The math depends on which fail model is true — pick one above to see how it changes things
Two Possible Fail Models — Unresolved
The game says "10% fail per cog start." This could mean two very different things:
Flat (10% every cog): Each cog independently has a 10% fail chance — the rate itself never changes. Survival to cog 6 = 0.9⁶ = 53.1%. Under this model, EV is always positive and grows with each cog — always push to max cogs.
Escalating (10%→47%): The fail chance itself grows each cog — cog 1=10%, cog 2=19%, cog 3=27%, cog 4=34%, cog 5=41%, cog 6=47%. Survival to cog 6 drops to just 10.9%. Under this model, EV peaks around cog 4 and going to cog 6 is barely break-even — pushing to max cogs is a mistake.
Community read: going past cog 3 "feels like a coinflip," and cog 4-5+ feels progressively scarier — consistent with the escalating model (53%→35%→21%→11%) more than flat (73%→66%→59%→53%). But this is subjective and unverified. Use the dropdown above to see both.
How It Works
Each cogwheel gives you vouchers plus 1 extra random voucher, and the next cog gives +1 more than the last — so vouchers escalate fast. Each cog start has a flat 10% chance to wipe: the event ends immediately and you lose everything.
The Hard Numbers
Complete all 6 cogs:53% of runs
Wipe before cog 4:27% of runs — zero reward
Breakeven (conservative voucher dist):cog 4
Breakeven (high rarity voucher dist):cog 2
EV at cog 6 (conservative):+141% of base run value
EV at cog 6 (high rarity):+547% of base run value
Frequent wipes on cog 4–5 are not bad RNG. Nearly half of all Roulette runs zero out. That is the mechanic working as intended.
Why Max Cogs Is Always Correct
Cog 6 gives 27 vouchers. Cog 3 gives 9. The runs that survive and pay out are so valuable that they more than cover the runs that wipe — mathematically, every early cashout costs EV regardless of how the run has gone so far.
"But I'm Seeing Rainbow Vouchers — Should I Cash Out?"
No. Seeing high rarity drops early means this is exactly the run you want to complete. Those early cogs are just the start — cog 6 on a high rarity run is where the real payout is. Cashing out at cog 3 because you already got a rainbow is leaving most of the value behind. The temptation to lock in a good start is understandable, but the math is clear: keep going.
When It's Okay to Stop Early
Under flat: stopping early is a variance-management trade — you give up EV for stability, not playing "better."
Under escalating: stopping at the ★-marked optimal cog (shown in the table above for your gear count) isn't a trade-off at all — it IS the better play. Pushing past it loses both EV and increases variance.
Encounter Rate & The Season Reality
Roulette is a Rainbow service (~1% encounter rate). At real-world mapping rates you see it roughly once per 100 maps. At full deep-space juice a failed Roulette is a lost map — clockwork IS the map value.
When you see it — per-encounter result:
Goes negative (wipe):~43% of encounters
Floor (P10):−1 map cost
Average win:+4.9× map cost profit
Reliable positive at:~500 maps (93%)
At ~1 encounter per 100 maps, 500 maps = ~5 Roulettes. That's roughly one full 90-day season of serious mapping before the EV advantage meaningfully materializes. The "feels like always skip" instinct is statistically valid for anyone outside streamer/nolifer territory.
Voucher rarity distribution is unverified — the difference between conservative and high rarity is dramatic and this remains the biggest unknown in the Roulette ranking. Mechanic data from the
Chinese community doc.