RAIN RNG · PRE-SUBMISSION PACKAGE · v1.2 · 2026-09-14 09:34 UTC
RAIN RNG — Pre-Submission Package for accredited test labs
Everything a GLI / BMM / iTech / eCOGRA reviewer asks for under GLI CSR §2.2 — RNG description, source, collection tools, statistical evidence with raw logs, reproducible build, process documents — published as-is, with the gaps stated. RNG component only (@rain/rng-core, @rain/rng-session, @rain/rng-node, rng-py, lab/). Not in this package: the casino contracts, games and Outperform markets.
NOT CERTIFIEDSELF-RUN EVIDENCEJOBS STILL RUNNING — COUNTS ARE LIVE, NOT ROUNDEDREPRODUCIBLE No laboratory has evaluated this package yet. Every number below was produced by us, on our hardware, with the tools in the zip; a lab will re-run them. Numbers marked in progress are read from the running jobs' checkpoints at build time.
1STATUS — PROVEN · IN PROGRESS · PENDING EXTERNAL
From LAB-SUBMISSION-REQUIREMENTS §7 (20 artifacts: HAVE 13 · PARTIAL 5 · MISSING 2), REVIEW-RESPONSE-R1 and
lab/results/PROGRESS.md. "Proven" means: in the repository, reproducible with the shipped tools.✅PROVEN — IN REPO, REPRODUCIBLE
- RNG description in GLI-19 §3 vocabulary, §3.3.2 a/b/c explicit — RNG-DESCRIPTION
- Source complete, readable, git history; vendored engines un-minified
- Reproducible build: 214 dist files fingerprinted,
lab:verify-buildpasses ×2 — BUILD - Collection tools (final-outcome + raw, production code path, seeded, checkpointed, adversarial modes)
- Dieharder chacha20 90 P / 5 W / 0 F on 61/80 combos (stream, never rewinds)
- NIST STS 1000 × 10⁶ bits: 188/188 both mechanisms, 0 uniformity fails
- TestU01 SmallCrush 15/15 both mechanisms
- Scaled outcomes chacha20: 10⁸ per game (dice, roulette, coin, reel), 2 × 10⁶ decks, 3 × 10⁸ adversarial — 8.04 × 10⁸, all total-χ² pass at α = .01
- Scaling proof (intBelow exact; legacy % N bias < N/2²⁵⁶) — SCALING-PROOF
- §3.3.2(c) one-pager — STATE-MODIFICATION
- Runtime KATs (11 vectors, NIST CAVP HMAC_DRBG + RFC 8439 + RAIN stream) before first use / 24 h /
/selftest; output inhibited on failure - Real-network latency measured (Virginia → Oregon), 4,000/4,000 rounds, 0 errors — REPORT
- Process docs: key management, incident response, change control, SDLC, hosting/HA/DR, security program; CI + CODEOWNERS + governance in repo
⏳IN PROGRESS — RUNNING, COUNTS LIVE
- Dieharder hmac-drbg: 52/80 combos done — 83 PASSED / 3 WEAK / 0 FAILED. Remaining:
rgb_lagged_sumntuples (28 combos, ≈ 0.44·(n+1) GB each ≈ 250 GB total). Why slow: HMAC_DRBG through the full ceremony path yields ≈ 1 MB/s per core; a lab host finishes it with the same resumable runner. - Dieharder chacha20:
rgb_lagged_sum14–32 (19 combos, ≈ 200 GB) deliberately not run — budget; runner resumable. - hmac-drbg 10⁸-per-game outcomes (
results/outcomes/hmac-1e8/): 270,000,000 scaled outcomes reached so far —The completed hmac sets (2 × 10⁷ per game, 2.0 × 10⁸ total incl. legacy comparisons) are already analysed in the report.game outcomes so far status dice 60,000,000 in progress — 600,000 / 1,000,000 rounds (60 %) roulette 70,000,000 in progress — 700,000 / 1,000,000 rounds (70 %) coin 70,000,000 in progress — 700,000 / 1,000,000 rounds (70 %) slot-reel 70,000,000 in progress — 700,000 / 1,000,000 rounds (70 %) - TestU01 Crush / BigCrush: not run (137 GB / 1.1 TB of words; 10–70 h at our producer rate). Stream runner shipped; documented as not run, not as passed.
- Contract bytecode fingerprint: no
solcin the lab environment;RngAnchor.solundeployed → anchoring path not "final".
