Early/provisional vs enough graded history
“Enough graded history” means forecasts were frozen before events, misses were kept, and you can see how stated percentages lined up with outcomes over a real sample—for that sport or market. “Early/provisional” means the board may be live, but that reliability story isn’t there yet. Treating every row as equally settled overstates confidence.
A quiet trust problem in sports modeling: a single board that mixes sports at very different stages of proof.
Some sports (or markets) have enough graded history that you can talk about how probabilities behaved in buckets over time—forecasts frozen before the event, outcomes recorded, misses kept. Others are earlier / provisional: useful to publish, but not ready to borrow the trust of a long graded record.
Plain-language bar:
- Enough graded history — pre-event forecasts, full outcomes (including misses), visible sample size, and a clear view of whether stated percentages roughly behaved like those percentages over time for that sport.
- Early / provisional — the model output is available, but the reliability story is still thin; say so instead of using one global badge for every row.
What to avoid: a blanket “our model is calibrated” claim that doesn’t name the sport, and quietly upgrading language without showing the graded work.
On TrueProps—not a sportsbook—we try to be explicit about maturity by sport. Where we have a public graded record you can inspect, we point people there. Where a sport is still early, we prefer provisional language over borrowed confidence. See the board and graded Results. Start free; if maturity labels are unclear, that’s useful feedback.
Questions
- Does provisional mean “bad”?
- No—it means “not yet backed by enough graded history.”
- Can a sport move from provisional to graded?
- Yes, as the public ledger grows—without deleting misses.
- Why not wait until everything is graded?
- Publishing early with honest labels beats silent confidence.