TWT liquidation map

• Longs: 93% of $182.8K, with price at their levels 24% of the time. Shorts: 93% of $6.1K, with price at their levels 15% of the time.
• Binance publishes only part of its liquidations, so its dollars are a floor.
• It scores the map; the lines on the chart are picked from a finer version of it and are tested on their own. See how the lines are being tested.
• Provisional model: no registered test covers this market yet.

Binance publishes only part of its liquidations, so its dollars are a floor. A daily record, not a test. See how the rails are being tested.

• Every minute the published map was ready is scored, the UTC day after, at 00:20 UTC. The window is the last 28 complete UTC days, 2026-09-19 to 2026-10-16 UTC.
• At the map's levels: in that minute price traded through a level where the map had long dollars (price fell to it) or short dollars (price rose to it), however few. A long liquidation counts on the long side, a short liquidation on the short side, wherever it printed.
• Time at the map's levels: the share of scored minutes in which that happened, each side weighted by its liquidation dollars. It is what the share at the levels would be if liquidations came at random minutes. They do not: they cluster in fast moves, and fast moves reach more levels, so this baseline flatters the map, most of all one with levels close to the price.
• The share at the levels divided by the time there is not printed: a map in reach half the time reads 2x at most however good it is, and fast moves lift any map.
• The daily score of the drawn levels (Binance, from 10 Oct 2026): every minute a one-minute bar reached a drawn level, the same side's reported liquidations that minute and the next two, against minutes of the same market and side that reached no level and whose move toward that side fell in the same one of 42 fixed size bands. Markets weigh evenly. It leads the record once it has 1,000 touches over three days; until then the one-off study does. Bybit and Hyperliquid wait for their registered tests.
• The study of the drawn levels (Binance, 10 to 13 Oct 2026): for every minute a one-minute bar reached a drawn level, the same side's reported liquidations that minute and the next two, against minutes of the same market and side whose move toward that side fell in the same one of 200 size bands but reached no level. A standard map (10x, 25x, 50x and 100x tiers from Binance open interest, the same number of levels) was scored the same way. It ran once, on data already recorded; it is not a registered test.
• Reported per map $: liquidations reported in those minutes divided by the dollars the map had at the levels price traded through. At 1.00x the venue reported as much as the map had.
• Binance throttles its liquidation feed, so its dollars are a floor; Bybit's feed is complete. Hyperliquid is not scored yet, and Bybit markets outside the registered Bybit test are held out.
• A daily record, not a test. It describes what happened each day; the rails' registered forward test is run separately.
• Model: v2-c643da479075-provisional (mid class), 12 days. Last computed 2026-10-17 00:20 UTC.

Newest first. Reached: minutes in which price reached the map's levels on that side. Map $: the dollars at the levels price traded through. Reported $: the liquidations reported on that side in those minutes.