Model validation

How to verify prediction-software accuracy

Start with denominators, date coverage, prespecified metrics and a complete log.

Updated 2026-09-20

This guide uses testable rule assumptions and does not infer any platform’s implementation from its name. It is educational material, not betting advice.

Key definition

Accuracy needs a clear denominator: which rounds were predicted, skipped, failed or produced no output.

How to verify it

Predictions must be created before outcomes and carry tamper-evident timestamps to rule out backfilling.

Limits and conclusion

Beyond headline accuracy, examine class imbalance, confidence intervals, calibration and improvement over a simple baseline.

Practical checklist

An independent reviewer should be able to recompute metrics from the complete log. Screenshots and selected hits are not evidence.

Practical checklist

  • Accuracy needs a clear denominator: which rounds were predicted, skipped, failed or produced no output.
  • Beyond headline accuracy, examine class imbalance, confidence intervals, calibration and improvement over a simple baseline.
  • Fix the rules, preserve a complete record, separate training from testing, and only then ask whether a model adds information beyond a simple baseline.