Duel fairness construction: HMAC-SHA512 · 52-bit doubles · cursor per event (extracted from the client bundle). Independently verifiable: yes. Client seed: settable. Fairness score 7.5/10. Snapshot 2026-08-18.
Duel originals derive outcomes from HMAC-SHA512 keyed with the server seed over "clientSeed:nonce:cursor"; the first seven bytes of each digest form a 52-bit uniform double and the cursor increments per event. We extracted this construction from the reference verifier bundle and our verifier (duel-sha512 engine) reproduces Duel rounds. Seed pairs rotate normally and the client seed is settable.
How to verify a Duel round
Save the committed server-seed hash (or the published chain/proof values on chain- and VRF-based games) before playing.
Set your own client seed where the scheme supports one.
After the session, rotate the seed pair or open the revealed values; confirm the revealed seed hashes to the commitment.
Recompute the round from the revealed values with the operator’s published per-game formula or our seed verifier.
A mismatch on any single round — after checking encoding and nonce indexing — is the only evidence of tampering that actually means something.
Originals and declared house edge
Verification proves a round was not altered after the bet; the declared edge is applied openly in the payout table and is unaffected by any of this.
Game
Declared edge
Notes
Dice
0.1%
0.1% advertised edge applies below the ~$25k potential-profit boundary; above it the staircase rises toward 1%.
Blackjack
not declared
No declared RTP in our catalogue.
Plinko
0.1%
Crash
not declared
No declared RTP in our catalogue.
Mines
0.1%
Beef
not declared
Keno
0.1%
Castle Roulette
not declared
Video Poker
not declared
Groomer's Van
not declared
Rock Paper Scissors
not declared
Anomalies and caveats
The house edge is not a single number: a public game-config endpoint ships a 91-tier staircase (0.1%–1.0%) keyed to potential profit — the first boundary sits near $25,000 of potential profit whatever the win chance (WagerDB audit, 2026-08-16). The interface describes this as bet-size based.
Several originals (Crash, Blackjack, Beef, Castle Roulette, Video Poker) carry no declared RTP in our catalogue.
Anomalies describe observations from public code, documentation or reporting. They are not accusations of cheating; where something is unproven we say so.
Scoring breakdown — 7.5/10
Verifiability: 2/2 — Can a player independently recompute a round end-to-end? 2 = full commit-reveal with a public algorithm; 1.5 = verifiable with caveats (mixed or partially undocumented constructions); 1 = commitment check only; 0 = nothing to check.
Player control: 2/2 — Does the player influence the draw? 2 = settable client seed, rotation reveals the server seed, per-bet nonce; 1 = player input absent on some games (e.g. hash-chain titles); 0 = no player input at all.
Documentation: 1/2 — Public documentation quality. 2 = per-game algorithm detail and/or sample code or a working verification sandbox; 1 = generic commit-reveal description; 0 = none found.
Edge transparency: 0.5/2 — Is the price of each game stated and stable? 2 = per-game RTP/house edge published and consistent with our catalogue; 1 = published but with gaps, wide spreads or marketing mismatches; 0.5 = dynamic/undisclosed edge behaviour; 0 = nothing published.
Track record: 2/2 — Known fairness-relevant incidents. 2 = none found; 1.5 = minor or contested; 1 = material public disputes relevant to trust in the fairness claim.
The score grades the fairness system and its transparency, not the casino overall. It says nothing about licensing, solvency or withdrawal reliability.
Method and sources
WagerDB reverse-engineering: duelEngine extracted from the reference verifier bundle; edge staircase read from the public, unauthenticated api/v2/dice/config endpoint (2026-08-16, see /research/house-edge).
Analysis based on public documentation, community verifiers and — where noted — the operator's public client-side JavaScript, snapshot 2026-08-18. Constructions can change; the date matters.