Tennis research reading list: 24 papers, six questions

Published · 24 papers in 6 groups

Twenty-four journal papers on tennis, grouped by the question each one studies. Each entry gives the full citation, a DOI link and the data the paper says it used.

It also says, in one line, whether the question can be re-run on the public sample of our point-by-point dataset. One BibTeX file at the end holds every paper and the dataset.

Download the BibTeX See the six questions

The six questions

  1. Are points independent and identically distributed? 4 papers
  2. How big is the serve advantage, and does serving first matter? 4 papers
  3. Do momentum and streaks exist? 5 papers
  4. Do players perform differently on important points? 4 papers
  5. How well do rating systems predict matches? 4 papers
  6. How long do matches last? 3 papers

What the public sample holds

The public sample of our point-by-point dataset is on Zenodo at doi:10.5281/zenodo.22048730. We read version 1.0.3 of the record on 18 November 2026.

Its point file holds 951,064 score states from 5,380 matches of June 2026, on the ATP, WTA, Challenger and ITF tours.

Each row gives the sets, games and points, who served, a tiebreak flag and the time the state was captured.

The match file lists 173,571 matches up to August 2026, with tournament, tier, surface, best-of and players. It has no scores and no winners.

Nothing in the sample gives rankings, odds, serve speed, first or second serve, rally length or shot detail.

A tape can start after the first point or stop before the last, and some rows have no server.

How to read the re-run line

A Yes means the question can be asked of the sample. It does not mean the answer will match the paper.

We were strict. Where a question needs complete tapes, set results or a season of history, we called it Partly or No.

Our re-test page runs simpler tests of three of these papers' questions on this sample and discusses a fourth. Their entries link to it.

Are points independent and identically distributed?

Suppose each player wins points on serve at a fixed rate, independent of every other point. Then those two rates settle the chance of winning the match. These papers test or use that idea.

  1. Franc J. G. M. Klaassen and Jan R. Magnus (2001). Are Points in Tennis Independent and Identically Distributed? Evidence From a Dynamic Binary Panel Data Model. Journal of the American Statistical Association 96(454), 500-509.

    doi:10.1198/016214501753168217

    Data
    Almost 90,000 points from Wimbledon matches, 1992 to 1995.
    On the sample
    Partly Our re-test ran simpler forms of both tests on this sample, the point after a won point and break points. See the re-test
  2. Paul K. Newton and Joseph B. Keller (2005). Probability of Winning at Tennis I. Theory and Data. Studies in Applied Mathematics 114(3), 241-269.

    doi:10.1111/j.0022-2526.2005.01547.x

    Data
    The 2002 US Open and Wimbledon tournaments, used to validate the theory and to show how a champion can be predicted.
    On the sample
    Partly Serve points won can be counted in games recorded without a gap, but not reliably for a whole match.
  3. Paul K. Newton and Kamran Aslam (2009). Monte Carlo Tennis: A Stochastic Markov Chain Model. Journal of Quantitative Analysis in Sports 5(3).

    doi:10.2202/1559-0410.1169

    Data
    Player and whole-field statistics from the 2007 men's ATP circuit.
    On the sample
    Partly Serve and return points won can be counted match by match, but over one month, not a season.
  4. Min Kyu Sim and Dong Gu Choi (2020). The Winning Probability of a Game and the Importance of Points in Tennis Matches. Research Quarterly for Exercise and Sport 91(3), 361-372.

    doi:10.1080/02701367.2019.1666203

    Data
    Point-by-point records of ATP men's singles matches, 2011 to 2015.
    On the sample
    Partly Score and server are in the rows, but only gap-free games can be used, and some deuce games are hard to read.

How big is the serve advantage, and does serving first matter?

These papers measure the server's edge and ask whether the order of serving in a set or a tiebreak changes who wins.

  1. Jan R. Magnus and Franc J. G. M. Klaassen (1999). On the Advantage of Serving First in a Tennis Set: Four Years at Wimbledon. Journal of the Royal Statistical Society: Series D (The Statistician) 48(2), 247-256.

    doi:10.1111/1467-9884.00186

    Data
    Almost 90,000 points from four years of Wimbledon.
    On the sample
    Partly Our re-test ran the serving-first question on this sample, but first and second serves are not recorded. See the re-test
  2. Jan R. Magnus and Franc J. G. M. Klaassen (1999). The Effect of New Balls in Tennis: Four Years at Wimbledon. Journal of the Royal Statistical Society: Series D (The Statistician) 48(2), 239-246.

    doi:10.1111/1467-9884.00185

    Data
    Four years of Wimbledon, including the 1995 championships, which were played with softer balls.
    On the sample
    No The sample marks neither ball changes nor first and second serves.
  3. Danny Cohen-Zada, Alex Krumer and Offer Moshe Shapir (2018). Testing the effect of serve order in tennis tiebreak. Journal of Economic Behavior & Organization 146, 106-115.

