Approve trading signals via XGBoost probability gate and dual-LLM ensemble voting.
npx skills add https://github.com/raosuraj23/alpaca-bot --skill predict-raosuraj23
Please help me install this Agent Skill. Skill: predict Source: https://github.com/raosuraj23/alpaca-bot/tree/main/.claude/skills/predict Command: npx skills add https://github.com/raosuraj23/alpaca-bot --skill predict-raosuraj23
This Skill automates rigorous signal approval for trading strategies by combining a machine-learning probability gate with dual-LLM ensemble voting.
Run the validator with a sample xgb_prob and market_implied_prob to get an approval decision
High-intent search queries and answers about installing and using this skill.
You validate trading signals by passing them through a three-layer system: an XGBoost probability gate checking win probability and edge, a dual-LLM AND gate using Gemini and Haiku, and a calibration layer tracking Brier scores.
Dual-LLM ensemble voting is a consensus mechanism requiring both Gemini and Haiku language models to independently approve a trading signal before execution, enforcing cross-model agreement to reduce false positives.
You test signal logic offline by running the standalone validator script provided in the scripts directory, supplying sample XGBoost probability and market-implied probability values to generate an approval decision.
Yes, the system applies XGBoost probability thresholds, requiring a minimum win probability of 0.55 and an edge of 0.04, to filter real-time technical analysis signals and research-driven alerts.
Calibration tracking works by calculating a Brier score to measure prediction accuracy and applying a Kelly scalar to adjust position sizing, ensuring probability estimates align with actual historical outcomes across strategy contexts.
You should use dual-LLM gating when strict cross-model consensus is required to approve signals, preventing single-model biases from triggering trades and ensuring higher reliability for research-driven alerts.
AlphaGBMCompare Polymarket probabilities with options-implied probabilities to surface mispricing signals.
PuckAPICalibrate sports prediction model probabilities using logistic or isotonic regression.
ovachieverPerform survival analysis on censored data with Cox models and ensemble methods.
bitandbytesAutomate ML meta-labeling for trading with triple-barrier labeling and XGBoost calibration.
