In one sentence
Clash Watchdog recorded 10,472 rounds of Crash Duel X (SmartSoft) from 11 Oct 2026 to 14 Oct 2026, 03:27 UTC; at a fixed 2x cash-out the measured return was 97.3% (95% CI 95.4% to 99.2%), against the provider's stated 96%.
Measured return vs the provider's claim
What is 95% interval?
95% interval
97.3%
The provider states an RTP of 96%.
Measured here at a fixed 2x cash-out: 97.3% (95% CI 95.4% to 99.2%) over 10,472 rounds, as of 14 Oct 2026, 03:27 UTC. The interval and the claim overlap, so the count so far is consistent with it. That is not proof that the game is fair.
Instant crash: rounds below 1.01x
Of 10,472 recorded Crash Duel X rounds (as of 14 Oct 2026, 03:27 UTC), 2.1% ended below 1.01x, shown as 1.00x: the instant crash (95% CI 1.8% to 2.4%). If the provider's 96% holds and the game follows the standard crash formula (the chance of reaching x is the RTP divided by x), about 5.0% of rounds would end below 1.01x. The formula's 5.0% lies outside this 95% interval, so over these 10,472 rounds Crash Duel X's share differs from the standard formula. That is a measurement, not a finding about fairness: the provider's claim is tested by the measured return at 2x.
How often rounds reached 2x, 5x, 10x and 100x
Shares of past rounds, each on its own count. They are not odds for the next round.
| Multiplier | Rounds that reached it | 95% CI | n | If the claim holds ‡ |
|---|---|---|---|---|
| 2x | 48.6% | 47.7% to 49.6% | 10,472 | 48.0% |
| 5x | 20.1% | 19.3% to 20.8% | 10,472 | 19.2% |
| 10x | 10.2% | 9.6% to 10.8% | 10,472 | 9.6% |
| 100x | 1.0% | 0.8% to 1.2% | 10,472 | 1.0% |
Highest multiplier recorded
5,000x†, the highest the lab has recorded for Crash Duel X in 10,472 rounds (as of 14 Oct 2026, 03:27 UTC). † Observed in recorded rounds, so it has no interval. It is a record of past rounds, not a target or a prediction. Crash Duel X's documented cap is 5,000x, and the lab's own record has reached it.
Where rounds ended
Every recorded round falls in one band. The hatched band is the instant crash. The counts are frozen at the snapshot.
| Band | Share | 95% CI | Rounds |
|---|---|---|---|
| <1.01x (Instant crash) | 2.1% | 1.8% to 2.4% | 220 |
| 1.01-1.5x | 32.7% | 31.8% to 33.6% | 3,426 |
| 1.5-2x | 16.5% | 15.8% to 17.3% | 1,732 |
| 2-3x | 15.5% | 14.8% to 16.2% | 1,619 |
| 3-5x | 13.1% | 12.5% to 13.8% | 1,375 |
| 5-10x | 9.8% | 9.3% to 10.4% | 1,030 |
| 10-30x | 6.8% | 6.3% to 7.3% | 711 |
| 30x+ | 3.4% | 3.1% to 3.8% | 359 |
Milestones
RTP @ 2x and the instant-crash rate at fixed round counts. The 95% interval narrows as rounds come in.
| Milestone | Period (UTC) | RTP @ 2x (95% CI) | Instant crash (95% CI) |
|---|---|---|---|
| 10,000 rounds | 11 Oct 2026 – 14 Oct 2026 | 97.1% 95.2% to 99.1% | 2.1% 1.8% to 2.4% |
| All rounds in this snapshot 10,472 rounds | as of 14 Oct 2026, 03:27 UTC | 97.3% 95.4% to 99.2% | 2.1% 1.8% to 2.4% |
Drift check: starts once 30,000 rounds are in complete blocks of 5,000 (10,000 so far).
How it was measured
Rounds were read from the provider's own free play-money demo, one row per round as it ended. No money was staked, and no operator account was used.
Every round the feed delivered is counted, in order. Rounds are never sampled, filtered or dropped to shape a figure. The lab records the rounds it sees, not every round ever played: gaps from reconnects follow the clock, not the outcomes.
A round counts as reaching a multiplier when its crash point is at least that multiplier. The fixed 2x cash-out returns twice the stake when a round reaches 2.00x and nothing otherwise; the measured return is the average of those returns.
Methodology lab-v2.0 (site) and crashlab-v2 (feed).
How to verify
The raw records behind this report are sealed one UTC day at a time. These sealed days, each with its row count and SHA-256 fingerprint, were published before the snapshot was taken:
| Day (UTC) | Rows | SHA-256 |
|---|---|---|
| 1,399 | 0ecd2f00f7cdb8ea8231f37825c3755611b5c34eac264e2e517f0593e24efd88 | |
| 4,217 | aaa99f77dabd780dd8f36f5aeef5d994704fe74811335e4a95520f03bfb0c20e | |
| 4,300 | 61a89b83df30e73dbabbac8659f1647346c004e938b36ada51a9522d73baff87 |
What this report does not show
- No prediction: these are shares of past rounds, and none of them says what the next round will do.
- Pending tests stay pending: the lab has published no result for lag-1 autocorrelation of consecutive rounds or for hour-of-day patterns.
- One game and one period: the figures describe Crash Duel X from 11 Oct 2026 to 14 Oct 2026, 03:27 UTC. They do not rank games or months. An edition is a dated cut of everything the lab has recorded for the game, not one month's rounds.
- A free demo, not a real-money table: no operator's real-money table was measured.
- Not proof that the game is fair, and not a recommendation to play it.
Cite this report
The id below names a frozen snapshot: it always means these exact figures. A game page's own id (such as cw:crash-duel:2026-10:v1) names a game and a month, and its figures keep moving.
Was this page useful?
Anonymous: no account and no cookie.