How the Umpiry Score & Rating work
What the Umpiry Score is
The Umpiry Score is a single number that summarises the context around a tennis match. Rather than relying on one thing — a current ranking, or recent results alone — it brings together several different categories of factors that influence how a match tends to play out, and condenses them into one compact, readable figure for each player. The idea is simple: a ranking or a scoreline tells you part of the story, but the circumstances around a match — who suits the surface, who has travelled, who is fresh — shape it too. The Umpiry Score is a way to read that context at a glance.
The factors behind the score
The Umpiry Score draws on five broad categories of factors. Each captures a different dimension of a match.
Matchup — how the two players stack up against each other: their longer-term standing, how each performs on the surface being played, and their previous meetings. The matchup is the backbone of the score.
Form — recent results and momentum. A player arriving on a run of wins is in a different place than one who has been losing early.
Fatigue — the physical toll of recent tennis. Long matches, a heavy run of matches in a short window, and deep tournament runs all add up and can affect how a player holds up.
Travel — the disruption of moving between events: crossing time zones, adjusting to a new location, and switching from one surface to another between tournaments.
Motivation — the context that raises or lowers the stakes, such as playing at home in front of a familiar crowd or the wider importance of a particular match.
No single category decides the score. They are considered together, so that strength in one area can be balanced against a disadvantage in another — much the way an experienced observer would weigh up a match.
The data it draws on
The factors behind the score are built from data, not opinion. That includes historical match results going back over two decades, player rankings, surface-specific performance, and contextual details such as schedules, locations and conditions. This information is aggregated and cross-checked from multiple sources, then normalised so that players, tournaments and results line up consistently across the platform. Keeping the underlying data clean and consistent is what lets the different factors be compared on a like-for-like basis. The score is updated regularly as new results come in, so it reflects the most current picture available before a match.
What the score is not
The Umpiry Score is a context tool, not a crystal ball. It is not a prediction of who will win, and it is not a betting tip or a recommendation to wager. Tennis is decided on court, and any match can turn on a single point, an injury, or a shift in momentum that no model can foresee. The score is meant to help you understand the circumstances around a match more clearly — to see why a contest might be closer or more lopsided than the rankings alone suggest — not to tell you what to do about it.
Some factors are applied only when the relevant data is available for a given match, so the depth of context can vary from one match to another. Where information is limited, the score leans on what is known and is presented accordingly. It is one informed read among many, not the final word.
How the Umpiry Rating works
The Umpiry Rating is a separate measure from the Umpiry Score above. Where the Score reads the context around a single match, the Rating is a season-long, career-spanning measure of how strong a player is right now. It is an Elo rating — the system Arpad Elo devised for chess, adapted for tennis — and unlike the Score, its workings are fully open.
Every player starts at a baseline of 1500. After each match the winner takes rating points from the loser; beating a stronger opponent transfers more than beating a weaker one. A new player’s rating converges on their true level after roughly 30–50 matches, so the starting baseline has long since washed out for anyone near the top of the lists.
How much a single result moves a rating depends on the importance of the event — the Elo “K-factor”:
- Grand Slams — 32
- Masters 1000, WTA 1000 and the Tour Finals — 24
- ATP and WTA 250 / 500 and International events — 16
- Challengers and ITF events — 8
Each player also carries separate clay, hard and grass ratings, calculated the same way but only from matches on that surface, so a clay specialist’s grass results never flatter their clay number. The ratings are recomputed daily from more than 438,000 matches dating back to 1996, which means the day after a result the lists already reflect it — the “updated daily” you see on the rankings.
Because Elo ratings depend entirely on these settings — the baseline, the K-factors, the start of the dataset — different Elo systems produce different numbers for the same player. The Umpiry Rating, the Tennis Abstract Elo, the UTR and the official ATP and WTA points rankings each use their own method, so values are not directly comparable across systems. What is meaningful is the order and the gaps within a single system. A rating from a long-inactive player reflects their level when they last played: Elo does not fade with time off, only with results, so we flag such ratings rather than hide them.
Alongside the Elo rating, the engine also computes an internal network rating (a Bradley–Terry model). Instead of updating one match at a time, it solves the whole web of results at once — if A beats B and B beats C, A’s rating already reflects the implied edge over C — with recent results weighted more heavily. In our backtests this read women’s matches slightly better than Elo, so the Umpiry Score’s matchup reading uses it for WTA matches. Elo remains the published rating on the rankings pages.
Win Probability
The win probability shown on match pages is a statistical estimate of each player's chance of winning, expressed as a percentage. The two percentages always sum to 100.
The model is calibrated: when it says 70%, players in that situation win about 70% of the time. We validate this with out-of-sample testing rather than tuning for headline accuracy.
The model uses pre-match information to estimate win probability: player ratings (overall and surface-specific), player profile factors, and match context (surface, round, tour). Recent form, head-to-head history, fatigue and travel are captured in the Umpiry Score factor breakdown alongside the probability — giving you both a single estimate and the context behind it.
Tennis tours differ structurally, so we train and validate a separate model for each: ATP & WTA, Challenger (men), and Challenger (women) & ITF. Challenger-level matches are inherently less predictable, which is why these carry a “higher uncertainty” note on match pages.
Models are validated using out-of-sample testing — predicting matches the model has never seen, on the most recent completed seasons. This mimics real conditions: predicting matches that haven't happened yet. Calibration is checked across the full probability range, comparing predicted probabilities against actual win rates.
