Hunter KillerCreate free accountCascade Autopsy #15: BTC, 2026-08-28
hunterkiller.io · All numbers from our multi-venue tape · Graded, not estimated
Summary
| Field | Value |
|---|---|
| Date/Time | 2026-08-28 14:01 UTC |
| Dominant Symbol | BTCUSDT |
| Total Liquidated | $212.5M |
| Duration | 5m 0s |
| Raw Events | 2,013 (independent re-read) |
| Venues | binance, bitfinex, bitmex, bybit, gateio, hyperliquid, kraken_fut, okx |
| Direction | predominantly LONG liquidations (79% of events), price fell |
Source: trading.db/liquidation_events, independent re-read confirms headline numbers.
Top Symbols by USD Liquidated
| Rank | Symbol | USD | Events | Share |
|---|---|---|---|---|
| 1 | BTCUSDT | $204.8M | 591 | 96.4% |
| 2 | ETHUSDT | $2.3M | 262 | 1.1% |
| 3 | XRPUSDT | $1.6M | 194 | 0.8% |
| 4 | SOLUSDT | $1.3M | 233 | 0.6% |
| 5 | XAUUSDT | $733,799 | 112 | 0.3% |
Timeline (UTC)
| Minute | Events | USD | Top Symbol | Venues |
|---|---|---|---|---|
| 14:00 | 286 | $2.9M | ETHUSDT | binance, bybit, gateio, okx |
| 14:01 | 77 | $1.2M | BTCUSDT | binance, bybit, gateio, okx |
| 14:02 | 893 | $9.2M | BTCUSDT | binance, bybit, gateio, hyperliquid, kraken_fut, okx |
| 14:03 | 39 | $76,595 | ENAUSDT | binance, bybit, gateio, okx |
| 14:04 | 520 | $198.0M | BTCUSDT | binance, bitfinex, bitmex, bybit, gateio, hyperliquid, okx |
| 14:05 | 468 | $3.9M | BTCUSDT | binance, bybit, gateio, hyperliquid, okx |
| 14:06 | 469 | $2.9M | BTCUSDT | binance, bybit, dydx, gateio, hyperliquid, kraken_fut, okx |
| 14:07 | 134 | $1.7M | SOLUSDT | binance, bybit, gateio, hyperliquid, kraken_fut, okx |
What the Magnet Field Showed BEFORE the Cascade
Snapshot: 2026-08-28T13:58:07Z, 3 minutes before cascade start. Anchor price at snapshot: $79,401.60
Model calibration (per-symbol): - precision@N (OOS) = 50.0% (n=348, CI 95% lower bound: 44.8%) | walk-forward window: 720h Calibrated: 2026-08-28T13:59:34Z
What this means: The predictive-liquidation model places levels in the top N ranked zones that matched real liquidation clusters in 50% of OOS evaluation windows (n=348, CI 95% lower bound: 44.8%). This is the measured zone-accuracy, not a forecast of whether a cascade will start.
