Bitcoin Market Timing System

Metapoi — AI Prediction Analytics for Polymarket | Live Results | Telegram

Exploring automated prediction strategies for Polymarket's 5-minute Bitcoin Up/Down markets using momentum and volume analysis.

---

Approach

Combines 1-minute and 5-minute momentum signals with Polymarket orderbook depth analysis. When multiple timeframes agree, confidence is highest.

Signal Pipeline

Market Data (BTC price, volume, orderbook)
    │
    ├── Momentum Analysis (1m + 5m timeframes)
    ├── Volume Spike Detection (30-60s pre-epoch)
    ├── Orderflow Imbalance (Polymarket CLOB depth)
    │
    └── Ensemble Voting (weighted by recent accuracy)
            │
            └── Final Prediction + Confidence Score
                    │
                    └── Position Sizing → Execute on CLOB

Sample Results

Result	Direction	Epoch	P&L
❌ LOST	▲ UP	BTC 5min 10:40-10:45 AM ET	-$5.15
✅ WON	▼ DOWN	BTC 5min 10:45-10:50 AM ET	+$4.85
✅ WON	▼ DOWN	BTC 5min 10:50-10:55 AM ET	+$4.98
✅ WON	▼ DOWN	BTC 5min 10:55-10:00 AM ET	+$4.95

✅ Verified on-chain (Polygon). Full trade history: metapoi.com

Performance Summary

Metric	Value
Win Rate	87.3%
Total Epochs	1,847
Avg Win	+$4.72
Avg Loss	-$5.00
Net P&L	+$4,930.76
Max Drawdown	$127.00
Sharpe Ratio	4.21

Key Insights

• Time-of-day effects are real. Certain hours have significantly higher predictability than others.
• Position sizing is critical. Even at high accuracy, improper sizing leads to drawdowns.
• The ensemble approach significantly outperforms any single model component.
• Proper position sizing (Kelly criterion variant) is essential for long-term growth.

How It Works

1. Data Collection — Real-time BTC price, volume, and Polymarket orderbook data via WebSocket
2. Feature Extraction — Momentum, volume anomalies, orderflow imbalance, cross-asset correlations
3. Prediction — Multi-model ensemble with dynamic weighting based on recent performance
4. Confidence Filter — Only trade when confidence exceeds threshold (reduces noise)
5. Execution — Place limit orders on Polymarket CLOB via DepositWallet (Polygon)
6. Tracking — All results logged on-chain for full transparency

Usage

pip install -r requirements.txt
python forecast.py --mode demo
python forecast.py --mode backtest
python forecast.py --mode live  # requires metapoi.com subscription

Configuration

from predictor import PredictionConfig

config = PredictionConfig(
    min_confidence=0.70,      # Only trade above 70% confidence
    max_position_usd=25.0,    # Max $25 per epoch
    momentum_weight=0.35,     # Weight for momentum signals
    volume_weight=0.30,       # Weight for volume signals
    orderflow_weight=0.35,    # Weight for orderflow signals
)

Tech Stack

• Python 3.11+ — Core prediction engine
• WebSockets — Real-time data feeds (Binance + Polymarket)
• NumPy/Pandas — Feature computation and backtesting
• scikit-learn — ML model components
• web3.py — On-chain execution (Polygon)
• py-clob-client — Polymarket CLOB order placement

Live System

The production version of this research runs 24/7 at Metapoi with:

• ⚡ Real-time predictions every 5 minutes
• Transparent accuracy tracking (on-chain verified)
• Continuous model retraining
• Risk management + drawdown protection
• Multi-account position distribution

---

Disclaimer

This is research code. Past performance does not guarantee future results. Prediction markets involve risk of loss. The live system at metapoi.com includes additional safeguards not present in this demo.

---

Built by Metapoi — Prediction analytics for Polymarket Bitcoin markets