Overview
What's now
CURRENT CHAMPION
V6 · V66 Cooldown(4,48)
+70,576% compound · +245% min α · 4/4 beats B&H · MDD 27.7%
GRU 2x128 ensemble · 5 seeds × 4 folds · 36 features · tb_vol_10_7 labeling
Last roundR141MTF filter · REJECT
Rounds since V66 beaten21paradigm exhaustion
Healthy bots5/5all running
Live capital (real)€109.53as of 2026-05-26 18:28 UTC · 2 invested · 3 cash
R&D signal: 21 rounds without beating V66 baseline. Refinement is exhausted — shift focus to NEW paradigms. R117 MACD_div (+179K% compound, decorrelated from V66) is the live lead. See pipeline for next experiment (R120 portfolio).
Production bots live capital · trades · state · refreshed at commit time (2026-05-26 18:28 UTC)
V2
GRU Ensemble (always-invested)
capital (live)€22.21
trades2
stateCASH
backtest min α—
V3
GRU + DailyRSI>80 + Trail 10%
capital (live)€23.00
trades0
stateCASH
backtest min α—
V4
GRU 3-Regime adaptive (V5.4 Robust-5)
capital (live)€23.31
trades0
stateINVESTED
backtest min α+117%
V5
V115_cmp (GRU + peak_drop + ratchet + regime cooldown)
capital (live)€20.01
trades0
stateINVESTED
backtest min α**+238%**
V6
V66 Cooldown(4,48) (GRU defensive uniform)
capital (live)€21.00
trades0
stateCASH
backtest min α**+245%**
Latest experimental results most recent 3 rounds
R135
mean-reversion
REJECT
2026-05-26
Production-grade bar-by-bar portfolio simulator
| fold | fold_name | method | v66_pnl | macd_pnl | bh_pnl | portfolio_pnl | alpha | beats_bh | n_days | avg_w_v66 |
|---|---|---|---|---|---|---|---|---|---|---|
| 0 | F0_Bear | fixed 25/75 (no clf) | 33.7 | -44.6 | -11.1 | -28.6 | -17.6 | False | 974 | 0.25 |
| 1 | F1_Bull | classifier-driven soft+25% floor | 28.9 | -9.8 | 58.0 | 22.2 | -35.8 | False | 668 | 0.36 |
| 2 | F2_Recovery | classifier-driven soft+25% floor | -1.1 | -32.6 | 205.5 | -22.2 | -227.7 | False | 670 | 0.385 |
| 3 | F3_Lateral | classifier-driven soft+25% floor | -15.8 | -67.8 | 24.9 | -41.3 | -66.1 | False | 633 | 0.559 |
R134
labeling
REJECT
2026-05-26
V66-zone forward label + purged walk-forward (anti-leak)
| fold | fold_name | train_size | test_size | train_acc | test_acc | recall_v66 | precision_v66 | f1_v66 | n_v66_true | confusion_matrix |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | F1_Bull | 915 | 668 | 1.0 | 0.754 | 0.146 | 0.054 | 0.079 | 48 | [[497, 123], [41, 7]] |
| 2 | F2_Recovery | 1583 | 670 | 0.997 | 0.422 | 0.159 | 0.317 | 0.212 | 327 | [[231, 112], [275, 52]] |
| 3 | F3_Lateral | 2253 | 633 | 0.974 | 0.452 | 0.488 | 0.409 | 0.445 | 285 | [[147, 201], [146, 139]] |
R132
hybrid-sma
REJECT
2026-05-26
Vol-shock proxy features for regime classifier (sentiment proxy without LLM)
| variant | fold | accuracy | recall_v66 | precision_v66 |
|---|---|---|---|---|
| baseline_r121b | 1 | 0.9925149700598802 | 0.9611650485436893 | 0.99 |
| baseline_r121b | 2 | 0.9791044776119403 | 1.0 | 0.8333333333333334 |
| baseline_r121b | 3 | 0.8643578643578643 | 1.0 | 0.12962962962962962 |
| shock_only | 1 | 0.7904191616766467 | 0.02912621359223301 | 0.06976744186046512 |
| shock_only | 2 | 0.7701492537313432 | 0.014285714285714285 | 0.011627906976744186 |
| shock_only | 3 | 0.976911976911977 | 0.0 | 0.0 |
| full_r132 | 1 | 0.9910179640718563 | 0.9514563106796117 | 0.98989898989899 |
| full_r132 | 2 | 0.9791044776119403 | 1.0 | 0.8333333333333334 |
| full_r132 | 3 | 0.8643578643578643 | 1.0 | 0.12962962962962962 |
Navigate
→ R&D History
All 40 rounds, sortable, filterable by paradigm and verdict
→ Production Strategies
V2-V6 deep dive: edge thesis, weaknesses, comparisons
→ Discarded Paths
Graveyard with failure mode + re-explore triggers
→ R&D Pipeline
Kanban: queued / in-flight / recently done
→ Ideas Inbox
Your idea capture — Claude analyzes on request