Level assessment: 12 REACHED / 15 MISSED (both shown; misses are never hidden)
| Leverage | Side | Predicted Price | Status | Mass (relative) |
|---|---|---|---|---|
| None× | long | $78,746.54 | REACHED | 206612056 |
| None× | long | $78,389.23 | REACHED | 159722934 |
| None× | long | $78,031.92 | REACHED | 123828285 |
| None× | short | $80,175.77 | REACHED | 109189952 |
| None× | short | $79,937.56 | REACHED | 104510038 |
| None× | long | $79,222.95 | REACHED | 104270123 |
| None× | short | $79,580.25 | REACHED | 95358639 |
| 50× | long | $78,607.58 | REACHED | 61762693 |
| 75× | long | $79,242.80 | REACHED | 21701062 |
| 100× | long | $79,560.40 | REACHED | 9264729 |
| 100× | long | $79,242.80 | REACHED | 8739845 |
| 100× | short | $80,195.62 | REACHED | 8209827 |
| None× | long | $77,793.72 | MISSED | 98454647 |
| 25× | long | $77,019.55 | MISSED | 94699213 |
| 10× | long | $72,255.46 | MISSED | 84482166 |
| None× | short | $80,533.07 | MISSED | 84393297 |
| 25× | short | $82,736.47 | MISSED | 80702920 |
| None× | long | $77,555.51 | MISSED | 75510056 |
| 10× | short | $87,500.56 | MISSED | 63556927 |
| 50× | short | $81,148.44 | MISSED | 54848644 |
| None× | short | $81,009.48 | MISSED | 54448705 |
| 5× | long | $64,315.30 | MISSED | 38170379 |
| None× | short | $81,485.89 | MISSED | 25199047 |
| 5× | short | $95,440.72 | MISSED | 23238783 |
| 3× | long | $53,834.28 | MISSED | 22091931 |
| 75× | short | $80,513.22 | MISSED | 19241729 |
| 3× | short | $106,239.34 | MISSED | 10461067 |
Caveats: These are predicted liquidation cluster ZONES from our walk-forward calibrated model, not leaked order-book data, not single prices but band estimates. REACHED = actual cascade price range came within 0.5% of the predicted level. MISSED levels are shown at equal weight, both reaches and misses are counted, never cherry-picked. The measured calibration (with n= and 95% CI lower bound) is stated in the Model calibration section above.
Venue Breakdown
Our tape covers multiple venues independently, cross-venue coverage is the moat.
Venues active: binance, bitfinex, bitmex, bybit, gateio, hyperliquid, kraken_fut, okx
(Per-venue USD breakdown available via the live API at /market/cascade_autopsy. The tape covers all venues simultaneously, the cascade was multi-exchange.)
Contagion Propagation
Historical co-fire correlation, NOT causal evidence. Shared macro shocks produce simultaneous liquidations across symbols. Minimum 3 co-fire samples required to appear.
| Responder | Co-fire n | Shock Beta | Same-bar Move | Next-bar Move |
|---|---|---|---|---|
| XRPUSDT | 5 | 1.90 | -0.06% | 0.16% |
| DOGEUSDT | 5 | 1.89 | 0.04% | -0.14% |
| SUIUSDT | 5 | 1.78 | -0.14% | 0.20% |
| ADAUSDT | 5 | 1.72 | -0.29% | -0.18% |
| SOLUSDT | 5 | 1.70 | 0.28% | 0.10% |
Interpretation: symbols with shock_beta > 1.0 historically moved MORE than the source symbol during similar cascade events. This is a historical observation (n shown per pair), not a forecast of future co-movement.
Accuracy Metadata
- Magnet model: OOS precision@N = 50.0% (n=348, CI 95% lower bound: 44.8%) | walk-forward window: 720h | calibrated: 2026-08-28T13:59:34Z
- Tape: 2,013 raw liquidation events from binance, bitfinex, bitmex, bybit, gateio, hyperliquid, kraken_fut, okx
- Capture span: ~52 days of rolling multi-venue liq tape
- Independent verify: headline USD/count confirmed by independent DB re-read (PASS)
Your Take
Is your bag on this list? BTCUSDT, ETHUSDT, XRPUSDT were the symbols hit hardest in this cascade. Were you watching the long level at $78,746.54? It graded REACHED on our tape. Reply with the symbol and level you had on - reaches and misses both count here, and telling us where the model was wrong makes the next autopsy better.
Share this autopsy: hunterkiller.io/go/liq?c=autopsy-share&r=TG-CHANNEL
Autopsy published directly by the hunterkiller.io automated pipeline. Data: 2,013 events from our multi-venue tape (real captured data, not model-estimated). Accuracy model is walk-forward calibrated with precision@N and OOS score.
marketing_assets/live/autocard-card-autopsy-2026-08-28t1401-btc.png