Evidence
Every liquidation of a tokenized stock on Lista Lending from 18 June 2026 to 22 September 2026, placed on the New York week. The risk builds while the market is closed; the money is lost when it opens.
Where the money is lost
81%of the dollars real borrowers lost were repaid in the first 90 minutes after an open: $24.8k of $30.6k, across 33 liquidations.
0 of 120liquidations of bStock collateral happened on a weekend, although Friday 8 PM to Sunday 8 PM is 29% of the clock.
81%of an ordinary week New York's regular session is closed. Prices that move then are settled in the first minutes of the next session.
Each star is one liquidation; its distance from the centre is the dollars repaid. Hollow rings are 87 tiny positions of one test address. Every star links to its transaction.
By session
The 120 liquidations with a bStock as collateral. Clock share is the part of the sample period each session takes, read from the NYSE calendar.
| Session | Clock | Liquidations | Real borrowers | Their dollars | All dollars | Dollar share |
|---|---|---|---|---|---|---|
| Regular session | 18.4% | 33 | 18 | $26,773 | $27,003 | 85.5% |
| Pre-market and after hours | 26.9% | 35 | 7 | $1,992 | $2,317 | 7.3% |
| Overnight | 22.7% | 49 | 5 | $1,795 | $2,239 | 7.1% |
| Holiday | 3.3% | 3 | 3 | $27 | $27 | 0.1% |
| Weekend | 28.7% | 0 | 0 | $0 | $0 | 0.0% |
| All sessions | 100.0% | 120 | 33 | $30,586 | $31,586 | 100.0% |
87 of 120 liquidations (73%) happened outside the regular session, but they carried 14.5% of the dollars. Most of them are the test address's small positions crossing the line at night.
Clock against dollars
The same table as two bars: how the clock divides, and how the dollars do.
Inside the day
The 33 liquidations of real borrowers, by when they landed.
| When | Count | Dollars | Share |
|---|---|---|---|
| First 90 minutes after a weekend or holiday | 3 | $17,924 | 58.6% |
| First 90 minutes after an ordinary night | 8 | $6,924 | 22.6% |
| Later in the regular session | 7 | $1,924 | 6.3% |
| While New York was closed | 15 | $3,814 | 12.5% |
| All organic liquidations | 33 | $30,586 | 100.0% |
The largest
The eight largest liquidations of real borrowers, each with its transaction on BNB Chain.
| New York time | Collateral | Repaid | Timing | Transaction |
|---|---|---|---|---|
| Mon 2026-07-27 10:35 | MUB | $8,959 | first 90 min after weekend/holiday | 0xe495b1b1...6a984f |
| Mon 2026-06-22 10:15 | SPCXB | $8,705 | first 90 min after weekend/holiday | 0xa840f16f...716383 |
| Tue 2026-07-07 09:55 | SPCXB | $4,450 | first 90 min | 0xd51b39fd...274ae5 |
| Tue 2026-07-28 09:39 | SPCXB | $2,129 | first 90 min | 0x09d3a24a...ccca41 |
| Mon 2026-08-10 18:53 | MRVLB | $1,079 | post | 0x26a1f065...79b155 |
| Wed 2026-07-22 13:55 | SPCXB | $918 | later | 0xe6ab795c...aa25bf |
| Wed 2026-08-19 00:46 | GLWB | $739 | overnight | 0xadcc99b1...45670b |
| Mon 2026-08-17 23:38 | QCOMB | $667 | overnight | 0x685d6eb1...3b6624 |
Method
- Every
Liquidateevent of Lista's Moolah contract (0x8f73b65b...) in a market that involves a bStock was read witheth_getLogs, from 18 June 2026 to 22 September 2026: 121 events. - Each was placed on the New York week by its block time, with the session read from the same NYSE calendar the contracts carry (holidays, early closes and daylight saving included).
- The repaid amount is valued in dollars with the Binance hourly close of the collateral at that hour; the loans themselves are dollar stablecoins.
- Rows are grouped by borrower: 33 from real borrowers, 87 from one test address (0x05e3a7a6...) that opened positions of $4 to $26 across many markets, and 1 in the one market where a bStock was the loan and not the collateral.
- "First 90 minutes" means 9:30 to 11:00 AM New York time on a trading day; "weekend" means Friday 8:00 PM to Sunday 8:00 PM.
