The labels you will meet
| Label | Where | Meaning |
|---|---|---|
| Lean / Medium / High | UFC PBE Algo | Probability band plus evidence requirements; thin samples capped at Medium |
| FULL / LIMITED / INSUFFICIENT DATA | UFC Fight Simulator | Whether there is enough verified Fight DNA to simulate reliably; LIMITED results are shown with reduced confidence |
| High / Medium / Low / Insufficient | Fight DNA | Sample size by bouts and observed seconds; Insufficient stats are withheld, not shown as zero |
| High / Medium / Low / insufficient | Tennis DNA | Sample behind each value; percentiles only at medium or high, same tour only |
| Minutes gate | Soccer Player DNA | No ranks below 450 nominal minutes; counts still shown |
| Tracking / official | NFL model products | Graded record lanes that are never merged |
| Shadow / not published | NHL, NBA | Models that run privately or not at all until they pass release criteria |
| Untrained | Boxing bout model; Tennis pick model | No pick or fair line is published |
No-calls are designed
Not in scope; bout off the card, changed or already fought; fighter identity not fully reconciled; no live model version; features could not be assembled; latest fight not yet in Fight DNA; a UFC debut; fewer than 20 of 33 features available; too close to call (below 55%); lock window closed.
Every one of these is a case where a number would look precise but be unreliable. Refusing to publish is the honest output.
Why the network works this way
A probability without its evidence invites overconfidence. Tiers, gates and no-calls push the uncertainty onto the screen, where you can see it. Read the label before the number.
Tier thresholds and gate criteria are product decisions and can change between model versions. The product's own model card or methodology page is the source of truth for the current rules.