🤖 XGBoost BTC ACTIVE
Gradient boosted decision tree trained on 2 years of BTC/USDT 1H klines. Predicts if price will move >0.1% up in the next 4 hours.
Accuracy54.1%
Training Data18,000 samples
Features51
AlgorithmXGBClassifier
Max Depth6
Estimators200
Learning Rate0.05
Target4H direction (>0.1%)
Last Trained2026-04-03
RSI(14)
EMA(9,21,50,100,200)
MACD
Bollinger Bands
ATR(14)
Volume Ratio
Momentum(1-24h)
Volatility(14,50)
Hour/DOW cyclical
🤖 XGBoost ETH ACTIVE
Same architecture as BTC model, independently trained on ETH/USDT data. Each coin gets its own model to capture unique patterns.
Accuracy54.2%
Training Data18,000 samples
Features51
Top Featureema_100
Last Trained2026-04-03
RSI(14)
EMA(9,21,50,100,200)
MACD
Bollinger Bands
ATR(14)
Momentum(1-24h)
📊 Technical Analysis Engine ACTIVE
Rule-based engine that computes RSI, EMA crossovers, BB position, MACD, and volume trends from real Binance kline data. Powers 14 of the 16 agents.
Indicators5 (RSI, EMA, BB, MACD, Vol)
Data SourceBinance Klines API
Lookback200 periods
Cache5 min TTL
Agents Using14/16
😨 Sentiment Engine ACTIVE
Real-time Fear & Greed Index from Alternative.me API combined with price momentum for sentiment scoring.
Data SourceAlternative.me API
UpdateEvery 1 hour
Current F&G—
🧠 LSTM Predictor PLANNED
Deep learning sequence model for time-series prediction. Requires TensorFlow + GPU training. Currently using a rule-based fallback.
StatusNot Trained
RequiresTensorFlow, GPU
ArchitectureLSTM (128→64→1)
🐋 On-Chain Whale Tracker PLANNED
Real whale movement detection from blockchain data. Currently using volume spikes as a proxy.
StatusProxy Mode
RequiresWhale Alert API ($)