Eligible is not liquidated. Ukemi measures the difference; once a stratum is committed, the gate hands back an upper bound on the amount liquidated, per stratum.
A measure of liquidation exposure on one lending venue: the lending book read at one declared block, the oracle price path the protocol actually consulted, and, once a stratum is committed, a conformal upper bound on the amount liquidated. Never a probability of being right.
What Ukemi is
- A book at a block: every account holding the collateral and a debt, read at one declared reference block and reduced to a digest.
- A realized oracle path: the sequence of prices the protocol consulted over the window, recorded from chain events, not a simulated market.
- A conformal region, per stratum: for a calibrated class, an upper bound on the amount realized. Never a probability.
- An abstention when it cannot know: too few calibration points for a population (under_calib), a read without quorum, an account it cannot evaluate. The output is a named state, not a number.
What Ukemi is not
- Not a forecast of the next price, and not a forecast that more liquidations will follow.
- Not a probability of liquidation for an account, and not a probability that a bound is right.
- Not a rating, a gauge or a ranking of a protocol, a market or an account.
- Not a risk parameter: it sets no threshold and no cap. A curator stays the curator.
- Not a claim about a new event: once a stratum is committed, its bound is calibrated on one episode; exchangeability across events is named, not assumed.
The liquidation-eligible-coverage class is served through the gate. In words, without the figures of the served clause, the served state is:
where a stratum's calibration is committed, the gate serves a conformal upper bound on the liquidable amount; on every other stratum it abstains (under_calib)
On the committed stratum, as read from the served gate on 2026-09-25:
- 170 calibration points
- bound margin 1261.84298996 in the lending venue's oracle base currency; the upper bound for a prediction in this stratum is the prediction plus this margin
- calibration digest e7e673664c03e3c5d15956d864f8379b6fe4660ed689be38a85add95d4eff334
The calibration digest identifies the calibration points this bound is computed from; the gate returns it with every answer on this stratum, so an answer can be matched to its calibration.
The hypothesis report of the calibration course on one recorded lending episode (pre-registered outcomes, counts and digests) is on the course page. Committing the strata that meet the floor to the served class is a separate, recorded step.
When a class holds enough calibration, the served region is a conformal upper bound on the amount liquidated: an open floor by construction, not a calibrated bound. How far the bound reaches says nothing about being right; a loose bound is a legitimate, informative answer.
The bound is conditional, and the condition is stated, not hidden:
the bound holds only if yhat was produced by the frozen close-factor rule on a mono-collateral WETH account at the first crossing, which the gate does not check
The gate has three outputs, and every one of them is an answer. It commits, with the region, when the population holds enough calibration points and the request falls inside a region narrow enough to act on. It defers while its clock is open and the region is too wide to act on. It abstains otherwise, with a typed reason, for example: too few calibration points (under_calib: the region is withheld, and on a stratum that is not committed the served answer counts no calibration point; the measured counts are on the course page), a missing or disputed input, a request outside the region, an exhausted budget.
The strata, the coverage level and the minimum number of calibration points for a population are written down before the run. A region is only ever emitted for a population that holds enough calibration points; outside it the answer is under_calib, and for a stratum that is not committed the served answer counts no calibration point. The largest-amount stratum is expected to stay under_calib for a long time; the course page shows every stratum's measured count against its floor.
One venue, one collateral class
One lending venue, core market, single-collateral WETH accounts. An account holding any other collateral is excluded, marked non_evaluable. In the threshold recompute the other legs are held at their book-block price: a declared limitation, not a repricing.
Few episodes
Once a stratum is committed, its coverage holds only under exchangeability with the calibration episode. The distance to a new event is named, never estimated away.
Open questions stay open
The served-price delay seen in the design episode is unexplained; on the course episode the lag is measured against a pre-registered bound, not explained. An account in another efficiency-mode category is excluded, marked non_evaluable: whether that category covers the collateral cannot be checked off-line.