Table Tennis Match Probability, Explained
Give one number per player, the chance of winning a point on their own serve, and this guide shows you how to turn it into game, deuce and match odds for best-of-3, 5 or 7 using the exact Markov chain of an 11-point game.
What the tool does and one quick example
Table tennis scores in games to 11, win by two, and the serve changes hands every two points. The whole match reduces to two probabilities: how often Player A wins a point when A serves, and how often B wins a point when B serves. From those two numbers you can compute everything else exactly.
Here is the hook. Suppose A wins 55% of points on serve and B wins 52%. That looks lopsided at the point level, but a single point is a small edge. Fold it through a full best-of-5 and A wins the match about 64% of the time. The edge grows as you stack points into games and games into a match, yet it never runs away the way tennis serve dominance does. That is the story this page explains in numbers.
When to use it, and when not
Use this when you have a defensible estimate of each player's serve-point win rate and you want the downstream odds: game win probability, chance the game reaches deuce, match win probability, and the distribution of final game scores (3-0, 3-1, 3-2 and so on).
Do not use it as an oracle. The single-number-per-player model assumes points are independent and that serve win rate is constant across the whole match. Real players tire, adjust tactics, and ride streaks. The tool offers a momentum mode to probe that, but the honest error bar on any pre-match prediction here is wide: a point-rate estimate off by two percentage points can swing match odds by five to ten points.
Serve matters far less than in tennis. Elite table tennis players win roughly 52 to 56 percent of points on serve, so games sit much closer to a coin flip and upsets are common. If you have used the Tennis Match Probability Calculator, expect flatter, closer numbers here.
The point model and the serve rotation
Two parameters drive the chain. Call them s_A and s_B: the probability that the server wins the point when A serves, and when B serves. In the defaults, s_A = 0.55 and s_B = 0.52.
Serve alternates every two points until 10-10. So the sequence of servers in a game from A's first serve is A, A, B, B, A, A, and so on. At 10-10 the rule changes: serve alternates every single point until someone leads by two. The probability that A wins any one point depends only on who serves it:
Here p_A^{(A\,serves)} is A's chance to win a point A is serving, equal to s_A. When B serves, A wins the point with probability 1 - s_B, since either the server wins or the receiver does. With the defaults, A wins 0.55 of A's serves and 0.48 of B's serves.
Building the game as a Markov chain
A game state is the pair of scores plus whose turn it is to serve, which is fixed by the point number. Track the state (a, b) where a is A's points and b is B's. From each state you branch two ways: A wins the point, or B does, each with the serve-dependent probability above. Absorbing states are any score where one side has at least 11 and leads by two.
The deuce region needs care. Once you reach 10-10, the game becomes a two-point-lead race with single-serve alternation. Because the server changes every point, the per-point win chance for A settles to an average of a serve point and a receive point. Over one A-serve plus one B-serve pair, A expects to win
With the defaults, q_A = \tfrac{1}{2}(0.55 + 0.48) = 0.515. From 10-10, the chance A eventually wins the game is the classic two-point-lead formula:
Plug in q_A = 0.515: the numerator is 0.2652, the denominator is 0.2652 + 0.2352 = 0.5005, so A wins from deuce about 0.530 of the time. A tiny per-point edge of 1.5 points becomes a 3-point edge once you demand a two-point margin, because the trailing player has to break serve twice in a row.
A worked example with the demo numbers
Best of 5, serve 55% and 52%
These are the demo defaults: s_A = 0.55, s_B = 0.52, format best-of-5, independent points. Run the chain and you get the game and match figures below.
- Compute A's chance to win a single game. Solving the full 11-point chain (including the deuce branch above) gives roughly g = 0.574. So even a 3-point serve gap plus a small receive edge makes A a modest favourite per game, not a runaway one.
- Turn game probability into match probability. Best-of-5 means first to 3 games. The match win probability is the sum over the ways A reaches 3 wins before B does.
- Write the best-of-5 sum with g = 0.574 and 1-g = 0.426:
P_{match} = \sum_{k=0}^{2} \binom{2+k}{k} g^{3} (1-g)^{k}
- Evaluate the three terms. The g^3 factor is
0.1892. The 3-0 term is 1 \cdot 0.1892 = 0.1892. The 3-1 term is 3 \cdot 0.1892 \cdot 0.426 = 0.2418. The 3-2 term is 6 \cdot 0.1892 \cdot 0.426^2 = 0.2060. - Add them: 0.1892 + 0.2418 + 0.2060 = 0.637. So A wins the best-of-5 about
64%of the time.
