |
| Pragmatic Limbo+ | Shuffle | — | — | 97.97% ±6.54 | 8,816 | $1.1M |
| keno | Rainbet | — | — | 283.67% ±5.08 | 132,816 | $684.3K |
| Le Fisherman | Rainbet | 96.33% | — | 110.81% ±13.83 | 8,252 | $482.0K |
| coinflip | Shuffle | — | — | 103.14% ±7.80 | 1,508 | $475.6K |
| Sugar Rush Super Scatter | Shuffle | 96.58% | — | 65.62% ±15.59 | 1,370 | $405.3K |
| Gates of Olympus Super Scatter | Shuffle | 96.50% | 96.50% | 82.00% ±12.53 | 1,945 | $375.2K |
| dice | Rainbet | — | — | 135.51% ±26.84 | 36,936 | $333.0K |
| Wanted Dead or a Wild | Shuffle | 96.38% | 96.38% | 101.06% ±58.22 | 1,547 | $294.2K |
Volume-weighted across each casino's published cells. The matched columns compare like for like: only cells where the configured RTP build is known.
| Casino | Games | Bets | Wagered | Realized (all) | Configured (matched, weighted) | Realized (matched) |
|---|---|---|---|---|---|---|
| Shuffle | 15 | 88,000 | $24.6M | 97.46% | 96.13% (3 games) | 95.35% |
| Rainbet | 45 | 558,487 | $6.9M | 174.16% | 96.38% (20 games) | 192.55% |
| Gamdom | 4 | 12,479 | $281.4K | 83.68% | — | — |
Every slot has a declared RTP — the figure in the provider's specification sheet. Most providers also ship the same game in several RTP builds (96%, 94%, 92% and lower), and each casino quietly picks one: the configured RTP, which is often below the headline. Neither number tells you what players actually received. The realized RTP on this page is computed from wagerdb's own bet feed — 1,150,120 observed bets, $77.8M in volume — as total payouts divided by total wagers per game per casino.
The comparison is only meaningful with statistical discipline. Every published cell shows its sample size and wagered volume, cells under 1,000 bets or $10.0K wagered are suppressed outright, and each realized figure carries a ±1σ standard-error band derived from the observed multiplier dispersion. Slot payouts are heavy-tailed: over a short window the realized value routinely sits several points above or below the configured value, and that is expected behaviour, not an anomaly.
Rows marked triple-verified have all three numbers above the evidence threshold — the fullest public picture of a slot's economics available anywhere. For the build-level view (which casino runs which RTP setting of the same game), see the RTP Builds Report.
Declared RTP is the headline return-to-player figure in the provider's game specification. Configured RTP is the specific build the casino actually runs — providers ship the same slot in several RTP configurations and each casino picks one. Realized RTP is neither: it is what players actually received back in the bets wagerdb's own feed captured, computed as total payouts divided by total wagers. The first two are promises and settings; the third is a measurement.
Three reasons, in order of how often they apply. First, variance: RTP is a long-run average over millions of rounds, so over a short observation window the realized figure swings — sometimes far — around the configured value. That is normal and expected, which is why every number here carries a ±1σ band and a sample size. Second, bet-size mix: realized RTP is volume-weighted, so a few large bets landing big multipliers move it more than thousands of small ones. Third, feed coverage: if our feed captures a non-representative slice of a casino's bets (for example a public feed that over-represents wins), the realized figure can sit persistently above or below the true value. A realized RTP far above 100% across a very large sample is a coverage artifact, not evidence that a casino pays out more than it takes in.
We suppress every cell with fewer than 1,000 observed bets or less than $10,000 wagered. Below that threshold a realized RTP number is mostly variance, and publishing it would be publishing noise dressed up as signal. The suppressed count is shown above the table — most pairs in the feed do not meet the bar.
It is one standard error of the mean, estimated from the observed dispersion of per-bet multipliers: se = stdev(multiplier) / √n. Roughly speaking, if the game were replayed under identical conditions, the realized RTP would land inside the ±1σ band about 68% of the time. Slots are heavy-tailed — rare huge multipliers dominate the variance — so the band can be wide even with thousands of bets.
It marks game–casino pairs where all three independent numbers exist and the realized cell passed the evidence threshold: the provider's declared RTP, the casino's configured RTP build, and our realized measurement from the feed. It is the fullest public picture of a slot's economics that exists anywhere.