How Picks Terminal works
Reading a fight page, a probability, and everything we do not claim to know
Picks Terminal is an independent combat-sports analytics publication. The terminal above is the live surface for the next UFC card: every bout on the card, the two fighters, their recorded physical and statistical profiles, and a written breakdown of how the matchup is likely to unfold. We are not a sportsbook and not a tipping service. We publish analysis, we archive it at a permanent URL, and we later grade our own predictions in public.
How to read a win probability
Each matchup carries a win probability for one fighter, expressed as a percentage. That number is an estimate of how often we would expect that fighter to win if the bout were contested repeatedly under the same conditions — not a forecast of what will happen on the night. A 65% favourite is expected to lose roughly one time in three. When our probabilities are well calibrated, the fighters we rate at 65% should win about 65% of the time across a large sample; when they do not, the error is visible on our model accuracy page. Nothing on this site is a guarantee, a lock, or advice to risk money.
What goes into a matchup breakdown
A breakdown is built from stored data, not from live speculation. For each fighter we hold a record, weight class and rank where known, physical measurements such as height and reach, stance, and striking and grappling rates drawn from public fight statistics: significant strikes landed and absorbed per minute, striking accuracy, takedown volume and takedown defence, submission attempts, average fight time, and the distribution of significant strikes by position (standing, clinch, ground) and by target (head, body, leg). Those inputs feed the comparison grids, the strike distribution profiles, and the combat signature and archetype labels you see on each fight page.
On top of those inputs sit the written layers: a short angle summarising the central question of the fight, key advantages and potential concerns for each side, and — for the fights we have published in full — a long-form article that walks through the striking exchange, the grappling picture, the cardio and durability questions, and a stated verdict.
How uncertainty and missing data are handled
Debutants and short-notice replacements frequently have no usable statistical sample. When a value is unknown, it stays unknown: a missing striking rate is displayed as unavailable and is never silently rewritten as zero, because a zero would drag every derived comparison and probability toward a fighter who simply has not been measured yet. Where a fight page shows no strike distribution chart, that is the honest state of the record rather than a rendering fault. The same rule applies to results — a bout that has not been officially settled is not scored.
How predictions are audited after the fact
Before an event begins we capture an immutable snapshot of every published probability. Those snapshots cannot be edited afterwards, which means our record cannot be quietly improved once results are known. When official results are recorded, each snapshot is scored with the Brier score — a standard measure of probabilistic accuracy — and grouped into calibration buckets so you can see whether our 60–70% picks really do land around two-thirds of the time. The full ledger, including the fights we got wrong, is public on the model accuracy page.
Where to go next
The methodology page documents our data sources, publication gates and correction process in detail. The schedule lists upcoming and completed cards with their permanent pages, and about covers who operates the site and how we stay independent of advertisers.
Standing74%
Clinch7%
Ground19%