📦 MOO→11:00 — все версии и стратегии

Полная эволюция MOO entry → 11:00 exit asymmetric strategies (v1 → v5 + engine v2 plugin refactor + v2.5)
1 LIVE (v1 Tier 1) 3 research-done (v2.5/g3a/...) 3 paper plugins (engine v2) 2 superseded (v2/v4) 1 BLOCKED (v5 layered) 10 итераций

🧬 Эволюция версий — что было, что осталось

v1 (Tier 1 only)
LIVE в M14 production · sub5_pump

Originally один setup: gap≥10% × px<5 × MOO→11:00 SHORT. Pre-period 2023 PF 2.17 ✅. Это единственный edge переживший все аудиты.

в M14 deploy PF 2.17

v2
SUPERSEDED · топ эдж rejected

Топ-1 эдж gap_le-10 × px_20_50 LONG +1.69% REJECTED — pre-PF 0.64 (2024-born). Все rvol_*x также REJECTED (per-trade Sh inflation от catalyst-day concentration).

edge rejected

v3 (5 invest × 3 audit)
FINAL strategy reference · paper hedge stack

5 параллельных расследований × 3 аудитора. Tier 1 (ALL 5 PASS) = v1 same. HEDGE STACK: LONG gap_-10_-5 × pmd_-5_-2 + SHORT gap_ge10 × px_lt5 → daily Sh 8.01, MaxDD -3.9%, 4/4 quarters +.

Sh 8.01 DD -3.9%

v4 LAYERS
SUPERSEDED · most failed permutation

151 layers passed (32 indicator + 112 cross-asset + 7 positioning). TOP indicator: MACD hist > 0 ΔSh +3.07. TOP cross-asset: DXY 5d ≤-0.34% ΔSh +4.63, VIX [34,45]. v5 audit revealed permutation failure.

layers exposed

v5 JOINT (per-trade Sh 14)
BLOCKED — illusion permutation 0/12 robust

Joint stacks Sh ≥9. Top 3-layer: gap_ge10×c5_le-10 + nz_gt200 + DXY_5d_LO + pos52_b=lo20 → Sh 14.99. H4 MONDAY×gap_ge10 Sh 13.81 N=191. AUDIT: 0/12 stacks robust permutation — within-ticker shuffle preserves Sh 14 → edge sits в asymmetric payoff DISTRIBUTION SHAPE, не в layer specificity. PAPER ONLY.

illusion disabled in scorer

Engine v2 plugin refactor
Apr 27 · 4 plugins · 87/87 tests

Refactored monolithic v1 → plugin engine: 4 plugins (organic/reports × pump/dump), 1%SL/7%TP/11:00 hard exit, M14-style 6-layer risk gates, full audit pipeline. organic_pump enabled, остальные 3 = paper-only.

23/23 plan tasks 87/87 tests

v2.5 (4-stream research)
Apr 27 EOD · pre-deploy gate

A1 exit 10:00 (79% TPs до 10:00) + A6 ORB filter (skip 33%) + A7 sweep bonus + B universe ADV 500K+. Stream D: g3a SHORT (ATR-stretch). Stream C: first5_directional (orthogonal).

×2-3 lift cache extension required

v6 (backlog)
Not started

ADV 1M+ universe для N≥200/fold. Fix DIRTY features (sector snapshot, H4 eoq future-leak, halt_today, VIX_VIXY proxy decay-ETF). Quarterly re-audit cadence.

