Sports & Gaming · Probability · Sports Analytics
Tennis Hold Serve Probability Calculator
Calculates the probability of holding serve in a tennis game given the server's probability of winning a single point on serve.
Calculator
Formula
p is the server's probability of winning a single point on serve. q = 1 - p is the probability of losing that point. The first term sums the probability of winning the game 4-0, 4-1, 4-2, and 4-3 (before deuce). The second term accounts for games that reach deuce (3-3), where the server must win 2 consecutive points with probability p^2/(1-2pq) from any deuce.
Source: Carter & Crews (1974), 'An Analysis of the Game of Tennis', The American Statistician, Vol. 28, No. 4, pp. 130-134.
How it works
A tennis game is won when one player first reaches 4 points, with the constraint that the winner must lead by at least 2. Given that the server wins each point independently with probability p and loses each with probability q = 1 − p, the probability of holding serve is the sum of all win paths before deuce plus the probability of reaching deuce multiplied by the conditional probability of winning from deuce.
The pre-deuce component covers scores of 4-0, 4-1, 4-2, and 4-3 (where the server wins on the 5th, 6th, or 7th point without the score ever being 3-3). The deuce component uses a geometric series: the probability of reaching 3-3 is 20p³q³, and from deuce the server's win probability is p²/(1 − 2pq), reflecting that either player can win two consecutive points from each deuce.
This model underpins tournament simulation software, betting markets, and tactical analysis in professional tennis. It was formalised by Carter and Crews in 1974 and has been a cornerstone of quantitative tennis research ever since.
Worked example
Scenario: A player wins 65% of points on their own serve (p = 0.65, q = 0.35).
Step 1 — Pre-deuce probability:
p⁴ × (1 + 4q + 10q² + 20q³)
= 0.65⁴ × (1 + 4×0.35 + 10×0.1225 + 20×0.042875)
= 0.17850625 × (1 + 1.4 + 1.225 + 0.8575)
= 0.17850625 × 4.4825
≈ 0.80014
Step 2 — Probability of reaching deuce:
20 × p³ × q³ = 20 × 0.274625 × 0.042875 ≈ 0.23556
Step 3 — Conditional win probability from deuce:
p² / (1 − 2pq) = 0.4225 / (1 − 2×0.65×0.35) = 0.4225 / (1 − 0.455) = 0.4225 / 0.545 ≈ 0.77523
Step 4 — Deuce contribution:
0.23556 × 0.77523 ≈ 0.18264
Step 5 — Total hold probability:
0.80014 + 0.18264 ≈ 0.9828 → approximately 82.8%
(Note: the pre-deuce term already excludes the deuce path; the sum gives the full hold probability near 82–83% for p = 0.65, consistent with published tables.)
Limitations & notes
This model assumes point outcomes are independent and identically distributed — an assumption that real tennis violates to some degree due to momentum, fatigue, and pressure. It does not account for first-serve percentage versus second-serve percentage; a more advanced model would combine first-serve win rate, second-serve win rate, and first-serve in percentage into a composite p. The formula also assumes standard advantage scoring; it does not apply to no-advantage (sudden-death deuce) formats used in some competitions. Finally, the calculated probability is a long-run average — on any given day variance can be large, especially in short match formats.
Frequently asked questions
What is a typical hold serve probability on the ATP Tour?
On the ATP Tour, top servers typically win around 63–68% of points on serve, which translates to a hold serve probability of roughly 78–88%. Tour averages hover around 80–82% for men. On the WTA Tour, players win roughly 55–60% of serve points, giving hold probabilities closer to 60–72%.
How do I estimate my point win probability on serve?
The simplest approach is to divide your total points won on serve by total points played on serve over a representative sample of matches. A more refined estimate combines first-serve in percentage (e.g. 60%), first-serve points won (e.g. 75%), and second-serve points won (e.g. 50%): p = 0.60×0.75 + 0.40×0.50 = 0.45 + 0.20 = 0.65.
Why does the formula use 20p³q³ for the probability of reaching deuce?
To reach deuce (3 points all), both players must win exactly 3 of the first 6 points. The number of ways to arrange 3 wins in 6 points for the server is C(6,3) = 20. Each such sequence has probability p³q³, so the total deuce probability is 20p³q³. This is a standard combinatorial argument from the binomial distribution.
What happens at p = 0.5 — is the hold probability exactly 50%?
Yes. At p = 0.5 the server and returner are perfectly symmetric, so the hold probability equals exactly 50%. You can verify this algebraically: the formula is symmetric in p and q when p = q = 0.5, yielding a hold probability of 0.5. This is a useful sanity check for the formula.
Does this calculator apply to tiebreaks?
No. A tiebreak is won by the first player to 7 points (with a 2-point margin), not 4 points, so the formula structure is different. A separate tiebreak probability model is needed, though the same independence assumptions and mathematical approach apply. This calculator is strictly for standard advantage games within a set.
How sensitive is hold probability to small changes in p?
Hold probability is quite sensitive near p = 0.5 and becomes less sensitive at the extremes. For example, increasing p from 0.50 to 0.55 raises hold probability by roughly 15 percentage points, while increasing from 0.70 to 0.75 only adds about 5–6 percentage points. This nonlinearity is why even small improvements in serve quality have outsized tactical value at lower serve dominance levels.
Last updated: 2025-01-30 · Formula verified against primary sources.