quantmodel — Optimum3 deeper 2nd pass (2017 sleeve mix; Jun/Aug hist misses)

Objective

Optimum3 deeper 2nd pass (2017 sleeve mix; Jun/Aug hist misses).

Notes: Near-tie corr_epsilon=0.009 + mom_edge=0.001: official 2015+ corr 0.907→0.913 CAGRΔ −1.51% PASS; hist-alloc 10/12→11/12 (Aug-2025 EWJ fixed). Jun-2026 PDBC outside top-6 + 2017~0.60 residual. Soft expand ~0.92 corr rejected (hist).

Approach

Reuses existing report research/findings/optimum3-corr-t128.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 t128):

a396cdb3 research: t128 Optimum3 near-tie corr 0.907→0.913 (#115)
a396cdb3 research: t128 Optimum3 near-tie corr 0.907→0.913 (#115)
Source report: research/findings/optimum3-corr-t128.md
Task-history entry: research/findings/task-history.json#t128

Results

Key table from source report (preserved):

variant	monthly corr	hist-alloc	notes
t122 baseline (stitch + thin top-3)	0.907	10/12	prior shipped
top_k=7 / corr_days=126 grids	≤0.91	≤7/12	rejected (breaks live months)
soft expand ratio~0.8	~0.92	6–8/12	higher corr, worse hist + 2017
corr_epsilon=0.009 + mom_edge=0.001	0.913	11/12	shipped

Conclusions and metrics are in research/findings/optimum3-corr-t128.md; see detail page for preserved tables/metrics.

Conclusions

Near-tie corr_epsilon=0.009 + mom_edge=0.001: official 2015+ corr 0.907→0.913 CAGRΔ −1.51% PASS; hist-alloc 10/12→11/12 (Aug-2025 EWJ fixed). Jun-2026 PDBC outside top-6 + 2017~0.60 residual. Soft expand ~0.92 corr rejected (hist).

Problems / follow-ups

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