Skip to content

Pennant strategy + detection registry

Append-only directory of every distinct strategy variant and every distinct detection-parameter set that has run under the Pennant research line. Both get permanent IDs the results matrix references:

  • Strategy IDs: PEN-<asset>-<NNN> (e.g. PEN-STOCK-001)
  • Detection IDs: PEN-DET-<label> (e.g. PEN-DET-baseline)

Detection IDs are short labels rather than dated cohort IDs so the matrix stays readable. Each detection ID below points to the canonical cohort parquets at cohorts/DET-*-<scan-date>/. If the detection code or universe changes and the same label is re-scanned, the cohort gets a new scan-date in its directory but the registry ID stays — the registry tracks parameter set, the cohort directory tracks parameter set + scan date.

Strategies

ID Name Asset Status Parent Created Spec Latest test Headline
PEN-STOCK-001 Phase 11b stock scaled-exit Stock active — 2026-05-11 PEN-STOCK-001.md PEN-TEST-004a–e Baseline cohort: $40,398 final from $10K (+304%, 7.50% CAGR, –38.1% max DD, Sharpe 0.543). V2 cohort: best Sharpe 0.617 with –21.4% max DD.

Detection variants

ID Params (pennant duration, flagpole duration) Status Created Cohort n events Latest test
PEN-DET-baseline pennant 5–15, flagpole 1–10 (production) active 2026-05-11 DET-BASELINE-2026-05-11 15,534 PEN-TEST-001a, 004a, 005
PEN-DET-v1 pennant 10–20, flagpole 1–5 active 2026-05-11 DET-V1-2026-05-11 5,155 PEN-TEST-001b, 004b
PEN-DET-v2 pennant 7–17, flagpole 1–5 active 2026-05-11 DET-V2-2026-05-11 7,428 PEN-TEST-002, 004c
PEN-DET-v3 pennant 6–17, flagpole 1–3 active 2026-05-11 DET-V3-2026-05-11 6,304 PEN-TEST-003a, 004d
PEN-DET-v4 pennant 6–17, flagpole 1–2 active 2026-05-11 DET-V4-2026-05-11 4,637 PEN-TEST-003b, 004e

All five detection variants share these other parameters: max_retrace_pct = 0.382, min_magnitude_pct = 12.0, min_atr_multiple = 4.0, volume_ratio_min = 1.5, trend_filter on (EMA-55 ≥ 10-day prior). Date range scanned: 2007-02-15 → 2026-05-08.

Conventions

  • Strategy ID = PEN-<asset>-<NNN>. Asset codes so far: STOCK, OPT (reserved for options). Sequential within asset.
  • Detection ID = PEN-DET-<label> where the label is a short lower-case word ("baseline", "v1", "v2", ...). Avoid embedded dates — the cohort directory carries the scan date.
  • Status = active / superseded / archived.
  • Parent = ID this entry derives from. Empty for fresh designs.
  • Cohort = relative link into cohorts/<dir>/ (detection variants only).

Adding a variant

  1. New strategy: pick the next unused PEN-<asset>-<NNN>; write strategies/<id>.md with full mechanics; append to the Strategies table above.
  2. New detection variant: pick the next unused PEN-DET-<label>; add a row to the Detection-variants table; the cohort itself lives under cohorts/DET-<UPPER>-<scan-date>/ with its own README.
  3. Reserve a PEN-TEST-NNN in ../../../ledger.md for the first run.
  4. After the run completes, append the result row(s) to ../results_matrix.csv via infra/update_matrix.py::append_row().

Lineage notes (historical, pre-Pennant directory)

Strategies that predate the Pennant research line (Phase 1 stock scaled-exit, Phase 2 options ITM 0.95× × 21–35 DTE, Strategy C 1.15× call-only audit anchor, etc.) are not in this registry — they live in build_v1/ and are catalogued in historical_recovery.md. If any historical strategy is re-run under the new harness, it gets a fresh PEN-<asset>-<NNN> at that point.