⬜PENDING EXTERNAL — NOT IN OUR HANDS
- Branch protection + second maintainer. Settings written out in GOVERNANCE; must be applied by a GitHub org owner; repo is single-maintainer today. CI file sits at
.github/workflows-pending/on the mirror (PAT lacksworkflowscope). - Penetration test — not contracted. Scope and annual cadence fixed in SECURITY-PROGRAM §2.
- External contract + node audit — not contracted; cadence per deployment / per major version.
- Laboratory evaluation — this package is what we would hand over at the pre-submission call; no lab engaged yet.
- Docker HA run (2 workers + Postgres + LB, SIGKILL mid-run):
ha-test.shshipped, not executed — no Docker/root here; Render's LB gives no kill-one-instance handle. - npm publish decision — packages pack cleanly (
npm pack --dry-run), names free; publishing is the owner's call — PUBLISHING.
2HEADLINE EVIDENCE
Two mechanisms are evaluated separately (GLI-19 §4.5.1): hmac-drbg (SP 800-90A HMAC_DRBG/SHA-256) and chacha20 (RFC 8439), both keyed per round from
r_k = keccak256(pRev_k ‖ hRev_k ‖ sessionSeed ‖ sessionId ‖ k). Full tables with p-values: REPORT-STATISTICAL.| SUITE | CHACHA20 | HMAC-DRBG | NOTE |
|---|---|---|---|
Dieharder 3.31.1, full -a battery, stream input (-g 200, never rewinds) | 90 PASSED / 5 WEAK / 0 FAILED 61/80 combos · ≈ 73 GB consumed | 83 PASSED / 3 WEAK / 0 FAILED 52/80 combos · IN PROGRESS | Not run: rgb_lagged_sum tail (data volume). A 4 GB file-mode run was found to rewind 19× and is quarantined, not evidence (§9 of the report). |
| NIST SP 800-22 STS 2.1.2 — 1000 sequences × 10⁶ bits, all 15 tests, default parameters | 188/188 result lines in CI · 0 uniformity P < 10⁻⁴ | 188/188 · 0 uniformity fails | One private STS tree per mechanism; first chacha run discarded (shared experiments dir) and re-run. |
| TestU01 1.2.3 SmallCrush | 15/15, p ∈ [0.06, 0.96] | 15/15, p ∈ [0.09, 0.92] | Crush / BigCrush not run (137 GB / 1.1 TB). Runner shipped. |
Scaled outcomes after intBelow/shuffle — χ² total, blocked-χ² KS (100 blocks), serial lags 1–8, runs, coupon, reel interplay | 10⁸ per game dice · roulette · coin · reel stops; 2 × 10⁶ decks (1.04 × 10⁸ placements); 3 × 10⁸ adversarial dice → 8.04 × 10⁸ | 2 × 10⁷ per game (2.0 × 10⁸ total incl. legacy % N comparisons); 10⁸ sets in progress: 270,000,000 so far | Every total-distribution χ² passes at α = .01 for every game, both mechanisms, all adversarial modes. Isolated single-statistic exceedances (e.g. lag-4 |z| 3.07/3.29 in two chacha adversarial sets) at the rate ≈ 15 statistics × 1 % predicts; not shared across sets. |
| Fisher–Yates shuffle | 52×52 card×position table χ² (2601 d.f.): chacha p = 0.1000, hmac p = 0.2708 · per-position KS · adjacency · first-card — all pass | Proof of 1/52! in SCALING-PROOF. | |
| Adversarial player input (GLI-19 §4.5.2(e) defence) | 3 modes — fixed, zero, grind — 10⁸ dice each (chacha) + 2 × 10⁷ each (hmac): statistically indistinguishable from the honest set on every aggregate statistic | The player can add entropy but cannot bias: hRev_k is committed and unknown. | |
| Reproducible build | 214 files SHA-256 in FINGERPRINTS.json; npm run lab:verify-build passes twice on the reference machine; CI re-checks drift | No solc here → contract bytecode not fingerprinted. | |