    doi:10.1016/j.jebo.2017.12.012

    Data
    1,701 men's and 920 women's tiebreaks from top-tier tournaments, 2012 to 2015.
    On the sample
    Partly Tiebreaks are flagged with a server, but some tiebreak rows lack the server and some tapes stop early.
  4. Silvia Montagna, Vanessa Orani and Raffaele Argiento (2021). Bayesian isotonic logistic regression via constrained splines: an application to estimating the serve advantage in professional tennis. Statistical Methods & Applications 30(2), 573-604.

    doi:10.1007/s10260-020-00535-5

    Data
    Grand Slam singles match data, used to model the serve advantage by rally length.
    On the sample
    No The sample has no rally length.

Do momentum and streaks exist?

Does winning a point, a game or a set change the chance of winning the next one?

  1. David Jackson and Krzysztof Mosurski (1997). Heavy Defeats in Tennis: Psychological Momentum or Random Effect? CHANCE 10(2), 27-34.

    doi:10.1080/09332480.1997.10542019

    Data
    The published record has no abstract. The paper's opening sets its subject as the sets within best-of-five-set matches.
    On the sample
    Partly Set scores are in the rows, but few June tapes are best of five. Our re-test discusses this paper and does not test whole sets. See the re-test
  2. Jingya Wang, Sihang Guo and Yuanyun Zhou (2024). A multidimensional momentum chain model for tennis matches based on difference equations. PLOS ONE 19(12), e0316542.

    doi:10.1371/journal.pone.0316542

    Data
    7,284 points from the Wimbledon 2023 men's singles, and 6,870 matches to test the model.
    On the sample
    Partly Our re-test ran a simpler streak test on this sample, but the running distance and net play it reports are not in the sample. See the re-test
  3. Philippe Meier, Raphael Flepp, Maximilian Ruedisser and Egon Franck (2020). Separating psychological momentum from strategic momentum: Evidence from men's professional tennis. Journal of Economic Psychology 78, article 102269.

    doi:10.1016/j.joep.2020.102269

    Data
    4,930 games from 141 Grand Slam men's singles matches.
    On the sample
    Partly Breaks and the games after them are in the tapes, but the interruptions the paper relies on are not marked.
  4. Craig A. Depken, John M. Gandar and Dmitry A. Shapiro (2022). Set-level Strategic and Psychological Momentum in Best-of-three-set Professional Tennis Matches. Journal of Sports Economics 23(5), 598-623.

    doi:10.1177/15270025221085715

    Data
    66,262 best-of-three professional matches, 2002 to 2020.
    On the sample
    Partly Earlier set scores are in every row, but a tape that stops early loses the last set, and nothing measures the skill gap.
  5. Peggy A. Richardson, William Adler and Douglas Hankes (1988). Game, Set, Match: Psychological Momentum in Tennis. The Sport Psychologist 2(1), 69-76.

    doi:10.1123/tsp.2.1.69

    Data
    163 matches played by 119 male and female players at three sanctioned tournaments, with game-by-game results and player interviews.
    On the sample
    Partly Game-by-game results can be read from the tapes, but the winner only where a tape runs to the end, and there are no interviews.

Do players perform differently on important points?

Does the chance of winning a point change when the point decides more, such as a break point or a game that ends a set?

  1. Peter G. O'Donoghue (2001). The Most Important Points in Grand Slam Singles Tennis. Research Quarterly for Exercise and Sport 72(2), 125-131.

    doi:10.1080/02701367.2001.10608942

    Data
    Points from 252 men's and women's singles matches at all four Grand Slams over two years.
    On the sample
    Partly Serve points can be counted at scores before deuce, but June 2026 holds only two Grand Slams, the French Open and Wimbledon.
  2. Gareth Knight and Peter O'Donoghue (2012). The probability of winning break points in Grand Slam men's singles tennis. European Journal of Sport Science 12(6), 462-468.

    doi:10.1080/17461391.2011.577239

    Data
    528 men's singles matches from all four Grand Slams in 2008 and 2009.
    On the sample
    Partly Break points before deuce can be read from the score, but June 2026 holds only two Grand Slams, the French Open and Wimbledon.
  3. Julio González-Díaz, Olivier Gossner and Brian W. Rogers (2012). Performing best when it matters most: Evidence from professional tennis. Journal of Economic Behavior & Organization 84(3), 767-781.

    doi:10.1016/j.jebo.2012.09.021

    Data
    Point-level data from twelve years of the US Open.
    On the sample
    Partly How much each point matters can be worked out from the score, but one month cannot show a career.
  4. Brian Goff, Dennis Wilson and David Zimmer (2025). Performing in high-pressure situations: the case of tennis players up a break and serving for the set. Applied Economics 57(39), 6138-6147.

    doi:10.1080/00036846.2024.2381129

    Data
    Over 136,000 games, July 2021 to September 2022.
    On the sample
    Partly Pressure games and holds can be read from the game scores and the server, but the sample holds few player attributes.

