Zero edge is not a marketing slogan — it’s a specific mathematical configuration. This page explains what happens inside the game when the house advantage is set to 0%: how multipliers are calculated, how outcomes are generated and verified, and where the “margin” actually lives in a standard game so you can see exactly what changes when it’s removed.
Where the Margin Lives in a Normal Game
Every casino game has a payout formula. The house cut is built into that formula — it’s the gap between what fair math would pay and what the game actually pays.
Dice: the simplest example
A dice game generates a random number between 0 and 100. You set a target and bet Over or Under. A standard way to express the payout formula is:
Multiplier = (1 / win probability) × (1 − house edge)
At 50% win probability:
| House Advantage | Calculation | Multiplier | Expected Loss per $1 Bet |
|---|---|---|---|
| 0% (fair) | (1/0.50) × 1.00 | 2.000x | $0 |
| 1% (Stake) | (1/0.50) × 0.99 | 1.980x | $0.01 |
| 3% (typical) | (1/0.50) × 0.97 | 1.940x | $0.03 |
The margin is the (1 − house edge) term. At 0%, it equals 1.00 — meaning the formula produces mathematically fair payouts. At 1%, the formula multiplies every payout by 0.99, shaving 1% off each win. The player never sees this directly; they just see a multiplier of 1.98x instead of 2.00x.
At 0% advantage, removing this term is all that changes. The random number generation, the seed system, the game interface — everything else stays the same. Only the payout table shifts to fair values.
Crash: the bust distribution
In Crash, the house cut is embedded in the bust distribution — the probability curve that determines where the multiplier stops each round.
In a simplified fair (0% edge) crash model, the survival function can be expressed as:
P(multiplier ≥ m) ≈ 1 / m
This means: roughly 50% of rounds reach 2x or higher, 10% reach 10x or higher, and 1% reach 100x or higher. An auto-cashout at any target m yields an expected value of approximately (1/m) × m = 1.00 — break-even.
In edge-bearing implementations (e.g. Aviator at 3% house edge), that survival curve is shifted downward so that the expected return falls below 100%. Fewer rounds reach high multipliers, and low-multiplier outcomes (including rounds that end before any meaningful cashout) become more frequent. The exact mechanism differs by provider — some inject a discrete “instant bust” rate, others shift the entire distribution — but the net effect is the same: the mathematical return drops to the configured RTP.
At 0% edge, whatever mechanism the provider uses is either removed or set to neutral, so the distribution produces fair expected payouts at every cashout target.
Mines: per-click multipliers
In Mines, fair cashout values are derived from the inverse of the cumulative survival probability after each successful click. On a 5×5 grid with 3 mines, the probability of surviving k clicks is calculated using combinatorial math (specifically, the hypergeometric distribution):
P(survive k clicks) = C(22, k) / C(25, k)
The fair multiplier at each cashout point is 1 / P(survive k clicks). With a house advantage, this value is reduced — the exact method depends on implementation (per-click tax, cashout-level tax, or hybrid).
The following table shows exact values for a 3-mine configuration:
| Cashout After Click | Survival Probability | Fair Multiplier (0%) | At 3% Margin (on cashout) | Difference |
|---|---|---|---|---|
| 1 | 88.00% | 1.136x | 1.102x | −0.034x |
| 2 | 77.00% | 1.299x | 1.260x | −0.039x |
| 3 | 66.96% | 1.494x | 1.449x | −0.045x |
| 5 | 49.57% | 2.018x | 1.957x | −0.061x |
| 8 | 29.57% | 3.382x | 3.281x | −0.101x |
| 10 | 19.78% | 5.055x | 4.903x | −0.152x |
Note: “At 3% margin” column uses a simplified model where the house cut is applied once to the final cashout value. Some implementations compound the reduction per click, producing slightly different numbers. Values are illustrative of the principle, not a specification of any particular provider’s math.
At 0% advantage, each cashout pays its true inverse survival probability — no reduction applied.
