Completed worlds
1 000Independent recorded seedsThe policies stayed fixed · chance changed
What happened in 1,000 other worlds
An apparent star may simply have received an extreme draw. Replication shows the full range of each policy's results instead of treating one rank as proof of skill.
Visible house-result percentile
96,5%Among replication-world GGRMedian policies ahead
244 / 1,00024,4%Policy manifest
FixedOnly declared stochastic choices and outcomes changeAcross all worlds
The final house result
-R 2MFinal P&LR 2M
Why replication matters
A single winner makes an easy story.
Among 1,000 policies, someone will look exceptional. The experiment does not infer ability from that maximum. It compares the visible result with the same policy's range, chance of profit, risk of ruin and median rank under new seeds.
- Hold policy and game definitions constant.
- Generate an independent world stream.
- Repeat the six-month schedule.
- Report distributions rather than a selected world.
The groups across worlds
How cohort medians moved
These ranges show how each cohort's median P&L changed across 1,000 complete worlds.
Small bets, frequent
-R 1 0995th–95th world range
-R 1 381 to -R 815
Small bets, infrequent
-R 2105th–95th world range
-R 258 to -R 162
Medium bets, frequent
-R 3 9375th–95th world range
-R 4 498 to -R 3 210
Medium bets, infrequent
-R 8415th–95th world range
-R 1 088 to -R 607
Large bets, frequent
-R 9 6625th–95th world range
-R 9 896 to -R 8 707
Large bets, infrequent
-R 1 7285th–95th world range
-R 2 287 to -R 1 161
Loss-chasing
-R 6 1765th–95th world range
-R 7 363 to -R 4 643
Strict limits
-R 1 2365th–95th world range
-R 1 558 to -R 916
Variable / random staking
-R 3 9925th–95th world range
-R 4 831 to -R 3 145
Random control
-R 2 5435th–95th world range
-R 3 315 to -R 1 773