Stress test
Published results of the project's pool simulation (packages/sim/results/2026-11-12, run on 2026-11-12). Every number below follows from the assumptions on this page. None is measured.
Assumptions
All of these are assumptions- Size of the runs
- 200 agents, 60 days, 1,000 seeded runs per scenario (seeds 1 to 1000)
- Assumed chance a contract changes, per 30 days
- Underwriting assumption, not a measurement. The pool's own declared values: immutable 0.5%, local upgradeable 5%, global by account 10%.
- Pool rules
- Capital adequacy k = 5, concentration limit 2 x capital, as the contract declares them.
- What the model leaves out
- Disputes and the arbiter, a wrong or late attestor, several underwriters, several tokens, and agents changing behaviour after a loss.
The full input of each scenario is in packages/sim/results/2026-11-12/assumptions.json, and the method in REPORT.md beside it.
One table, six scenarios
1,000 runs each| Measure | Normal operation | One target changes | Mass correlated claims | Underwriter withdrawal | Underwriter withdrawal, lock switched off (counterfactual, not the contract) | Capital not enoughThe scenario that breaks the pool |
|---|---|---|---|---|---|---|
| Starting capital, NEAR | 3,000 | 3,000 | 3,000 | 3,000 | 3,000 | 100 |
| Chance the pool is insolvent, 95% interval | 0.0% (0.0% to 0.4%) | 0.0% (0.0% to 0.4%) | 0.0% (0.0% to 0.4%) | 0.0% (0.0% to 0.4%) | 0.3% (0.1% to 0.9%) | 98.8% (97.9% to 99.3%) |
| Ending capital, median (5th to 95th), NEAR | 4,765 (3,814 to 5,105) | 4,165 (3,457 to 4,580) | 2,911 (2,183 to 3,450) | 3,511 (2,871 to 4,012) | 1,339 (531 to 1,857) | 108 (13 to 550) |
| Claims paid, median, NEAR | 115 | 720 | 1,950 | 1,393 | 1,101 | 1,857 |
| Most owed and unpaid at once, median (95th), NEAR | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 0 (0) | 707 (1,227) |
| Solvency ratio, median (5th) | 0.44 (0.36) | 0.38 (0.32) | 0.29 (0.25) | 0.33 (0.27) | 0.21 (0.19) | 0.01 (0.00) |
| Time to insolvency, median days | not insolvent within 60 days | not insolvent within 60 days | not insolvent within 60 days | not insolvent within 60 days | 41.1 (3 runs) | 30.3 (988 runs) |
"Insolvent" means the pool refused at least one claim because it was larger than the book's capital. A solvent scenario says only that it was not insolvent within 60 days.
The pool breaks in this scenario. The pool starts with 100 NEAR, with no adequacy rule and no concentration limit, against all the coverage the agents want to buy. On day 30 the upgrade authority with the most coverage behind it changes everything it controls, and every exposed agent claims within two hours. This is the scenario that breaks the pool. It breaks in 99% of runs, a median of 30.3 days in. It is shown as it happened.
What each scenario does
- Normal operation
- Independent, infrequent code changes, at the pool's own declared P(event). No scripted event.
- scripted event: none scripted. underwriter withdrawal: none. pool rules: on.
- One target changes
- On top of the random changes, the implementation with the most coverage on it changes on day 20.
- scripted event: day 20, the most exposed implementation changes. underwriter withdrawal: none. pool rules: on.
- Mass correlated claims
- On day 30 the upgrade authority with the most coverage behind it changes everything it controls at once, and every exposed agent claims within two hours of the report becoming claimable.
- scripted event: day 30, the authority with most coverage behind it changes everything it controls. underwriter withdrawal: none. pool rules: on.
- Underwriter withdrawal
- On day 25 the underwriter tries to withdraw half its shares, daily until it works; on day 40 a mid-sized change lands. The contract refuses withdrawals while any coverage is open or in its reporting period.
- scripted event: day 40, the most exposed implementation changes. underwriter withdrawal: day 25, 50% of free capital. pool rules: on.
- Underwriter withdrawal, lock switched off (counterfactual, not the contract)
- The same run with the contract's withdrawal lock removed, to show what the lock protects. The contract does not behave like this.
- scripted event: day 40, the most exposed implementation changes. underwriter withdrawal: day 25, 50% of free capital, lock switched off. pool rules: on.
- Capital not enough
- The pool starts with 100 NEAR, with no adequacy rule and no concentration limit, against all the coverage the agents want to buy. On day 30 the upgrade authority with the most coverage behind it changes everything it controls, and every exposed agent claims within two hours. This is the scenario that breaks the pool.
- scripted event: day 30, the authority with most coverage behind it changes everything it controls. underwriter withdrawal: none. pool rules: off: no adequacy rule or concentration limit.