quantmodel — BBDF deeper 2nd pass (2024/2026 residuals; SPY-vs-cash hist-alloc)

Objective

BBDF deeper 2nd pass (2024/2026 residuals; SPY-vs-cash hist-alloc).

Notes: BBDF abs_1m_rescue: corr 0.917→0.923 (PASS CAGRΔ −0.69%). 2026 yearly 0.71→0.86; Feb-2026 UUP recovered. SPY dual-1m rejected (~0.80). Findings: black-box-df-corr-t127.md

Approach

Reuses existing report research/findings/black-box-df-corr-t127.md — tables/metrics preserved verbatim via source_report pointer. Body below summarizes grounded evidence from that report and adjacent artifacts.

Files / code / data changed

Associated commits (inferred from log grep t127):

4c52131e research: t127 BBDF abs-1m rescue (corr 0.917→0.923) (#114)
4c52131e research: t127 BBDF abs-1m rescue (corr 0.917→0.923) (#114)
Source report: research/findings/black-box-df-corr-t127.md
Task-history entry: research/findings/task-history.json#t127

Results

Key table from source report (preserved):

variant	monthly corr	hist-alloc	notes
t123 baseline (weighted 13612 + stitch)	0.917	4/12 MAE~0.20	prior shipped
vol_lookback=84	0.917	4/12	+0.0006 only — reject
SPY dual 1m / abs-1m gate	~0.80	5/12	fixes Jun-2026 cash; destroys ITD corr
never_spy / always_spy / rank sleeves	≤0.88	mixed	reject
SHY hurdle / abs-only def / EWMA vol	≤0.91	mixed	reject
border_or_abs (±1% below BIL if abs>0)	0.921	3/12	close 2nd
bil_or_abs_1m rescue	0.923	3/12 MAE~0.22	shipped

Conclusions and metrics are in research/findings/black-box-df-corr-t127.md; see detail page for preserved tables/metrics.

Conclusions

BBDF abs_1m_rescue: corr 0.917→0.923 (PASS CAGRΔ −0.69%). 2026 yearly 0.71→0.86; Feb-2026 UUP recovered. SPY dual-1m rejected (~0.80). Findings: black-box-df-corr-t127.md

Problems / follow-ups

See source report or PR discussion for follow-ups. No unsupported follow-ups fabricated.