Probabilities are recalculated regularly until the match starts, using each player's latest ratings. At the start of the match the number is frozen — what you see on a live or finished match is always the final pre-match estimate, never a hindsight revision.
What it is not: not a betting recommendation — Umpiry does not offer, facilitate or promote betting; the probability is statistical information and any decision is yours. Not a certainty — even a strong favourite loses sometimes; a 78% favourite still loses roughly one match in five. Not all-knowing — the model does not factor injuries, illness, on-court retirements, or motivation, and sudden news can make a pre-match number stale.
Umpiry Surprise Score
The Umpiry Surprise Score measures how big an upset a result was. It is simply 100 minus the winner's pre-match win probability (as a percentage): if the model gave the eventual winner a 22% chance, the result scores 78. A score of 70 or above marks a genuine shock — the winner was a clear underdog, with a pre-match win probability of 30% or below.
Crucially, the score uses the win probability frozen at the start of the match, never recomputed with hindsight — so it reflects how surprising the result was at the time, not how it looks after the fact.
Our win probability model has been live since 5 June 2026. Tournaments held before that date do not have frozen pre-match probabilities available, and we do not retroactively assign them — so the Surprise Score and our Upset Audits cover events from that date onward.
Draw luck
On a tournament page, “Draw luck” shows how much the bracket helped or hurt each contender. We run a Monte Carlo simulation of the actual draw to get every player's title chance, then run the same simulation over many random draws — the field shuffled into the bracket with no seeding — to get their title chance in an average draw.
The difference, in percentage points, is the draw luck: a player whose real bracket gives them a better title chance than an average draw got a kinder path (this includes the advantage a top seed gets from being kept apart from the other contenders), while a lower number means a tougher route. It is a property of the draw as it was made, so it stays fixed once the bracket is set — and it is descriptive context only, not a prediction or a betting tip.
Field strength & title run
On a tournament page, “Field strength” describes how strong the draw is by Umpiry Rating: the number of players in the draw, their average rating, the average of the eight strongest entrants, and the single highest-rated player. It uses each player's current rating, so we only show it for the live season — for past editions the ratings no longer reflect the field as it was at the time.
Once the final is decided, the “Title run” summarises how hard the champion's path actually was: how many opponents they beat, the average rating of those opponents, and the toughest single win. We trace the real bracket, so byes and walkovers are not counted as wins. These are descriptive numbers about one edition — we do not normalise them into a cross-tournament “difficulty multiplier,” which would need a season-wide baseline we do not yet publish.
Rivalry Score
The “Greatest rivalries” ranking scores every pair of players who have met at least five times at tour level. The score rewards three things: how often they played, how close the head-to-head stayed (an even record counts for more than a lopsided one), and how much was at stake — finals and Grand Slam meetings add the most.
Crucially it uses no current rating, so retired greats are ranked on the same footing as today's players rather than fading from the list as their rating decays. Head-to-head counting follows the official convention — tour-level matches only, walkovers excluded — and it is a relative ranking, descriptive context only.
Match Greatness Score
The Match Greatness Score ranks how memorable a completed match was. It combines what was at stake — the tournament's level and the round — with the strength of the two players (their pre-match ratings) and the drama of the scoreline: how many sets and games were played, and whether it went to tiebreaks.
A Grand Slam final that goes the distance between two of the best in the world scores highest; a routine early-round win scores low. It is a relative ranking — only the order matters — and it drives our match of the day, match of the week and greatest matches of the year. Descriptive context only, not a prediction.
Live momentum
On a live match page, the momentum view traces how the balance of a match shifts game by game, so you can see at a glance who has the upper hand right now and how the contest swung to get there.
It reads two things in parallel. The first is the flow of play itself — who is serving, who is holding, and the pressure points such as break points won and saved — which is what gives a set its rhythm. The second is the movement in aggregated live market prices, which we turn into a clean in-running probability by stripping out the built-in margin those prices carry (a step often called “de-vigging”), so what remains is an estimate that sums to 100% across the two players.
The result is a live read, not a forecast: it describes the state of the match as it stands, and it moves as the match moves. Like the pre-match win probability, it is statistical information only — not a betting recommendation and not a claim about what happens next. A single break of serve can turn it around in minutes.
Clutch Rating
The Clutch Rating captures how a player performs in the moments that decide matches — break points, deciding sets and tiebreaks — on a 0–100 scale relative to the rest of the tour. It draws on our full match history, needs a minimum sample before a player qualifies, and moves slowly as new results come in. It measures pressure performance, not overall strength: two players of a similar level can have very different Clutch Ratings, which is what makes it useful alongside the Umpiry Rating when sizing up a tight match.
Freshness
Freshness is a current snapshot of a player's physical workload, on a 0–100 scale where a higher number means more rested. It looks at how much a player has been on court recently — how many matches and games they've played in the last week or two, how long and gruelling those matches were, and how many days of rest they've had — and updates every day. It is a measure of recent load, not form or quality, and is most useful for sizing up a tight match where one player arrives fresh and another is deep into a long run.
Responsible play
Umpiry provides data and analysis for informational purposes only. It is an 18+ platform that supports responsible play and does not offer, accept or promote betting of any kind. If you choose to gamble, do so responsibly: set limits, never wager more than you can afford to lose, and seek support if betting stops being fun.