What this does not show
- The sample is young: about three months of one lending market, 33 liquidations of real borrowers worth $30.6k. A handful of large ones carry most of the dollars.
- The 87 test positions mark when prices crossed the line, not losses. They are counted in the session table and drawn as hollow rings, and left out of every statement about real borrowers.
- No weekend liquidation in this sample is not a promise of none. The same liquidators did act on weekends in markets without stocks.
- Dollar values use an hourly close, so a single liquidation can be off by that hour's move.
- Timing says when a loan was liquidated, not when its price first crossed the line: a closed market defers the bill, it does not cancel it.
Backtest: the same closures, with and without a shield
697 closures of the twelve listed stocks replayed on a market with a 75% liquidation LTV. With the shield, the keeper repays before each close until the loan survives that stock's worst-1% gap at a health factor of 1.05.
| Starting LTV | Closures | Liquidated, no shield | Liquidated, shielded | Shield acted in | Average repay | Overnight: no shield / shielded | Weekend: no shield / shielded | Holiday: no shield / shielded |
|---|---|---|---|---|---|---|---|---|
| 60% | 697 | 0 | 0 | 0 of 697 | none | 0 / 0 of 554 | 0 / 0 of 121 | 0 / 0 of 22 |
| 65% | 697 | 0 | 0 | 24 of 697 | 2.0% of the debt | 0 / 0 of 554 | 0 / 0 of 121 | 0 / 0 of 22 |
| 70% | 697 | 5 | 2 | 652 of 697 | 2.6% of the debt | 5 / 2 of 554 | 0 / 0 of 121 | 0 / 0 of 22 |
| 72% | 697 | 27 | 2 | 697 of 697 | 5.1% of the debt | 24 / 2 of 554 | 2 / 0 of 121 | 1 / 0 of 22 |
Assumptions to read before quoting it
- The cushion is assumed large enough to fund every shield, and collateral is never sold.
- No venue minimum loan is applied.
- The data has no earnings flag, so earnings nights are tested against the smaller ordinary buffer. The liquidations that remain with the shield are moves beyond that buffer.
- The bStock's premium to its underlying and any oracle lag are ignored: the worst move is taken straight from the closure's gap and its lowest hourly print.
- One sample of about three months, twelve stocks. A buffer set at the worst 1% is, by construction, exceeded about one time in a hundred.
The gap each stock is sized for
The worst 1% of close-to-open down-gaps per stock and kind of closure, as set on the Session Oracle. Shields are sized against them and the feed's band starts from them.
| Stock | Overnight | Weekend | Holiday | Earnings |
|---|---|---|---|---|
| NVDA | 4.17% | 7.37% | 4.50% | 5.02% |
| SPY | 1.55% | 2.23% | 1.59% | no earnings |
| QQQ | 2.02% | 3.18% | 2.64% | no earnings |
| TSLA | 5.85% | 5.51% | 10.52% | 8.80% |
| AAPL | 2.55% | 2.76% | 3.99% | 7.42% |
| GOOGL | 2.47% | 3.87% | 3.08% | 7.70% |
| META | 3.18% | 2.99% | 3.18% | 24.46% |
| MSFT | 2.41% | 2.60% | 2.63% | 8.44% |
| AMZN | 3.00% | 3.91% | 3.26% | 11.39% |
| COIN | 7.73% | 9.46% | 4.73% | 22.00% |
| CRCL | 5.13% | 4.66% | 3.57% | 4.47% |
| MSTR | 7.70% | 12.09% | 6.05% | 6.61% |
Reproduce it
The data files are in the repository. Nothing below needs a key.
python research/check_figures.py- Recomputes every published figure from the raw data and reports how many match. The study, its scripts and its method are described in research/README.md.
pnpm backtest- Replays data/closure-windows.json through the same planner the keeper runs and rewrites data/backtest-lltv75.json, the table above.
pnpm --filter @ballast/web test- Checks that every figure on this page and on the front page is the one computed from data/liquidations-week.json and the calendar.
Files: data/liquidations-week.json, data/closure-windows.json, data/backtest-lltv75.json, config/bsc-mainnet.json, config/nyse-calendar.json. Their fields and sources are listed in data/README.md.