How well do rating systems predict matches?

Elo, Glicko, Bradley-Terry models and the official rankings, compared as forecasts of who wins.

  1. Ian McHale and Alex Morton (2011). A Bradley-Terry type model for forecasting tennis match results. International Journal of Forecasting 27(2), 619-630.

    doi:10.1016/j.ijforecast.2010.04.004

    Data
    ATP tour results and surfaces, with official rankings and ranking points for the comparison models.
    On the sample
    No The match file holds no winners or rankings, and one month of tapes is too short a history.
  2. Leighton Vaughan Williams, Chunping Liu, Lerato Dixon and Hannah Gerrard (2021). How well do Elo-based ratings predict professional tennis matches? Journal of Quantitative Analysis in Sports 17(2), 91-105.

    doi:10.1515/jqas-2019-0110

    Data
    Every match at Wimbledon 2018 and 2019, the 2019 French and US Opens and the 2020 Australian Open. It compares odds, rankings and Elo.
    On the sample
    No The sample has no odds or rankings, and its results cover June 2026 only.
  3. Giovanni Angelini, Vincenzo Candila and Luca De Angelis (2022). Weighted Elo rating for tennis match predictions. European Journal of Operational Research 297(1), 120-132.

    doi:10.1016/j.ejor.2021.04.011

    Data
    Over 60,000 men's and women's professional matches.
    On the sample
    No A rating needs a long run of results, and the sample's results cover one month.
  4. Jack C. Yue, Elizabeth P. Chou, Ming-Hui Hsieh and Li-Chen Hsiao (2022). A study of forecasting tennis matches via the Glicko model. PLOS ONE 17(4), e0266838.

    doi:10.1371/journal.pone.0266838

    Data
    Men's and women's singles matches at the Grand Slams, 2000 to 2019.
    On the sample
    No The sample has no rankings, and its results cover one month.

    PLOS ONE published a correction on 2 December 2024 that adds an author affiliation.

How long do matches last?

How long matches take, in points and in time, and what changes that.

  1. Stephanie Ann Kovalchik and Martin Ingram (2018). Estimating the duration of professional tennis matches for varying formats. Journal of Quantitative Analysis in Sports 14(1), 13-23.

    doi:10.1515/jqas-2017-0077

    Data
    Ball-tracking data and public shot-level statistics, which feed a match simulation.
    On the sample
    Partly Points per match need a tape with no gaps, which the sample cannot prove, and capture times are not shot timings.
  2. Otto Kolbinger, Simon Großmann and Martin Lames (2019). A closer look at the prevalence of time rule violations and the inter-point time in men's Grand Slam tennis. Journal of Sports Analytics 5(2), 75-84.

    doi:10.3233/jsa-180277

    Data
    3,475 serves from the 2016 Australian Open men's singles.
    On the sample
    No Capture times mark when a score was seen, not the end of a rally or the next serve, and umpire actions are not recorded.
  3. Chengzhi Wang and Steve Drekic (2026). Boosting markovian tennis prediction: Ensembling and point-specific methods for match outcome and duration. Journal of Sports Analytics 12, article 22150218251412670.

    doi:10.1177/22150218251412670

    Data
    Men's and women's professional matches from 2011 to 2022, with 2014, 2018 and 2022 held out for testing.
    On the sample
    Partly June 2026 serve rates are in the tapes, but not the years of history the models learn from.

How we built the list

We found every paper by searching the Crossref index by topic. Then we resolved each DOI at doi.org and checked its title, authors, year and journal against the Crossref record.

The checks ran on 18 November 2026. The year shown is the year of the printed issue where Crossref has one, so a few papers appeared online a year earlier.

Each data line comes from the paper's abstract and title, in our words. One paper has no abstract on record, so its line comes from the paper's opening section.

We do not summarise what any paper found. For that, read the paper.

Related: tennis findings, re-tested runs simpler tests of three of these questions on our data, with the code. The point-by-point dataset is free for non-commercial academic research.

Download the BibTeX

One file holds all 24 papers and, at the end, the citation for our dataset. That last entry is the one the academic dataset page gives.

Download the BibTeX file

@dataset{livetennisapi_2026_tennis_pbp,
  author    = {{Live Tennis API}},
  title     = {Tennis Point-by-Point Dataset --- ATP, WTA, Challenger, ITF},
  year      = 2026,
  publisher = {Zenodo},
  doi       = {10.5281/zenodo.22048730},
  url       = {https://doi.org/10.5281/zenodo.22048730}
}