Session Improvements Summary

Bayesian Confidence Calculation: Implemented alternative confidence calculation using proper Bayesian evidence accumulation with log-odds and likelihood ratios
Dual Calculation System: Both additive and Bayesian methods calculated simultaneously for comprehensive A/B testing capabilities
Market Prior Integration: Bayesian method starts with Polymarket orderbook probability as prior, anchoring calculations to market reality
Quality Factor System: Signal evidence strength modulated by per-signal quality factors (0.7-1.5x multiplier on log-likelihood)
Database Schema Update: Migration 007 added additive_confidence, additive_bias, bayesian_confidence, bayesian_bias, market_prior_p_up columns
Configuration Toggle: BAYESIAN_CONFIDENCE flag in settings allows switching between additive and Bayesian methods
A/B Testing Workflow: Collect 100+ trades with both methods stored, then compare win rates using SQL queries
Advantages: Bayesian method naturally handles conflicting signals (they cancel out), properly combines independent evidence, and provides better uncertainty handling

polyflup-history