How MarketLens forecasts learn from real outcomes

Closing the loop\nWhen a forecast's target date passes, MarketLens compares the forecast with the observed close.\n\n## What changes\n- Models with lower realised error on that instrument receive more weight.\n- The most recent session is re-checked each time: what would yesterday's model have predicted for today?\n\n## What does not change\nNo forecast is ever edited after the fact. Learning only affects new forecasts.