Calibration
How closely forecast probabilities match observed frequencies across a sufficiently large comparable sample.
The language of
probability and price.
Short, extractable definitions for the concepts FORM/PRICE uses. Every metric needs a formula, a cohort and an evidence boundary.
Closing line value (CLV)
A comparison between the price captured at decision time and a defined price near market close, using the same market and source policy.
Drawdown
The largest peak-to-trough decline in a sequential unit ledger under a stated staking method.
Edge
The difference between a model-estimated probability and a comparable market probability after the bookmaker margin is addressed.
Expected value (EV)
For a one-unit decimal-odds decision, model probability multiplied by odds, minus one. It is an uncertain estimate, not realised profit.
Fair odds
The break-even decimal price implied by a probability: one divided by the probability.
Hit rate
The share of settled selections that win. It must be read beside average price, sample size and market definition.
Implied probability
The raw probability represented by decimal odds, calculated as one divided by the odds before margin adjustment.
Log loss
A scoring rule that penalizes probability forecasts, with especially large penalties for confident predictions that are wrong.
Market overround
The amount by which the raw implied probabilities of all outcomes exceed 100%, commonly used as a view of bookmaker margin.
Minimum acceptable price
The lowest current price at which a model estimate still passes its declared value and uncertainty buffer.
No-vig probability
An estimate of market probability after removing the bookmaker margin under a stated method.
Over/Under 2.5 goals
A two-sided football market settled on whether the match contains at least three total goals or no more than two, subject to the operator's settlement rules.
Prospective record
A forecast and price record frozen before the outcome is known and retained without hindsight rewriting.
Return on investment (ROI)
Net unit profit divided by total units staked for a defined, complete cohort.
Brier score
The mean squared difference between a probability forecast and a binary outcome. Lower is better within a comparable task.