| Runtime KATs / self-test | 11 known-answer vectors (NIST CAVP HMAC_DRBG ×4, FIPS/RFC, RAIN stream vectors both mechanisms) before first reveal, every 24 h, GET /selftest; RNG endpoints 503 in error state; fingerprint check vs manifest | SP 800-90A §11.3; GLI-19 §2.3.2. | |
REAL-NETWORK LATENCY — UNMODIFIED @rain/rng-node ON RENDER (GCP us-west1, OREGON, BEHIND CLOUDFLARE) · CLIENT AWS us-east-1 (VIRGINIA) · 1,000 SPINS PER RUN · 0 ERRORS
| ANIMATION GAP | TRANSPORT | PIPELINING OFF — p50 / p95 (ms) | PIPELINING ON — p50 / p95 (ms) | READING |
|---|---|---|---|---|
| 0 ms (no animation — pathological) | HTTPS keep-alive | 87.8 / 102.0 | 86.7 / 100.1 | Honest boundary: with nothing to hide behind, ON ≈ OFF. Pipelining overlaps the round trip; it cannot beat physics. |
| WebSocket | 76.9 / 100.1 | 72.4 / 83.6 | ||
| 60 ms (headline — harsher than any real slot) | HTTPS keep-alive | 88.5 / 101.8 | 36.0 / 49.0 | Critical path ≈ max(0, RTT − animation): ≈ 90 ms RTT − 60 ms gap → ≈ 30 ms visible. p99 119.3 → 67.5 (HTTP), 101.5 → 57.9 (WS). |
| WebSocket | 91.1 / 100.2 | 23.3 / 39.7 | ||
| 1,000 ms (a real reel; 300 spins) | HTTPS keep-alive | 87.7 / 103.9 | 1.2 / 1.6 | RTT ≪ animation → the network leaves the critical path; what remains is local settle + DRBG (single-digit ms). The reveal is already there. |
| WebSocket | 71.2 / 85.1 | 1.0 / 1.5 |
Application RTT baseline (GET /healthz keep-alive) p50 ≈ 101–108 ms on a transcontinental path — a pessimistic placement; an operator co-locating within one region would see ≈ 10–30 ms. Not shown: concurrent-player throughput, Postgres-backed persistence, browser → operator → node chains. Raw: results.json, results-gap0.json, results-gap1000.json, loadtest-*.txt, network-baseline.txt, deploy-manifest.txt — in the zip. Full write-up: REPORT.
3PACKAGE INDEX — MAPPED TO GLI COMPOSITE SUBMISSION REQUIREMENTS v2.0 §2.2
Every document is rendered here with the site chrome and shipped verbatim (markdown) in the zip and on the mirror branch. The markdown is the document of record.
| GLI CSR §2.2 ITEM | DOCUMENT(S) | STATUS |
|---|---|---|
| Package index | lab/README · EVIDENCE-INDEX | HAVE |
| RNG Source Code — final, complete, compilable, commented, edit history, binaries | packages/rng-core, rng-session, rng-node, rng-py on the mirror branch ↗ · CHANGELOG | HAVE (RngAnchor.sol undeployed) |
| RNG Final Outcome Collection Tool + Raw Output Collection Tool | lab/tools/collect-outcomes.mjs, collect-raw.mjs, analyze-outcomes.mjs, suite runners — in the zip; described in lab/README | HAVE |
| RNG Description and Documentation (GLI-19 §3 order; §3.3.2 a/b/c) | RNG-DESCRIPTION · docs/RNG · STATE-MODIFICATION | HAVE |
| Technical Source Code Description ("instantiation to final outcome") + every reduction | SCALING-INVENTORY · SCALING-PROOF | HAVE |
| Statistical data analysis with raw outputs | REPORT-STATISTICAL · EVIDENCE-INDEX · raw logs in the zip (Dieharder per-test logs, STS finalAnalysisReport + experiments, SmallCrush logs, outcome analyses); raw .bin/.csv listed below | HAVE, self-run; hmac Dieharder + 10⁸ sets in progress |
| Software verification (lab recompiles; signatures match) | BUILD · FINGERPRINTS.json | HAVE |
| Runtime self-verification (GLI-19 §2.3.2; SP 800-90A §11.3) | packages/rng-node/src/selftest.ts (mirror) · RNG-DESCRIPTION §10 | HAVE |
