Should the strongest serve first? Strategic ordering of servers in doubles tennis

ABSTRACT

Do athletes adapt to changing conditions within the match, and how? Can their strategic choice of server lead to a better outcome in a match? Leveraging point-by-point data on Grand Slam tennis doubles, we show that better historical serving performance in singles matches leads to a higher probability of being assigned as the ‘first server of the match’ in doubles. We also show adaptive athletes’ behaviour during doubles matches: poorer serving performance in the previous set leads to a higher probability of being replaced as a ‘first server’ in the following set by the teammate. Both of those strategic choices lead to performance improvement: up to $5\mathrm{\%}$ more points can be won if the strongest athlete serves first.

KEYWORDS:

• Team formation
• doubles tennis
• servers ordering
• strategic decision

JEL CLASSIFICATION:

• Z20

Acknowledgments

The author is grateful to Dr. Dmitry Dagaev for his valuable help and fruitful discussions. Also, the author thanks the audience of the research seminar of the Laboratory of Sports Studies, HSE University (August 1, 2022)

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ Assume Team A (players A1, A2) and Team B (players B1, B2). If team A is serving first, then the sequence of serving players would be as follows: A1 $\to$ B1 $\to$ A2 $\to$ B2 $\to$ A1 $\to$ B1 $\to$ etc.

² At least three years of previous singles matches data have been used. For example, for athlete $i$, who was playing doubles at 2022 Wimbledon (27 June − 10 July), singles matches data from July 2019 until June 2022 have been used. Different rolling-window lengths have been used in the sensitivity analysis.

³ A pair is included in the data set if and only if each athlete from the pair participates in more than 10 singles games.

⁴ Previous set to the one when changing serving sequence is determined.

Additional information

Funding

Support from the Basic Research Program of the National Research University Higher School of Economics is gratefully acknowledged.