Stake.com Provably Fair Audit: 250,000 Rounds Tested — Full Report
We independently tested 250,000 Stake.com game rounds using NIST SP 800-22 statistical tests. Result: 37/37 passed, 10.0/10 EXCELLENT. Here's the full report.
TL;DR: Stake Passes With Flying Colors
Score: 10.0 / 10 — EXCELLENT
Tests passed: 37/37
Games tested: 5 (Dice, Limbo, CoinFlip, Crash, Roulette)
Total outcomes analyzed: 250,000
Methodology: NIST SP 800-22 + Custom Gaming Tests
Seeds used: Real Stake.com seeds (independently sourced)
This is not an opinion piece. This is a mathematical audit. Every number, every test, every seed is verifiable. If you don't believe us — good. Download the data and check yourself.
Data Source & Verification
Transparency about where our test data comes from is non-negotiable. Unlike other review sites that simply claim "we tested it," we disclose exactly what seeds we used, where they came from, and how you can verify our results independently.
Seed Origin
| Parameter | Value |
|---|---|
| Server Seed Hash | fe8ff24859844caa... (full hash available in raw data) |
| Client Seed | HsWCaELP0o |
| Source | Public GitHub repository (DeepSearch_HashGames) |
| Nonce Range | 0 – 49,999 per game (250,000 total) |
| Hash Algorithm | HMAC-SHA256 |
Independence Statement
- Seeds were NOT provided by Stake.com. They were obtained from an independent, publicly accessible third-party repository.
- FairPlay Audit had no contact with Stake before or during this audit. No payment was received. No affiliate relationship existed at the time of testing.
- HMAC-SHA256 outputs were independently regenerated using our own Scorecard Tool v2.0 and verified against known Stake.com game results.
- Anyone can reproduce our results using the same seed pair and nonce range. The algorithm is deterministic — identical inputs always produce identical outputs.
Honest Limitations
We believe in stating what our data can and cannot prove:
- These seeds originate from third-party collection. While HMAC-SHA256 verification confirms they produce valid Stake.com game outcomes, we cannot independently confirm which specific player sessions they belong to.
- This audit covers one seed pair. Stake rotates server seeds. Our results apply to this specific seed chain. A manipulated casino could theoretically use fair seeds for some rounds and unfair seeds for others.
- Future audits will include our own verified test accounts as an additional data source — combining large-scale public seed analysis (statistical power) with first-party seeds (provenance proof). This mirrors how certified testing labs like GLI operate: automated mass testing plus manual spot checks.
Why does this matter? Because "trust me, I tested it" is not an audit. An audit requires verifiable data, disclosed methodology, and honest limitations. If a site doesn't tell you where their seeds came from, their "audit" is just an opinion with extra steps.
What We Tested
We independently obtained real Stake.com server seeds and client seeds from verified game rounds. Using our Scorecard Tool v2.0, we generated 250,000 outcomes across 5 games and subjected them to 37 statistical tests.
Our testing methodology mirrors how GLI (Gaming Laboratories International) and eCOGRA audit traditional casinos — but applied to provably fair crypto games. The key difference: we test the raw uniform floats from the HMAC-SHA256 byte generator, not just the game outcomes. If the source bytes are uniformly distributed, the game is mathematically fair.
Test Results by Game
Dice (50,000 rounds) — PASS 8/8
| Test | Standard | Result | p-value |
|---|---|---|---|
| Monobit Test | NIST SP 800-22 | ✅ PASS | > 0.01 |
| Runs Test | NIST SP 800-22 | ✅ PASS | > 0.01 |
| Block Frequency | NIST SP 800-22 | ✅ PASS | > 0.01 |
| Cumulative Sums | NIST SP 800-22 | ✅ PASS | > 0.01 |
| Chi-Square GoF | Standard Statistics | ✅ PASS | > 0.01 |
| Kolmogorov-Smirnov | Standard Statistics | ✅ PASS | > 0.01 |
| Serial Correlation | Standard Statistics | ✅ PASS | > 0.01 |
| Runs Up/Down | Standard Statistics | ✅ PASS | > 0.01 |
The raw floats generated by Stake's Dice game show no detectable bias, no serial correlation, and pass all four NIST randomness tests at 99% confidence.
Limbo (50,000 rounds) — PASS 8/8
Identical test battery, identical results. All 8 tests passed with p-values well above the 0.01 significance threshold. The HMAC-SHA256 output for Limbo shows true uniform distribution.
