Inside the TenINSIDETHETENInside the TenINSIDETHETEN
Receipts · Accuracy

Inside the Ten grades itself, in public

Every week we freeze the projections we're standing behind and publish their content hash before kickoff. Once the week completes, we score that exact frozen prediction against what actually happened — and against public baselines — with one documented methodology. Nothing is revised after the fact, and we show every week, including the ones a baseline beat us.

How the scoring works

  • • Freeze + hash. Each week's predictions are content-hashed (SHA-256) and timestamped before the games — tamper-evident.
  • • One methodology. The same code scores this page and our internal report — MAE / RMSE / bias / rank correlation / top-N hit rate / band coverage over predicted-vs-realized pairs, pooled season-to-date.
  • • Comparative. The model is scored beside named public baselines under identical eligibility — not graded against itself.
  • • Verifiable. Each row links its frozen hash; re-score the stored prediction and you get the same number.