| Platform items §2.9 / §2.11 — change control, SDLC, key management, incidents, hosting/HA/DR | process/CHANGE-CONTROL · process/SDLC · process/KEY-MANAGEMENT · process/INCIDENT-RESPONSE · process/HOSTING-HA-DR | HAVE ×3 · PARTIAL (change control, SDLC: single maintainer, CI not yet executed on GitHub) · PARTIAL (HA test not run) |
| Pen-test / external audit (GLI-19 App. B.9; §6) | process/SECURITY-PROGRAM | MISSING — third parties; not contracted, said so |
| Repository controls (GLI-19 App. B) | docs/GOVERNANCE · .github/workflows/ci.yml, CODEOWNERS (mirror) | PARTIAL — org owner must apply |
| Integration / hosting the counterparty | docs/INTEGRATION-OPERATOR · docs/HOUSE-NODE | HAVE |
| Readiness & requirements analysis | docs/LAB-READINESS · docs/LAB-SUBMISSION-REQUIREMENTS | — |
| Engineering evidence (not a §2.2 item) | Real-network latency | HAVE |
| Review record · distribution | REVIEW-RESPONSE-R1 · PUBLISHING | — |
| Game description / pay tables | out of RNG scope — game configs vendored in packages/games/vendor | PARTIAL (not this package) |
4DOWNLOADS — WITH SHA-256
Hashes computed at build time from the files served here. Also in SHA256SUMS.txt. Verify:
sha256sum -c SHA256SUMS.txt.| FILE | WHAT | SIZE | SHA-256 |
|---|---|---|---|
| rain-rng-package-v1.2.zip | lab/ — every document, tool (source), result summary, per-test Dieharder logs + .DONE markers, STS finalAnalysisReport/freq/RUN, SmallCrush logs, outcome analyses + meta, run logs, process docs, top-level docs/, CI + CODEOWNERS (raw .bin / big .csv excluded — listed below) | 1.6 MB | 30326467dffcb090da48f231a05fc9feed44c3f957b4c0cc98beb9d32e02f66f |
| rain-rng-package-v1.2-sts-experiments-chacha.zip | companion: NIST STS experiments/ tree for chacha20 (stats.txt + results.txt of all 15 tests, 1000 sequences) | 5.0 MB | dee3fff16949e5e79b79036ed9125b705d77ce1e65e41dd5b5f4c71bb7f5c93b |
| rain-rng-package-v1.2-sts-experiments-hmac.zip | companion: NIST STS experiments/ tree for hmac-drbg | 5.0 MB | 20e0071f7b55de5048226682aaf488c4fd9962dc87d8a5a15f3c13e2447236c1 |
| RAIN-RNG-Whitepaper-v1.2.pdf | Whitepaper v1.2 (PDF) — adds the real-network pipelining measurement | 2.1 MB | 00e157874d242a5fbbfb8eb07fa176a85e37e010723b16e9f163895aee7e6151 |
| RAIN-RNG-Whitepaper-v1.2.md | Whitepaper v1.2 (Markdown source) | 103 kB | 53786135d277ade66a847107c689c75b24b55ad9c5ed2f65e1bc345068aabc14 |
| FINGERPRINTS.json | reproducible-build manifest — SHA-256 of 214 dist files, toolchain pins | 25 kB | f5d6754fe298f38f7792480056da3f959e3be30f1fae43d78da95147937a2990 |
| EXCLUDED-ARTIFACTS.tsv | size + sha256 of every raw artifact not in the zip | 3 kB | f8ae48c38e23a47954e804974aef15ca1a127308403ae1df0d0841c7965f2276 |
| RAIN-RNG-Whitepaper-v1.1.pdf | history | 2.1 MB | 2cd6805b35941cbb1955e0021e98aacd2a2d2b8da075eec64c611cbd42887b77 |
| RAIN-RNG-Whitepaper-v1.0.pdf | history | 1.8 MB | b5d37bb492d08226c773c7947ac0e396eff38fcc4c25fd9d2441e7a540374075 |
NOT IN THE ZIP — RAW ARTIFACTS (25 files, NaN GB)
The raw generator output and the full outcome CSVs are too large for a static site. They are available on request (object-storage link) and — more usefully for a lab — regenerable bit-for-bit with the shipped tools from the recorded seeds (results/raw/<mech>/SEED-BIG, SEED-STREAM; per-set masterSeed in each .meta.json). Files still being written (hmac-1e8/*.csv) are hashed at their state at build time and will change.