How Provably Fair Ensures the Advantage Is What They Claim
Provably Fair (PF) is the verification mechanism. It doesn’t set the house cut — the payout formula does that. What PF does is prove that the random outcome wasn’t manipulated after your bet was placed.
Note: The following describes the standard seed-based PF pattern (server seed + client seed + nonce) used by many crypto originals including Duel. Other implementations exist — for example, some platforms use a reverse hash chain approach. The core principle (pre-commitment to outcomes) is shared, but the technical details vary.
The Seed System
- Server seed: Generated by the casino. Its SHA-256 hash is shown to you before the round. The casino cannot change the seed after showing you the hash — any change would produce a different hash.
- Client seed: Set by you (or auto-generated). You can change it anytime. Because the casino has already committed to the hashed server seed before receiving your client seed, they cannot alter it after the fact without changing the hash — which you would detect.
- Nonce: A counter that increments with each bet (0, 1, 2, …). Prevents reuse of seed combinations.
How the Outcome Is Generated
The three inputs are combined: HMAC_SHA256(server_seed, client_seed:nonce). The output is a deterministic hash. A mapping function converts this hash into the game result (a dice number, a crash bust point, a mine placement grid).
After a series of rounds, the casino reveals the unhashed server seed. You can then:
- Run
HMAC_SHA256(revealed_server_seed, your_client_seed:each_nonce)for every round. - Apply the same mapping function to each hash.
- Verify that the result matches what was shown to you in-game.
If any result doesn’t match, the casino altered the outcome — which is provable fraud.
What PF Proves (and Doesn’t Prove)
Proves: The outcome was determined before your bet. The casino didn’t change results after seeing your wager. Individual rounds are fair in the sense of being pre-committed.
Does not prove: The overall return level. PF verifies individual outcomes, not the aggregate payout formula. A game can be perfectly provably fair and still have a 97% RTP — because tamper-proof randomness and a negative payout formula can coexist. To verify the house cut, you would need to check the multiplier tables against the fair math — which we do on each game audit page. At Duel, the multiplier tables match the 0% formula within rounding (based on published materials, April 2026).
What Exactly Changes When the Advantage Goes to 0%
Here’s the complete list of what changes and what stays the same:
| Component | At 1–3% House Cut | At 0% House Cut |
|---|---|---|
| Random number generation | HMAC_SHA256 | Same |
| Seed system | Server + client + nonce | Same |
| Game interface | Standard UI | Same |
| Bet placement | Standard | Same |
| Payout formula | Multiplied by (1 − margin) | Multiplied by 1.00 (no reduction) |
| Dice multiplier at 50% | 1.98x (1%) / 1.94x (3%) | 2.00x |
| Crash bust distribution | Adjusted downward from fair curve | Fair inverse distribution |
| Low-multiplier outcomes (Crash) | More frequent in many edge-bearing designs | Occur at natural rate only |
| Mines per-click multiplier | Each cashout reduced by margin | Each cashout at fair conditional odds |
| Expected player return | Negative (by the margin %) | Zero |
The change is surgical: one coefficient in the payout formula shifts from (1 − margin) to 1.00. Everything else — the randomness, the verification, the interface — is identical.
Why 0% Advantage Is Not “Free Money”
A common reaction: “If the house cut is zero, I can’t lose in the long run, so I should just keep playing until I’m ahead.” This reasoning has two flaws:
Flaw 1: “Long run” requires infinite play
Your average return per bet converges toward fair value over a very large number of bets. Over any finite number, your actual balance will be above or below zero — possibly far from it. A 1,000-round session at 50% win chance has a standard deviation of approximately ±$316 at $10/bet. That means a result anywhere from −$316 to +$316 is within one standard deviation — entirely consistent with fair game math. Going −$300 is not unusual at all (it falls within one standard deviation). These are not margin-driven losses — they’re pure variance, and they’re real.
Flaw 2: Bankroll limits truncate the distribution
If your bankroll is $500 and variance pushes you to $0, you’re done. You can’t play the remaining rounds that would bring your average return back toward the expected value. The math only promises convergence over unlimited play; real players have limited money and limited time. This is the gambler’s ruin problem.
