Help

Glossary

One definition, one place. How it works cites the first four; the product’s own pages cite the rest.

Probability
A number between 0 and 1 for an outcome happening. BetDNA produces one per event from that sport’s own evidence, not from a single formula stretched across every sport.
Confidence
How much evidence stands behind that probability. Two events can share a probability and deserve completely different treatment because one of them is built on almost nothing.
Implied probability
The market price, converted. Decimal odds of 2.50 imply 40%. This is the number a probability has to disagree with to be worth anything.
Edge
The gap between the two, after the margin built into the price. A small positive gap on a well-evidenced event is worth more than a large one on a coin-flip.
De-vig
Taking the bookmaker’s margin back out of a set of prices so the implied probabilities add up to one. It is what turns a price into something a probability can be compared with.
Tier
Tiers grade the stake by the model’s probability for the pick: Platinum from 78% · Gold from 68% · Silver from 55% · Bronze from 45% · Punt from 38% · below 38%, no stake band. A tier grades the stake; it is not what makes a pick a pick.
Calibration
Whether a model that says 70% is right about 70 times in a hundred. Graded only against forecasts made before the event.
Log loss
Minus the natural logarithm of the probability the model gave the outcome that happened, averaged over the forecasts. Lower is better; a confident wrong forecast costs far more than a hesitant one.
As-of
Graded with only what was known before kick-off. A forecast written after the result, or rewritten with the result in hand, never counts.
Replay
A match played out on screen from its real result and timeline. It is not a prediction and it does not change one.
Board
The fixtures BetDNA carries for a window of days, every one with BetDNA’s view on it or a stated reason there is none.
Window
The days the board covers: 3 back, today, and 2 ahead - 6 days.