**Good news:** The existing NPZ is already on disk at `DB/fast_cache_akela_shortlist_1m_30d.npz` (8.9 MB). No data fetch is needed to get started.

Here is the 6-phase execution plan:

---

## Phase 1 — Verify prior results on existing NPZ
Run all 3 configs against the existing `DB/fast_cache_akela_shortlist_1m_30d.npz`:
```bash
cd obw_platform
python3 backtester_dual_long_short_fast_pack_v2.py \
  --cfg configs/h4_freedommoney_ratio_best_v2.yaml \
  --npz ../DB/fast_cache_akela_shortlist_1m_30d.npz \
  --symbol "FREEDOMMONEY/USDT:USDT"
```
Accept if ratio is within 5% of 12.69. Captures the JSON output, computes ratio.

## Phase 2 — Fetch fresh FREEDOMMONEY data from BingX (5000 × 1m bars)
```bash
python3 fetch_build_cache_v17.py \
  -i obw_platform/universe/universe_short_top15_1m.txt \
  -t 1m --back-bars 5000 \
  -o DB/combined_cache_1m_freedommoney_5000b.db \
  --exchange bingx --ccxt-symbol-format usdtm --fresh
```
Falls back gracefully if FREEDOMMONEY not on BingX.

## Phase 3 — Build NPZ from SQLite
```bash
python3 obw_platform/build_fast_ohlcv_npz_from_db_v2.py \
  --db DB/combined_cache_1m_freedommoney_5000b.db \
  --symbol "FREEDOMMONEY/USDT:USDT" \
  --out DB/fast_cache_freedommoney_1m_5000b.npz --ohlc-mode native
```

## Phase 4 — Re-evaluate on enlarged window
Same 3-config runs on the new NPZ, compare to Phase 1.

## Phase 5 — Tune if ratio degrades below 8.0
Target: `tpPercent`, `subSellTPPercent`, `callbackPercent`, `maxLongInvestPct`, `maxShortInvestPct`. Reject if `margin_calls > 0`, `|unrealized| > 60`, or `MTM_DD < -20%`.

## Phase 6 — Deliverables
Update `AGENT_STATE.md`, append to `EXPERIMENT_LEDGER.csv`, save best config, write ranked CSV to `_reports/freedommoney/`, pack snapshot.

---

Shall I proceed with execution starting at Phase 1?