Tennis scores represent a natural language domain that offers the unique opportunity to study the effects of discourse constraints on prosody with strict control over syntactic and lexical variation. This study analyzed a set of tennis scores, such as 30–15, from live recordings of several Wimbledon and Davis Cup matches. The objective was to uncover how the prosody of these scores depended on three factors: (a) the local discourse context, which varied because, after every point, either the first or the second number was different than it had been in the score reported just previously; (b) the global situation, which varied because either the serving player or the receiving player had the lead in the current game; and (c) the speaker role, which varied as the umpire or a radio reporter uttered the scores. This study excised 96 tennis scores (spoken in English) by 2 umpires and 2 live radio reporters. These scores were labelled and perceptually evaluated in terms of perceived prominence and acoustically analyzed in terms of three prosodic characteristics: loudness, speech rate, and pitch. The perceptual and acoustic analyses both showed that the prominence patterns were affected by the global situation and speaker role, but not by the local context. For both speaker roles, prosodic prominence patterns correlated with whether the game, so far, was proceeding favorably for the serving player. In the umpire data, a favorable situation correlated with extra emphasis on the first word in a score, whereas in the reporter data, an unfavorable situation correlated with more emphasis on the second word.