| PATH | SIZE | SHA-256 (at build) |
|---|---|---|
path | NaN B | sha256 — raw artifacts excluded from rain-rng-package-v1.2.zip; regenerable from recorded seeds with lab/tools; available on request. Hashes taken 2026-09-14T09:27Z; hmac-1e8/*.csv were still being written. |
lab/results/outcomes/chacha/card-shuffle.csv | 309.0 MB | e3bc8929b97b4809e784c5b30b7c8b9e73d75287ceb20753786d32cb5ed053b7 |
lab/results/outcomes/chacha/coin.csv | 208.2 MB | ac1ae962777fa6377520f1eacb8a679731dbd8e591aa707bb7d2b9d6959c2b92 |
lab/results/outcomes/chacha/dice-adv-fixed.csv | 300.2 MB | 5641c43b0a801c824cd4f181e2c833685f38a4903bcd2b5b7892cff5406df532 |
lab/results/outcomes/chacha/dice-adv-grind.csv | 300.2 MB | 8820089b500552d73d52b30da689be9dd100b92d4e551e11cbaa17bbea43f287 |
lab/results/outcomes/chacha/dice-adv-zero.csv | 300.2 MB | b70f396edb29539d61ed70a74895f1c32a30563f362e8122748ddab6291c9373 |
lab/results/outcomes/chacha/dice.csv | 300.2 MB | 6d2cb8b6e016147aac4829626b70f7b249853858a2cccbba63b9c0b02460ec0e |
lab/results/outcomes/chacha/roulette.csv | 281.2 MB | 275e34e074a96ee3dd99382a9be59b56b532115b5be7d1fc2e33d5a892fc74ec |
lab/results/outcomes/chacha/slot-reel.csv | 414.4 MB | c9640dac566a5aaae6a1899417559f6fa51a61c798e782e79e0934ed860c3ed0 |
lab/results/outcomes/hmac-1e8/coin.csv IN PROGRESS | 140.5 MB | 41b6c3b25024153d56de8f965715301e1ad2bdfcb3641efaa3b4ac2dc64d1396 |
lab/results/outcomes/hmac-1e8/dice.csv IN PROGRESS | 194.1 MB | b088909456040649cd5d198bc704f2c2afa6f06d9257f42472bfbb267c85219a |
lab/results/outcomes/hmac-1e8/roulette.csv IN PROGRESS | 189.8 MB | 81b9f723198e9004b8ee9803c8b5328fb69746efc63291d42739aad3c331dce7 |
lab/results/outcomes/hmac-1e8/slot-reel.csv IN PROGRESS | 281.6 MB | 64d7f63b4e171a92ed2afdeaaf275730dce179aea9b4d152ced0c56f00cf9262 |
lab/results/outcomes/hmac/card-shuffle.csv | 61.5 MB | ab41d8dab1e9883244c87f43b78ed00fd59529553dc3e2a98b96b8bed2b07208 |
lab/results/outcomes/hmac/coin.csv | 41.6 MB | 724570642927323c03a88ee311c833ad2162e169f164c370e7604d6ea617d7fb |
lab/results/outcomes/hmac/dice-adv-fixed.csv | 60.0 MB | 72dbda59833d15b19b7abd3ea54fd90c5b864d5a5f187656544714c1df2af1e9 |
lab/results/outcomes/hmac/dice-adv-grind.csv | 60.0 MB | 438f5121fabc2558570f9d0713fd01f1e308f9eaad7029edd2b4cfded76d0a77 |
lab/results/outcomes/hmac/dice-adv-zero.csv | 60.0 MB | f30e9890bea64cbc5fdaa55f6a68b8928b1747b044e87357e67a231304587918 |
lab/results/outcomes/hmac/dice.csv | 60.0 MB | 0a7155e7f1e98c36e6407b20de57ce9028c057ff6e1eb839ccfcb74ec54a5107 |
lab/results/outcomes/hmac/legacy-dice.csv | 248.1 MB | ca2af5cde45010fa214d3efad80b99282d7c8d15ded30037679b944a0fcd6cff |
lab/results/outcomes/hmac/legacy-roulette.csv | 244.3 MB | 36300947bc4ffe3f6d56d6805269681418cb552dad8c2b9dffeb0ede67a756da |
lab/results/outcomes/hmac/roulette.csv | 56.1 MB | 110bef6e427f854f23b4762977a53d5b26cf4754d7613e2d669140c33b1e9c81 |
lab/results/outcomes/hmac/slot-reel.csv | 81.1 MB | 90a5d336c3b18ca3029844e5aaca59aa93bf5cc9967f46b104585d821840b8f9 |
lab/results/raw/chacha/raw-32gbit.bin | 4000.0 MB | d8b314f6a9b8726238cdbd8b8b04c947de17eb4f009c4fee75b01a6d3e07dfda |
lab/results/raw/hmac/raw-32gbit.bin | 4000.0 MB | 68a170a0fb14a4746c8acbcf4ac36e32eb2a807d8f0c9b240ab51028662e43d3 |
▸FOR AUDITORS
- Repository: github.com/Quecko-Org/rain-playmarket-fe-dev · branch
