Max CAGR at Target Vol (SPX and 60/40) — expanding 60m/36m capped-simplex reject (qm-y8j.5)
Rejected qm-aaaz · 2026-08-18
Can a max-CAGR-at-vol Meta at SPX vol and at 60/40 vol beat its benchmark CAGR at matched vol via diversification return? Expanding annual January walk-forward 60m initial/36m per sleeve, capped-simplex 5%/20%/max10, sample cov ddof1, anti_overfit 0.002, monthly+static only, next_close 1-bar lag.
No. Both variants underperform on feasible OOS 2010-02-28..2026-07-31 (198m, 17 Jan signals, panel 2006-03-31..2026-07-31 245m) net 10bps next_close and across gross/2x: SPX delta -0.04367 (0.08053 vs 0.12420) and 6040 delta -0.03587 (0.06222 vs 0.09809). Vol-band 0.8-1.2x fails for SPX 0.746 while 6040 1.028 passes — overall infeasible. Regime 1/5 (SPX) and 0/5 (6040) fail 3/5 majority. Extra-lag Jan->Mar preserves failure (-0.04408 SPX, -0.03810 6040). Calibration 1991-1999 insufficient_history (0 eligible signals). Cost survival fails at 0/10/20bps. Not promoted.
Provenance
Research narrative
Question
Can max-CAGR-at-vol at fixed benchmark vol beat the benchmark CAGR via diversification return under expanding 60m/36m capped-simplex protocol?
Protocol
Expanding strictly_before January signal, 60m initial floor, 36m per sleeve, capped-simplex projection, sample cov, anti_overfit 0.002, next_close lag, 10bps base/20bps stress, vol-band 0.8-1.2x, calibration 1991-1999 vs untouched 2000-2026 preserved.
Result
Both variants fail OOS net win, vol-band (SPX), regime majority, extra-lag, cost 2x; calibration insufficient_history; not promoted.
Equity vs S&P (aligned, rebased to 1.0)
Aligned window, both rebased to 1.0 at start — Aligned window, both rebased to 1.0 at start for fair comparison · S&P dashed, experiment solid. X = Date, Y = Equity (rebased to 1.0).
Signal driver & thresholds (x-aligned)
Signal timeline (sparse triggers)
Triggers when overlay takes over base strategy
Stats
| CAGR | 0.00% | Sharpe | 0.00% |
|---|---|---|---|
| Sortino | 0.00% | maxDD | 0.00% |
Disposition
reject — not promoted to catalog; research preserved via experiments producer without fabricating catalog entry. Disposition is reject per G13 numeric gates.
Follow-on work
- No explicit follow-on; file new issue if needed
Errors & data gaps
Calibration 1991-1999 theoretical insufficient_history on ETF panel (first feasible signal 2010-01-31) flagged not counted as pass — panel starts 2006-03-31 so early calibration cannot be scored without synthetic histories Expanding walk-forward only on ETF panel 2006-2026 245m common; as_monthly modeled history not used for this evaluation but available for comparison; daily frequencies excluded monthly+static only Vol-band 0.8-1.2x fail for SPX variant (ratio 0.746 outside band, realized 0.1076 vs target 0.1441) while 6040 passes 1.028 — infeasible vol matching limits investability and shows narrow plateau OOS underperformance at matched vol net base 10bps: SPX CAGR delta -0.04367 (0.0805 vs 0.1242, Sharpe 0.775 vs 0.887) and 6040 delta -0.03587 (0.0622 vs 0.0980, Sharpe 0.706 vs 1.099) covering gross (0bps) and stress (20bps) — fee not rescuing failure, Sortino also below benchmark Throttling not applied: starts 32 workers 4 full locked defaults; liquidity drill drop_youngest_DBC not recomputed due to 600s kernel cost flagged insufficient_history; worst_month/worst_year representative -0.092/-0.114 not threshold gates
Follow-up
gh issue create --title "Follow-up Max CAGR at Target Vol SPX and 6040 expanding 60m36m capped" --body "Follow-up to qm-aaaz: Can a max-CAGR-at-vol Meta at SPX vol and at 60/40 vol beat its benchmark CAGR a" --label "priority:p2" --label "pipeline:research" --label "agent:ready"