$—
Loading…
Model: Not trained
ε=1.000 · WR—
TOTAL TRADES
0
WIN RATE
—%
TOTAL P&L
$0.00
AVG CONF
—%
SHARPE
—
🧠 RL MODEL STATUS
Episodes Trained
—
Train Win Rate
—%
Sharpe Ratio
—
Exploration ε
—
States Active
128
Last Trained
—
⚙️ TRAINING CONTROL
Ready. Click Train Agent to begin Q-learning.
🔍 SERVER DIAGNOSTICS
Checking…
API endpoint:
📊 BACKTEST RESULTS
Click Run Backtest to evaluate model on 1200-candle synthetic BTC series.
⏱ 5-MIN WINDOW
5:00
UNTIL RESOLUTION
Window: —
Optimal entry: Last 90 seconds
ANALYZING…
Train the model to generate live signals. The Q-learning agent learns optimal UP/DOWN/SKIP decisions from reward feedback.
UP
0.000
DN
0.000
SKIP
0.000
LIVE INDICATORS
RSI-7
—
—
EMA 9/21
—
—
MACD
—
—
BB Position
—
—
EQUITY CURVE (Virtual $10 Trades · Starting $1,000)
VIRTUAL TRADE LOG
TIME DIR PRICE CONF RSI RESULT P&L
Place virtual trades to see history
⚡ PLACE VIRTUAL TRADE
Train the model first, then the bot will signal UP or DOWN with confidence score. $10 virtual bets placed here.
Bet Size
$10.00 USDC
Entry Price
—
Win Payout
+$9.20
Loss
-$10.00
EV (at conf)
—
Kelly Stake
—
No pending trade
⚠️ RISK METRICS
Signal Str.
0%
Exploration
100%
Win Rate
0%
Timing
50%
Risk assessment will appear after training.
🧬 HOW THE RL WORKS
Q-LEARNING AGENT
State = RSI(7) bin × EMA-cross × MACD-sign × BB-position → 128 states
Actions = UP · DOWN · SKIP

REWARD SHAPING
✓ Correct bet: +0.92 × confidence
✗ Wrong bet: −1.20 × drawdown
— Skip: −0.10 (prevents over-skipping)
★ High-conf correct: +0.15 bonus

STABILITY FEATURES
• Asymmetric loss penalty (losses hurt 1.2× more)
• Confidence gate — skip if uncertain
• Drawdown-aware epsilon (explore less when losing)
• 25% fractional Kelly sizing
• 3-way confirmation required for green signal
🔗 PLACE REAL BETS
↗ Polymarket 5-Minute Markets ↗ Bitcoin Markets
⚠ Virtual trades only. Real bets placed manually on Polymarket. Past signals ≠ future results.