Advertised house edge: 1%.
Actual: up to 2.98%.

The captured game configuration declares rtp: 99 and house_edge: 1. It matches across September, seven logged-in reads on 8 October and the 9 October capture.

For a two-decimal target x above 1, RTP(x) = 0.99 × 0.98 × x / (x − 0.02). The instant-bust rule takes 1%; a second reduction in the multiplier formula lowers the return further. This assumes uniform HMAC output and excludes payout caps.

How the formula produces the return

Let u = h/e. The current non-instant crash point is (1 − u/50)/(1 − u). For a target x above 1, it reaches x when u ≥ (x − 1)/(x − 0.02). With uniform hashes, that gives a reach probability of approximately 0.99 × 0.98/(x − 0.02), including the separate instant-crash rule. Multiply by the target payout x to get the return.

Remove h/50 and the non-instant formula becomes 1/(1 − u). Its reach probability is 1/x. With the 1% instant-crash rule, that becomes 0.99/x, so x × 0.99/x = 0.99: a 99% expected return.

const e = 2 ** 52;
const result =
  Math.floor(100 * e / (e - h)) / 100;
return result.toFixed(2);

Rounding and precision: for a two-decimal target x, rounding the crash point down to two decimals doesn’t change whether it reaches x, so the corrected formula returns 99% at every target (METHOD.md, section 1).

This is a proposed correction, not a change Gamdom has made. It assumes the full target multiplier is paid.

All 99,900 tested targets from 1.01× to 1,000× return below 99%. The exact target-boundary check and an independent 60-million-hash simulation agree with the curve.

We also reconstructed 57,428 rounds between 19 September and 9 October from the chain linked to the published commitment. At every tested target, their realised return is within two standard errors of the formula’s curve. These were reconstructed, not individually observed, rounds.

57,428 reconstructed rounds: realised return versus formula

Reconstructed outcomes fall within two standard errors of the formula at each tested target.

Extra expected cost per bet = stake × (99% − RTP at the cash-out target). Equivalently: stake × 1.98% × (1 − 0.98 ÷ (target − 0.02)). At 2×, that is $10 per $1,000 wagered.

For a manual cash-out that won, use the multiplier at which it cashed out. A lost manual bet has no observed final cash-out target in the public feed; the player’s intended strategy cannot be recovered from that record alone.

The figures compare expected returns, not an individual refund entitlement or a claim that a particular losing round should have won. Caps, rounding and other account rewards require separate review.

My bet lost by a hair. Would a 99% game have paid me?

Not necessarily. A correct 99% game need not produce the same result for any single round. Different correct formulas can settle that round differently.

What changed is the odds. At 1.5×, the win probability is about 0.45 percentage points lower, roughly one bet in 224. Across all those bets, that difference is 0.67% of the stake. Our proposed compensation method spreads that expected cost across affected bets instead of guessing which individual losses should become wins.

We recorded nearly every bet in Gamdom's public Crash feed from 26 October, 19:19 UTC, to 28 October, 15:06 UTC: 5,381 rounds, 99,119 bets from 925 players, $850,279 wagered. Gamdom numbers every Crash bet, and the recording holds 99,119 of the 99,143 bets numbered in that time.

The feed reveals the cash-out only for winning bets. We estimate the target mix by weighting each winning stake by the inverse of its chance of winning. That reconstructed mix gives a 97.881% expected return, 1.119 percentage points below the stated 99% (95% range 1.005 to 1.246 points). Counting the losing bets too, whose targets sat above their round's crash point, gives 1.120. Two shorter recordings, on 21 September and 8 October, give 0.94 and 1.15 points.

That is $9,514 of extra expected cost in 44 hours. Per day, the window ran at $466,117 of Crash wagering and $5,215 of extra expected cost. It covers every hour of the day, most of them twice. Gamdom's bet numbers put it at 97% of the average bet rate from 21 September to 28 October.

The low case uses 27 October, the one full day in the window ($442,320 wagered), at the low end of the edge range: $1,346,883 over 303 days. The central case is $1,580,281. The high case uses the faster bet rate of 21 September to 8 October, at the high end of the edge range: $1,961,655.

The period starts on 2 January 2026: Gamdom's verifier says it covers Crash results from that date. The extrapolation assumes late October's pace held across the period. Gamdom's bet numbers cover only the five weeks from 21 September, and the 10 largest players made 53% of the 44 hours' wagering, so seasonality and how often high rollers play could move the total.

These are scenarios, not confidence bounds. They measure the expected cost of worse odds, not what the sampled players happened to win or lose. Gamdom should publish the relevant history so the report can replace assumptions with a complete total.