CoinFlip (50,000 rounds) — PASS 7/7
Note: Runs Up/Down test excluded for CoinFlip — this test is not meaningful for binary outcome data. All applicable tests passed. Coin fairness ratio: 50.03% / 49.97% — statistically indistinguishable from a fair coin.
Crash (50,000 rounds) — PASS 7/7
Crash games use a different HMAC call structure (no :round suffix). After normalizing for this, all applicable tests pass. The instant crash rate (1.00x multiplier) was measured at ~3.9%, consistent with a 4% house edge on Crash.
Roulette (50,000 rounds) — PASS 7/7
Chi-Square Goodness of Fit across all 37 numbers (0–36) shows uniform distribution. No number was hit significantly more or less than expected. Roulette-specific distribution test: PASS.
What This Means
After analyzing 250,000 game outcomes across 5 different games, here's what the data tells us:
- The RNG is fair. All NIST SP 800-22 tests pass at 99% confidence. The raw bytes from HMAC-SHA256 are uniformly distributed.
- No detectable manipulation. Serial correlation tests show no dependency between consecutive outcomes. The house isn't "adjusting" results.
- House edge is as advertised. Crash's ~3.9% instant crash rate matches the stated 4% house edge. Dice and Limbo outcomes match their theoretical distributions.
- The system is verifiable. Every outcome can be independently reproduced using the published seeds and HMAC-SHA256.
What This Does NOT Mean
Let's be honest about the limits of this audit:
- This does not guarantee future fairness. We tested a specific seed pair. Stake could theoretically change behavior with different seeds.
- This does not cover all games. Stake offers hundreds of games. We tested the 5 core provably fair games.
- This does not mean you'll win. The house edge is real. Over time, the casino will profit. That's math, not manipulation.
- This does not replace personal verification. You should always verify your own rounds using Stake's built-in tools.
Our Methodology
We follow the FairPlay Audit Fairness Testing Framework, which is based on:
- NIST SP 800-22 — The U.S. National Institute of Standards and Technology's standard for testing random number generators. Used by government agencies, security firms, and certified testing labs worldwide.
- Raw Float Testing — Like GLI and eCOGRA, we test the raw uniform floats from the byte generator, not game-specific outcomes. This is the scientifically correct approach.
- Significance Level α = 0.01 — 99% confidence. A test passes only if the probability of the observed data under the null hypothesis (fairness) exceeds 1%.
- Minimum 50,000 rounds per game — Ensures statistical significance. Most "audit" sites test fewer than 1,000 rounds.
Our Scorecard Tool is open source. The code, methodology, and raw data will be published on GitHub for independent verification.
Scoring System
| Score | Rating | Meaning |
|---|---|---|
| 9.0 – 10.0 | EXCELLENT | All or nearly all tests pass. No concerns. |
| 7.0 – 8.9 | GOOD | Most tests pass. Minor anomalies, no evidence of manipulation. |
| 5.0 – 6.9 | FAIR | Some tests fail. Investigate further before trusting. |
| 3.0 – 4.9 | POOR | Multiple failures. Statistical anomalies detected. |
| 0.0 – 2.9 | FAIL | Systematic failures. Evidence of non-random behavior. |
Stake.com: 37/37 tests passed = 10.0/10 = EXCELLENT
About This Audit
| Field | Value |
|---|---|
| Auditor | FairPlay Audit (The Ex Pit Boss) |
| Date | June 2, 2026 |
| Casino | Stake.com |
| Methodology | FairPlay Audit Fairness Testing Framework v2.0 |
| Standards | NIST SP 800-22, Chi-Square GoF, KS Test |
| Sample Size | 250,000 outcomes (50,000 per game) |
| Confidence Level | 99% (α = 0.01) |
| Data Source | Public GitHub repository (independently sourced) |
| Independence | No payment from Stake. No prior contact. |
| Reproducibility | Full seeds, code, and data available |
This audit was conducted independently. FairPlay Audit received no compensation from Stake.com. We have no affiliate relationship with Stake at the time of this audit. Our goal is mathematical truth, not marketing.
The Ex Pit Boss has 23 years of casino industry experience, including work in European casinos and international gaming operations.
See also: BC.Game Audit (50,000 rounds, 10.0/10) | Roobet Audit (50,000 rounds, 10.0/10) | Our Independence Policy