What counts as a loss
The permanent-loss event used for every lending-vault rating and loss measure.
The loss eventLink to this section: The loss event
A loss event occurs when the vault finishes one year with more than 1% of its starting value permanently lost. Value is measured in the asset the vault uses for accounting.
| Element | Definition |
|---|---|
| Measure | Decline in the vault's value at the end of the period |
| Horizon | one year |
| Threshold | More than 1% |
| Starting point | The vault's value at the start of the assessment |
| Unit | The vault's accounting asset |
Probability of loss, written P(loss), is the share of simulation paths that finish beyond this threshold.
Only the value at the horizon counts
- Starting value The vault's value on the day the assessment year begins.During the year the vault's value moves with the markets it lends to.
- Temporary decline The value can fall during the year and recover before it ends. That is not a loss event.Whether the value recovers or not, only the closing value is measured.
- End of the year The vault's value is measured once, at the end of the year.The closing value is compared with the starting value and the loss threshold.
- Permanent loss If the closing value is more than the threshold below the starting value, the year counts as a loss event.
How loss reaches the vaultLink to this section: How loss reaches the vault
A collateral price fall may push a borrower into liquidation. The liquidator receives collateral and uses available trading, redemption, or withdrawal routes to repay the debt. Any debt left unpaid becomes bad debt for the lending market. The vault bears its share of that bad debt.
This settlement path matters more than the price move alone. A large move can leave depositors whole when liquidation and recovery work well. A smaller move can create loss when a borrower is highly leveraged or the collateral is difficult to recover.
One distributionLink to this section: One distribution
The model keeps the final vault loss from every simulation path. The headline measures all come from this one distribution:
- P(loss): how often loss exceeds 1%.
- Expected loss: average loss across all paths.
- Conditional loss: average loss among paths that cross the threshold.
- Percentile loss: the loss level reached at a stated point in the distribution.
- Tail loss: average loss among the most severe paths.
Continue with Mapping the vault.