Reading a row
Every row is one claim. Here is what its numbers mean.
A row is one claim about one player, team or game — “at least 1 hit”, “over 229.5 passing yards”, “to win”. It carries a few things, and they answer different questions.
The tokens
- The percentage, largest on the row
- Our probability that the claim lands. Where a market has a calibration curve applied it is calibrated: of claims we print that number on, about that share have landed. Where no calibration curve is applied, the number is the model’s raw estimate, and the board says so.
- Typical
- How often this kind of claim typically lands with no model at all. It is the bar’s tick, and the number a row is read against where one has been measured — a probability is only interesting next to the typical rate it beats or misses. A team or game line reads against Even instead, and a three-way match result against Even 33, its even split. A player prop has neither, so its bar carries no tick at all: the row shows the probability and names what it was read against in words. A run line is different again: we hold no book price for it, and its line is not set to split the outcome, so it draws no tick either — its row states its line in its name.
- The points above typical
- How far our number sits above the typical rate, in points.
- Flagged
- Our chance clears the cutoff we set for that stat. Only some stats can be flagged. It ranks rows; it isn’t a promise.
- Projection
- Our model’s estimate for a stat we don’t flag today. We still show it and grade it hit or miss.
Deliberately no example probabilities or record counts on this page: those are published figures needing their own records under our claims standard, and the honest version is the live one. (A market line like “229.5” above is neither — it is the book’s number, not ours.) Open any board and the tokens above are on its rows — the typical rate and the points above it only where they have been measured.
Our number, and the book’s
The big number is our model’s chance that it happens. If a sportsbook’s number is shown, it sits next to ours.We never combine them into one number.
One habit worth forming
Open a row’s detail before acting on it. The headline number is the least interesting thing there: behind it are the matchup history, the sample sizes and the caveats — and thin samples say so rather than being presented as a rate you can lean on. Where no calibration curve is applied, the detail says that too.
Where a factor breakdown exists, it is there too. Factors are served for the MLB batter markets; every other row shows the projection and its calibration state, and says plainly that a breakdown for it is not built yet. The detail sheet never invents one.
Where the numbers come from is Methodology. What they have actually done is Results, which is open without an account.