SQUEEZE RADAR — predicted vs real, live

Every price burst on BTC/ETH/SOL gets a live probability of completing a ±0.16% move, scored at +20/60/120/180/240/300s by a fixed logistic model. Every prediction and its outcome is logged automatically — this table recomputes from that log. No selection, no edits. Columns = predicted-probability buckets; each cell shows average predicted → realized frequency.

checkpoint0-20%20-40%40-60%60-80%80-100%
+20s17% → 15% n=44631% → 31% n=632950% → 51% n=808169% → 71% n=496691% → 90% n=4324
+60s16% → 16% n=165131% → 33% n=696249% → 54% n=627869% → 72% n=447090% → 90% n=3333
+120s15% → 17% n=318930% → 35% n=699349% → 55% n=516369% → 74% n=329989% → 89% n=1914
+180s14% → 17% n=467329% → 36% n=658849% → 57% n=412169% → 75% n=225588% → 89% n=1028
+240s13% → 15% n=597829% → 37% n=592849% → 58% n=296069% → 73% n=146287% → 90% n=449
+300s11% → 15% n=750129% → 37% n=482649% → 59% n=204768% → 74% n=81687% → 85% n=199

Bursts resolved so far: 24352 · completed: 13792 (57%) · updated 2026-08-10 16:30 UTC (10-min cache)

Confirmed icebergs (level absorbed ≥1× its displayed size, multi-tranche tracking): 105926 resolved — held 6 · broken 29378 · pulled 4755 · consumed 71787 · median executed/displayed 2.1× (survival stats only — no directional claim. Frey & Sandås (QJF 2017): a heavily-executed iceberg = demand already met → LESS follow-through; our 527k-episode base-rate study agrees: P(hold) does not rise with the multiple)

Honest notes: scoring is live since 2026-06-10 — early samples lean on a high-volatility regime; the mix broadens as days accumulate. The model predicts WHEN/WHETHER a move completes, never direction (we measured direction predictability at these horizons: it is not there). Not financial advice. ← Squeeze Radar