Appearance
Accuracy
This is FIN's report card — it shows how well predictions matched what actually happened in the market.
What's in the Table
| Column | What It Means |
|---|---|
| Pred Date | When the prediction was made |
| Target Date | The date the prediction was for |
| Predicted | What the system predicted ($) |
| Actual | What actually happened ($) |
| Error % | How far off the prediction was (as a percentage) |
| Band Hit | Did the actual price fall within the expected range? (Yes/No) |
How to Read Error %
- Green (< 5%): The prediction was close — good job
- Amber (5–10%): Somewhat off — worth noting
- Red (> 10%): Significantly wrong — something unexpected happened
A prediction can have a large error % but still be a "Band Hit" if the actual price fell within the confidence band. That means the system correctly judged how uncertain it should be.
Band Hit: The More Important Metric
The system doesn't just predict a single number — it predicts a price range (the confidence band). A Band Hit means the actual price landed inside that range.
| Band Hit | What It Tells You |
|---|---|
| Hit | The system correctly estimated the uncertainty. The actual price was where expected, even if the exact number was off |
| Miss | The actual price was outside the predicted range — the system was too confident or got the direction wrong |
A high band hit rate (70%+) means the system is well-calibrated.
The Feedback Loop
Accuracy data doesn't just sit there — it drives improvement:
Prediction made→Market close→Actual price compared to prediction→Error calculated→System learns→Next prediction is better
What to Watch
- Track your trend — is error % going down over time? That means the system is learning.
- Watch for patterns — if predictions are consistently too high, the system should self-correct.
- Check band hits — a low hit rate means the system is either too confident or missing the right direction.
- Big errors are clues — a sudden large error usually means an unexpected event (check the Surprises page).

