The verdict grade scale
WeBall2 min readUpdated Aug 15, 2026Reviewed Aug 15, 2026
Every pregame selection on the board receives exactly one grade. Five of them are opinions about price; two are refusals to give one.
| Grade | Meaning | What to do |
|---|---|---|
| GOOD | Model probability clearly beats the best available price. | Bet it at the listed price or better. Below that price, re-check the edge. |
| LEAN GOOD | Positive edge, but modest — directional, not a hammer. | Playable at a reduced stake, or a reason to take this side if you were already leaning. |
| NEUTRAL | Market and model agree. No bet. | Pass. You are paying the vig for no edge. |
| LEAN BAD | Slightly overpriced — the other side is closer to fair. | Pass, and check the opposite side. |
| BAD | Clearly overpriced against the model's number. | Pass. If you want this game, the other side is where the value sits. |
| THIN DATA | Not enough sample to trust the number. Informational only. | Don't act on this grade. It is shown so you know we looked. |
| UNRATED | Displayed but not graded — no model covers this market yet. | Use it as a price reference. There is no WeBall opinion attached. |
The five ratings
The rating grades all answer the same question — how does the model's probability compare to the market's fair price? — and differ only in the size and direction of the gap.
GOOD and BAD are the two ends: a gap large enough to survive its own uncertainty, pointing in one direction or the other. LEAN GOOD and LEAN BAD are the same thing at a size that does not comfortably clear the model's error bars, which makes them tiebreakers rather than signals. NEUTRAL means the model and the market agree, which is the correct answer most of the time and is worth publishing rather than hiding.
The two refusals
THIN DATA means a model produced a number and the confidence gate rejected it. The bet may well be good; there is simply not enough sample behind the probability to say so honestly. Common in the first weeks of a season, for newly promoted teams, and for players with few logs.
UNRATED means there is no model for this market, or the entity behind it could not be resolved to history with enough certainty to risk a wrong-player or wrong-team grade. The price is real; the opinion does not exist.
How a grade is chosen
Two inputs decide it: the size of the edge and the confidence behind it.
A large edge with low confidence does not become GOOD — it becomes THIN DATA, because a big disagreement built on a small sample usually means the model is wrong rather than the market. Confidence gates the grade before edge sizes it.
What a grade is not
It is not a prediction, and it is not advice. It is a statement about whether a price is better or worse than the outcome deserves, produced mechanically and published with its own uncertainty attached.
The known limitations page is explicit about where these grades are least trustworthy, and no track-record claims are made anywhere until the calibration ledger has enough settled bets to support them.
Common questions
Why are there two kinds of "no grade"?
They mean different things. THIN DATA means a model ran and the confidence gate rejected the result. UNRATED means no model exists for that market, or the entity behind it could not be resolved to history. One is a data problem, the other a coverage gap.
Does GOOD mean the bet will win?
No. It means the price pays more than the outcome deserves. A GOOD bet at 55% loses 45% of the time, and a run of losses inside that is completely ordinary.