SUPERPOWERS ARENA -- The Horse Race
5 ML systems compete. Honest metrics. May the best algorithm win.
Live Loading...
🛠 SYSTEMS UNDER ACTIVE REPAIR — Major Overhaul In Progress

A 28-agent AI research audit found 10 critical bugs. All are now patched. A 90-day backtest on 4h data validated the fixes: System B hit 56.6% WR (Sharpe 9.91). Systems A and C were redesigned based on backtest results. Fresh picks from the corrected pipeline start now.

📈 Backtest Proof (90 days, 4h, 10 pairs, 0.35% costs)
System Trades Win Rate Sharpe Status
The Regime (B) 1,656 56.6% 9.91 PRIMARY
The Filter (A) 313 39.3% -3.13 REDESIGNED
Neural Net (C) 90 44.4% -1.11 TP LOWERED
Carry Trade (D) -- 60% (exp) -- NEW
Momentum (E) -- 55-60% (exp) -- NEW
⚡ The Filter (REDESIGNED)
RSI divergence replaces raw oversold
5-bar swing detection for entries
Regime filter from System B
Confidence threshold raised to 0.65
Circuit breaker at 0% WR
🎯 The Regime (PRIMARY - 56.6% BT)
Backtest: 56.6% WR, Sharpe 9.91
HMM cascade regime detection
4h primary, momentum + low-vol filter
2x weight in ensemble
Standalone picks at 50% size
🧠 The Neural Net (TP FIXED)
Signal direction fixed (was backwards)
TP lowered: 3x→2x ATR
Hold time extended: 20→30 bars
SL widened to 2x ATR minimum
Circuit breaker pauses at 0% WR
💰 The Carry Trade (NEW)
Funding rate contrarian carry
Exploits overleveraged longs/shorts
Binance perpetual funding rates
Expected WR: 60% (19-115% annual)
Research: R001, R005, R026
🚀 The Momentum (NEW)
Cross-sectional momentum ranking
Top 3 / bottom 3 by 7d momentum
Liu et al. 2022 JFE (Sharpe ~2.1)
Expected WR: 55-60%
Research: R001, Liu et al. 2022
🚀 All Fixes Deployed (Feb 24-25)
✓ 10 critical bugs patched from 28-agent audit
✓ Confidence thresholds raised (A:0.65, B:0.55, C:0.75)
✓ Transaction costs raised to 0.35%+ (was 0.20%)
✓ Stop-losses widened (2x ATR min + 1.5% floor)
✓ Cross-sectional momentum rank feature (Sharpe ~2.0)
✓ Funding rate 5-feature decomposition (21 features total)
✓ Ensemble coordinator with B-primary weighting
✓ Circuit breakers halt trading at 0% WR / 8% DD
⚡ System B: Regime direction filter blocks BUY in downtrends
⚡ System A: Dynamic ML threshold (PANIC=0.85, WARNING=0.75)
⚡ System B: F&G fear gate widened to <25 (was <15)
⚡ 3 toxic strategies hard-killed (-$18K total losses)
📅 Timeline to Proof
Feb 24 (NOW) All critical bugs patched, new features deployed, systems scanning on 4h Feb 25-28 First picks from corrected pipeline. Expect 5-15 new trades. Early signal quality visible. Mar 1-7 30+ trades accumulated. Statistical significance begins. Ensemble coordinator active. Mar 7-14 50+ trades = Deflated Sharpe Ratio test. First “PROVEN” badge possible if WR > 55% & Sharpe > 1.0. Target 55-65% win rate, Sharpe > 1.5, max drawdown < 10%
Loading...
Head-to-Head Comparison
SystemInceptionLast UpdatedTradesWin RateSharpeMax DDTotal PnLStatus
Loading all systems...
Strategy Breakdown (All Systems)
Age: | Direction:
Loading strategies...
Equity Curves Overlay
System A: The Filter
System B: The Regime
System C: The Neural Net
System D: The Carry Trade
System E: The Momentum
Consensus Signals (2+ Systems Agree)
Analyzing system signals...
Individual System Dashboards
The Filter
ML Signal Quality Filter
--
Win Rate
🎯
The Regime
Market Regime Detector
--
Win Rate
🧠
The Neural Net
Deep Learning Predictor
--
Win Rate
💰
The Carry Trade
Funding Rate Contrarian Carry
--
Win Rate
🚀
The Momentum
Cross-Sectional Momentum Ranking
--
Win Rate