How to Verify Zero Edge Yourself
You don’t need to trust anyone’s claim. Here’s how to check:
Method 1: Audit the multiplier tables
For Dice: set different win chances (10%, 25%, 50%, 75%, 95%) and check whether the displayed multiplier matches 1 / win_probability. At 0% house cut, they should match exactly (within display rounding). At 1%, they’ll be 1% lower. This is the fastest and most direct check — if the payout table itself shows a reduction, the margin is visible immediately.
Method 2: Verify individual rounds via PF
After a series of rounds, reveal the server seed and verify each round’s outcome using the HMAC_SHA256 formula. This confirms individual fairness (outcomes weren’t manipulated) but does not confirm the aggregate return level.
Method 3: Statistical sampling (advanced)
Record 10,000+ rounds and calculate the actual return rate. With sufficient volume, the observed return should cluster around the claimed RTP, but the acceptable deviation depends heavily on game variance and sample size. Low-variance games (Dice at 50%) converge faster; high-variance games (Crash targeting 50x) require far more rounds. This method provides the most direct evidence of the claimed return level but requires significant volume and careful record-keeping.
For a detailed PF verification walkthrough, see How to Verify Provably Fair.
Frequently Asked Questions
Is the house advantage really exactly 0.00%, or is there rounding?
Two different checks apply here. First, you can audit the payout tables directly: if the Dice multiplier at 50% win chance reads 2.000x, the formula matches 0% — this is detectable immediately, no sampling needed. Second, whether there’s a sub-0.01% bias hidden in unresolved decimal places would require either access to the source code or a very large statistical sample (hundreds of thousands of rounds). At Duel, the visible multiplier tables match the fair formula within display precision (3–4 decimal places, based on published materials, April 2026).
Can the casino change the return level without telling me?
The payout formula (multiplier tables) is visible in the game interface. If the multiplier for 50% win chance changes from 2.000x to 1.998x, the house cut has shifted from 0% to 0.1%. You can detect this by checking the tables before each session. PF does not guard against formula changes — only against outcome manipulation within a given formula.
Do all Duel Originals use the same mechanism?
The principle is the same (remove the (1 − margin) coefficient), but implementation differs by game type. Dice adjusts the multiplier formula. Crash adjusts the bust distribution. Mines adjusts per-click payouts. Plinko adjusts lane multiplier calibration. The math is game-specific; the concept is universal.
Does Provably Fair guarantee 100% RTP?
No. Provably Fair guarantees that individual outcomes weren’t manipulated — it proves pre-commitment to results. It says nothing about the payout structure. A game can be perfectly provably fair and still run at 97% RTP, because tamper-proof randomness and a negative-EV payout formula can coexist. To verify the return level, you need to audit the multiplier tables, not just the PF system.
Can a game be Provably Fair and still have a house advantage?
Yes — and most provably fair games do. PF protects against outcome manipulation; the house cut is a separate parameter in the payout formula. Stake Crash (99% RTP) and Aviator (97% RTP) are both verifiable via PF, but both have a built-in margin. The distinction between “fair randomness” and “fair payouts” is critical.
How does the casino make money at 0% house advantage?
Zero-edge games are typically offered within a broader business model. Duel, for example, funds zero-edge play through daily allowance limits ($50,000 turnover/day), after which RTP drops to 99.9%. The zero-edge offering functions as a differentiation tool and acquisition channel, not as the sole revenue source. See Zero Edge Allowance Explained for a full breakdown of the mechanics.
Bottom Line
Zero edge is a one-line change in the payout formula: the coefficient (1 − house margin) becomes 1.00. This makes every multiplier, every bust distribution, and every per-click payout mathematically fair. The randomness engine (Provably Fair), the game interface, and the rules don’t change.
You can verify individual outcomes through PF, and you can verify the return level by checking the multiplier tables against fair math. The system is transparent by design — not because you’re asked to trust it, but because you’re given the tools to check it.