sdk-v2↗ — a mirror branch (force-pushed snapshot) on an existing org repo; a dedicated repository is pending. Mirror head0e891111d0c6; source-of-truth commit for this page9636833236e4ac4898c00cdfa995afffe1da1a60(lab counters refreshed after the mirror push — the mirror lags by the running-job refreshes only). - Build & test:
git clone -b sdk-v2 https://github.com/Quecko-Org/rain-playmarket-fe-dev.git rain-sdk-v2 && cd rain-sdk-v2 npm ci && npm run build && npm run typecheck npm test # unit: KATs (FIPS 180-4, RFC 4231, NIST CAVP HMAC_DRBG ×32, RFC 8439, keccak), ceremony, crash-persistence, selftest, e2e npm run test:py # Python mirror byte-identical (26 pytest) npm run lab:verify-build # rebuild → SHA-256 of 214 dist files == lab/FINGERPRINTS.json (run it twice) bash lab/tools/status.sh # what the lab jobs did / are doing - Re-run the evidence:
lab/tools/run-dieharder-stream.sh <mech> <workers>(resumable),run-sts.sh,run-testu01-crush-stream.sh,run-outcomes.sh; report regenerates withreport-tables.sh fill. - Verify a live round without any of this: walkthrough with an Arbiscan example.
- Contact: via the operator address in the deployment log / rainsandbox.xyz; disclosure policy and SLAs in SECURITY-PROGRAM.
▸FOR LABS — PRE-SUBMISSION QUESTIONS WE WANT ANSWERED
From LAB-SUBMISSION-REQUIREMENTS §8 step 9. Every lab offers a pre-submission call; these are the points we would put on the agenda.
- Player-supplied entropy classification. How will you treat the player's
pRev_kunder GLI-19 §4.5.2(e) ("associated equipment shall not influence or modify the behaviors of the game's RNG") and AGCO 4.26 ("impervious to the player")? Our position: the player can add entropy but cannot bias — one-page proof in STATE-MODIFICATION / RNG-DESCRIPTION App. A; adversarial-input runs in the report. % Nlegacy vsintBelow. The certified path is rejection sampling (bias = 0). The deployed Solidity verifiers useuint256(r) % N(bias < N/2²⁵⁶, ≈ 10¹⁵³ draws to detect). Does the legacy reduction pass source review as-is, or must live games move tointBelowbefore a platform submission? (SCALING-PROOF)- Draw counts. We have 10⁸ scaled outcomes per game for chacha20 and 2 × 10⁷ (10⁸ running) for hmac-drbg. What volume per game and per mechanism do you require, and do you re-collect on your hardware from our tools?
- Scope of "the RNG under test". Component RNG certificate (UKGC/MGA style — generator + scaling only) vs GLI-19 platform submission that pulls the on-chain contracts in scope?
- Two mechanisms. Certify hmac-drbg and chacha20 as two implementations (§4.5.1), or one only?
Precedents, stated honestly: BMM certified Chainlink VRF v2 under GLI-19 (2 Sep 2022); iTech Labs issued UKGC-RTS RNG certificates to BC.Game (2019, 2020) and Hash Games (2024) for server-seed provably-fair systems. Both certified the generator + scaling under the ordinary RNG standard. No lab has publicly certified a two-party commit-reveal where the player is an entropy source and a contract enforces the commitment. Expect an engineering-level review, not a template job — and do not expect the certificate to say "provably fair".
Scope reminder. This package covers the RNG component. The casino contracts (ChannelManager, verifiers, HouseBond), the games and the Outperform markets are documented elsewhere on this site and are not part of this submission; the Outperform markets do not use RAIN RNG at all (house-signed price oracle, disclosed). Both reference deployments remain play-money and pre-audit. Whitepaper: v1.2 PDF · MD. Page generated 2026-09-14 09:34 UTC; live counters re-read on every rebuild.