Insight: When returning Williams’s serve, Wozniacki should have positioned herself closer to the T. In addition, her most effective returns were those hit down the middle.
In Part 1, I analyzed the rallies in the 2014 US Open Women’s Final. However, two of the most frequently hit shots in tennis are the serve and return. In fact, 258 of the 614 total shots in this match, or over 40% of them, were serves or returns. Furthermore, Serena Williams’s serve, which averaged 100 mph to Caroline Wozniacki’s 93.8 mph, gave Williams a considerable advantage over Wozniacki. Specifically, Williams won 65% of her service points, whereas Wozniacki won only 49% of hers. Therefore, serving or returning better may have helped Wozniacki in this match. While much of that improvement must take place on the practice court, better tactical choices can also help players overcome disadvantages.
As discussed in Part 1, I focus my specification search on specifications backed by prior knowledge. One such specification is whether second serves hit wide are more or less effective than those hit down the T. In contrast with first serves, second serves are usually intended to be returned. However, wide ones have the benefit of opening up the court for the server’s next shot. Meanwhile, the returner can compensate by positioning herself farther wide, but it is also possible for her to overcompensate and stand too far wide. As shown in Table 1, Williams hit 9 of her 23 second serves wide, where WIDE = 1 if the serve landed in the service box’s wide half. Table 2 shows her wide second serves were actually 49.2% less effective than her T serves, which is significant at the 5% level (the large coefficient makes the result significant despite a small sample size). Thus, from Wozniacki’s perspective, T second serves were more dangerous than wide ones, and she had to return them more often. As a result, shifting her returning position closer to the T may have allowed her to win more points.
I then analyze how far back Wozniacki stood on returns and where she aimed them. In Table 1, I define BEHIND = 1 when she stood behind the baseline at contact. In addition, I define DTM = 1 when her return went down the middle. I analyze down the middle returns in particular because, per my prior knowledge, they are the easiest to make and can jam a server still recovering from the service motion. Finally, I define the control variable FIRST = 1 when Williams’s serve was a first serve. FIRST is a proxy for the speed, spin, and direction of each serve because first serves tend to be faster, flatter, and placed farther from the body than second serves. It is also a proxy for Wozniacki’s starting position because Wozniacki started farther back on first serves. Per Table 2, standing inside the baseline was about 33% more effective than standing behind it. Also, returns hit down the middle were about 25% more effective than those hit to the corners, regardless of where she stood. DTM’s coefficient is insignificant due to the small sample, but since it is nearly significant and supported by my knowledge, I consider it meaningful.
Although Wozniacki was ultimately outmatched, she still won 43% of 114 total points, or only 16 fewer points than Williams. Therefore, a few tactical improvements may have made the match closer. It is also possible that additional tactics, such as approaching the net, would have worked had Wozniacki tried them. However, she only approached on four points, so even though she won three of them, it is impossible to analyze whether this tactic was truly effective. Thus, trying multiple tactics in a match not only increases the likelihood of finding a tactic during the match that works, but it also provides quantitative evidence of what worked and what did not.
Table 1: Summary Statistics
Table 2: Win Probability Regressions