Notice how the match odds (64%) exceed the game odds (57%), which exceed the deuce odds and the point odds. Every layer of best-of amplifies the favourite. That is the single most important intuition on this page.
Reading the outputs
The tool reports four things. Game odds are A's chance to win one game. Match odds fold that into the chosen format. The deuce probability tells you how often the game reaches 10-10, which for two close players is surprisingly high, often 15 to 25 percent. The game-score distribution splits the match odds into 3-0, 3-1 and 3-2 style outcomes.
Read the score distribution as a shape, not a single number. With the defaults, A's 64% match chance breaks down roughly into a 3-0 sweep, a 3-1, and a tight 3-2, each meaningfully likely. A favourite who is only slightly ahead wins more of their matches in five games than in three.
Common mistakes
The first mistake is confusing serve win rate with overall point win rate. A's overall share of points is not s_A; it is the average of A serving and A receiving, which is q_A = 0.515 in the defaults. Feeding 0.55 in as an overall rate would overstate A badly.
The second is assuming a longer format always helps you. It only helps the favourite. If you are the underdog, a best-of-7 is worse for you than a best-of-3. With the defaults, B's chance falls from 38.3% in best-of-3 to 34.9% in best-of-7.
The third is treating momentum mode as more accurate by default. It is a different model, not a more precise one.
Momentum mode runs 20,000 simulated matches point by point, so it carries Monte Carlo noise of about \pm 0.3 to \pm 0.5 percentage points on match odds, and it takes a few seconds. Set momentum strength to 0 and it should reproduce the exact independent-points result within that noise. If it does not, your two runs simply landed on different random draws.
What momentum strength actually changes
The momentum parameter nudges the next-point win chance toward whoever just won a point. At 0.03 the effect is mild and realistic. At 0.1 you get heavy streaks. Streakiness barely moves the match favourite, since boosts to the leader and the trailer roughly cancel over a long match. What it does move is variance: with strong momentum you see more blowout games and more improbable comebacks, so the game-score distribution spreads toward the extremes (more 3-0 sweeps and more 3-2 thrillers, fewer middling 3-1 results). Watch the score distribution, not the headline match number, to see the effect.
Related tools
Once you have a per-game probability, the Playoff Series Probability Calculator does the best-of-N folding for any sport. For rating-based matchups, the Elo Rating Calculator converts two ratings into an expected score you could use as a starting point. The same serve-to-match logic appears in the Volleyball Match Probability Calculator, the Badminton Match Probability Calculator and the Squash Match Probability Calculator. For goal-based sports, see the Soccer Score Simulator, the Penalty Shootout Probability Calculator and the League Season Simulator.
Frequently asked questions
Where do I get a player's serve-point win percentage?
From match statistics if you have them, or by estimating. For two evenly matched club players, start around 53% each and adjust. Remember the elite range is only 52 to 56 percent, so anything above 60% implies a large skill gap.
Why is my match odds number higher than my game odds number?
Because best-of formats reward consistency. Winning three games out of five needs the favourite to be lucky less often than winning one game does. With the defaults, game odds are 57.4% and best-of-5 match odds are 63.7%.
How often does a game reach deuce?
More often than most people guess. For two players near 53% serve each, roughly one game in five reaches 10-10. The closer the players, the higher that figure climbs.
Does it matter who serves first?
Very little across a match, because servers alternate by game. The tool averages over the coin toss for game 1, so the reported number does not depend on the result of the toss.
Is the independent-points model realistic?
It is a good first approximation for the match favourite and match odds. It understates the spread of game scores, since real points cluster. Use momentum mode to see how much wider the distribution can get.
Related tools
Elo Rating Calculator
Expected score and rating change for any two Elo ratings and K-factor.
Playoff Series Probability Calculator
Chance of winning a best-of-3/5/7 series from your single-game win probability.
Soccer Score Simulator (Poisson)
Win/draw/loss and most likely scorelines from two expected-goals numbers.
Badminton Match Probability Calculator
From serve-rally win % to game and match probability with the 30-point cap.
League Season Simulator
Monte Carlo the rest of the season: title, top-N and relegation odds from the current table.
Squash Match Probability Calculator
From serve-rally win % to game and match probability with PAR-11 scoring.