- beats top-q long (live) [OOS]: ret +413.5%, SR +0.82, maxDD -51.5%
- beats top-q long (live) [IS]: ret +25.3%, SR +0.26, maxDD -15.0%
- beats any long [OOS]: ret +209.6%, SR +0.62, maxDD -51.5%
- beats any long [IS]: ret +25.3%, SR +0.26, maxDD -15.0%
- misses any SHORT [OOS]: ret -52.6%, SR -0.35, maxDD -57.1%
- misses any SHORT [IS]: ret -37.6%, SR -0.16, maxDD -48.4%
- misses bottom-q SHORT [OOS]: ret -54.6%, SR -0.33, maxDD -59.0%
- misses bottom-q SHORT [IS]: ret -37.6%, SR -0.16, maxDD -48.4%
As-reported edge_hunt10 methodology: 60d sticky basket, surprise>0 (or negated for the miss side), 10bps cost proxy on churn, IS<2016<=OOS.
ALL DONE
Read
- Beat side (live book) works: top-quintile surprise basket OOS +413% / SR +0.82, resid vs EW market not shown but it is the validated sleeve.
- Miss side as SHORT: OOS -54.6% / SR -0.33, IS also negative. Long-the-miss-basket earns +52% OOS (both buckets carry positive market beta), so shorting misses bleeds. corrMkt -0.62, resid SR +0.38 but the raw -0.33 is the honest gross verdict.
- The Reddit asymmetry (misses drift 4.5x more) does NOT reproduce as a net shortable edge on this ADJ panel at 60d hold; the drift lives inside long-beta buckets on both sides. Parked; PEAD stays long-only top-quintile.