← all research

Input: data/baselines/eqmom_live_v1.csv (2006-08-14..2026-08-12, already net @10bps; config resid+sector+book+credit). Split 2016-01-01. Overlay turnover charged on top.

Condition [2] proxy-vs-book beta (C7)

| beta proxy | corr(beta_book, beta_proxy) | mean beta_proxy | mean |diff| | |---|---|---|---| | universe beta (old) | -0.372 | 0.946 | 0.379 | | selected beta (fix) | +0.684 | 1.141 | 0.204 | | book (reference) | +1.000 | 1.103 | 0.000 |

Proxy PASS if corr >= 0.70. The production fix uses the selected basket.

extra overlay cost = 10bps one-way

VariantOOS RetOOS SROOS maxDDIS SRboot-p vs D0
D0 baseline+1176.2%1.17-34.1%0.51-
D1 SPY>200d+685.5%1.14-21.5%0.610.9072
C7 hedge (universe beta (old))+934.3%1.22-25.7%0.520.8878
C7 hedge (selected beta (fix))+932.4%1.22-25.7%0.520.8886
C7 hedge (book beta)+934.3%1.22-25.7%0.560.8878
D1+C7+661.0%1.16-18.8%0.550.9216

extra overlay cost = 25bps one-way

VariantOOS RetOOS SROOS maxDDIS SRboot-p vs D0
D0 baseline+1176.2%1.17-34.1%0.51-
D1 SPY>200d+624.4%1.10-21.9%0.550.9342
C7 hedge (universe beta (old))+865.3%1.18-26.0%0.470.9372
C7 hedge (selected beta (fix))+863.5%1.18-26.0%0.470.9386
C7 hedge (book beta)+865.3%1.18-26.0%0.520.9372
D1+C7+564.8%1.09-20.7%0.450.9538

Read / vote rule

  • Adopt an overlay only if OOS p<0.05 vs D0 OR a structural DD cut >=10pp with SR within -0.10, AND the edge survives net@25bps.
  • D1 all-flat is the documented candidate; C7 book-beta is a best-case ceiling; the selected-beta row is what would actually ship.