Forecast Scorecard
The honest answer to “does it work?”. Every model is judged against simple baselines on the same stocks and days. A model only counts as working if it beats them consistently, by more than luck could explain.
Evidence so far · 5 days (~1 week)
Towards the 3-month target0%
Days scored: 0 / 60Forecasts scored: 0 / 1,000 per model
First verdicts after 20 days. “Proven” needs 60 days. 0 forecast days recorded so far; 0 forecasts are waiting for their result.
Nothing scored yet for this horizon. Next-day forecasts are scored the evening after they're made, 5-day ones a week later and 20-day ones about a month later. See the latest forecasts →
Nothing scored yet
Scores appear as forecasts reach their horizon. Next-day forecasts are first scored the evening after the first run.
How to read this
- Hit rate: how often the predicted direction (up or down) was right. Stocks rise on most days, so the “always up” baseline already scores above 50%.
- vs baseline: the model's daily hit rate minus the baseline's, on the same stocks. This is the number that matters.
- Chance it's luck: how likely an edge this size would appear by chance. Under 5% is convincing; under 1% is strong.
- Probability skill: are its percentages more accurate than the baseline's? Above 0 means yes.
- Ranking skill: do the stocks it rates best actually do best? Even +0.03 to +0.05 is useful if it lasts.
- Top-10 picks vs avg: how much better its 10 favourite stocks did than the average stock: what you'd earn by following it.