backlog post v2.5 deploy

📋 Каталог setup'ов — все стратегии MOO→11:00

# Version Setup Direction Trigger (rule) Sharpe PF / DD Status Verdict
1 v1 / v3 Tier 1 sub5_pump SHORT gap≥10 × px<5 — (PF 2.17) DD -3.9% LIVE M14 ✅ Single proven edge — pre-period verified
2 v3 hedge stack LONG gap-down dump bounce LONG gap_-10_-5 × pmd_-5_-2 8.01 daily DD -3.9% paper ✅ Combined with #1 = hedge stack
3 v3 Tier 2 c5 fade SHORT SHORT gap≥10 × c5_-10_-5 ~3.99→7.06 (+v4) PF 3.13 paper 🟡 v4 layer base — without layers WF unstable
4 v3 Tier 2 extreme c5 fade SHORT gap≥10 × c5_le-10 ~stable 3y paper 🟡 mean_3m 2.25× baseline UP
5 v3 Tier 2 day-high fade SHORT gap≥10 × dh_-20_-10 ~stable 3y paper 🟡 distance-from-high signal
6 v2 mid-cap LONG bounce LONG gap_le-10 × px_20_50 +1.69% mean pre-PF 0.64 REJECTED ❌ 2024-born edge, no pre-period evidence
7 v4 indicator + MACD hist > 0 layer SHORT gap≥10 × c5_-10_-5 × MACD_pos 7.06 (ΔSh +3.07) PF 3.13 audit failed 🔴 v5 perm test 0/12 robust
8 v4 cross-asset + DXY 5d weak SHORT gap≥10 × DXY_5d ≤-0.34% +4.63 ΔSh audit failed 🔴 same — perm illusion
9 v4 cross-asset + VIX panic regime SHORT gap≥10 × VIX [34,45] +4.16 ΔSh audit failed 🔴 same — perm illusion
10 v5 joint 3-layer mega stack SHORT gap≥10 × c5_le-10 × nz_gt200 × DXY_5d_LO × pos52_b=lo20 14.99 per-trade cap 18.5% BLOCKED 🔴 illusion — within-ticker shuffle preserves Sh
11 v5 H4 seasonality MONDAY tilt SHORT MONDAY × gap≥10 × px<5 13.81 N=191 WF 9/9 BLOCKED 🔴 1 cut 2025-09 used; needs 2024-09 cut
12 v2.5 / Stream A v1 + A1 exit 10:00 SHORT v1 + force-close 10:00 +18% perf research ✅ 79% TPs до 10:00 — code-only
13 v2.5 / Stream A v1 + A6 ORB filter SHORT v1 + skip if ORB break PM-high +0.49 Sh research ✅ skip 33% trades, quality up
14 v2.5 / Stream A v1 + A7 sweep bonus SHORT v1 + size 1.25× if sweep&reverse +5% research ✅ liquidity-sweep institutional flow
15 v2.5 / Stream B universe expansion SHORT v1 + ADV 500K+ (от 2M) WR 52→53.7% ×2 cap research ✅ counterintuitive — small fades cleaner
16 v2.5 / Stream D g3a SHORT (ATR-stretch) SHORT open ≥ prev+2×ATR(14), no ADR ~4.5 daily all 4y + research ✅ vol-normalized vs raw gap%
17 v2.5 / Stream C first5_directional L+S flat-open + first5 move >2% ~4.0 daily WR 82.3% research ✅ orthogonal — institutional flow signal
18 Engine plugin organic_pump SHORT v1 setup в plugin form ~1.5-3 daily PF 2.17 live ✅ enabled в strategy.yaml
19 Engine plugin org_dump (LONG bounce) LONG gap ≤-10% × px<5 × bounce TBD est PF 1.8 paper 🟡 needs audit driver run
20 Engine plugin reports_pump (earnings fade) SHORT earnings AMC + next-day pump >+10% TBD est PF 1.5 paper 🟡 needs audit driver run
21 Engine plugin reports_dump (earnings bounce) LONG earnings AMC + next-day dump <-10% TBD est PF 1.5 paper 🟡 needs audit driver run

📊 Сравнение версий — charts

Sharpe эволюция версий
v1 → v2.5 + audit verdict (real vs reported)
Status distribution (21 setup)
Где каждый setup сейчас
Reported vs verified Sharpe — audit gap
Per-trade Sh inflation vs daily honest Sh (v4/v5 layered example)
Direction breakdown — где edge живёт
Из 21 setup'а — SHORT доминирует (asymmetric pump-fade)

🧠 Ключевые уроки эволюции v1→v5

📌 Что выжило все аудиты

  1. v1 sub5_pump — единственный edge с pre-period 2023 PF 2.17. Все остальное либо 2024-born либо permutation-illusion.
  2. v3 hedge stack — combined LONG+SHORT даёт Sh 8.01, MaxDD -3.9% (5× lower DD vs v1 alone).
  3. v2.5 A1 exit 10:00 — code-only insight survived all sanity checks (79% TPs данные).

⚰️ Что НЕ выжило аудиты

  1. v2 mid-cap LONG — pre-PF 0.64, born в 2024 (нет structural edge).
  2. v4 layered — layers ΔSh +3-5 illusion (within-ticker shuffle preserves Sh = noise).
  3. v5 joint Sh 14 — per-trade inflation: edge sits в payoff distribution shape, не layer specificity. Bootstrap CI lower 1.3.
  4. rvol_*x indicators — catalyst-day concentration inflates per-trade Sh, daily Sh much lower.
⚠️ Recurring lesson
Per-trade Sharpe inflates 2-3× vs daily Sh. Bootstrap по dates ≠ trades. WF catches what per-trade misses. Never deploy on per-trade Sh alone — always demand daily Sh + permutation test + WF stability + pre-period evidence.