Klaassen & Magnus's 22 Myths of Tennis - Myth 13

Zebra prints and back-to-back matches haven’t been the only strange things to appear at this year’s French Open. The washout that has been a catastrophe for the 2016 Roland Garros schedule has created a myriad of unusual pressures and demands for the players (and a few of them have been vocal about their dislike of the situation). As a consequence of all this, player adaptability and grit have taken higher places in the list of skills needed to advance to the next round than would be the case in drier conditions.

Yet even under more normal conditions, when we can set aside the X factor of weather, there is still a lot of debate over the most important ingredients to being a champion. In this week’s myth, Klaassen and Magnus attempt to unravel part of this debate by pitting first and second serve ability against each other.

Myth 13: “A player is as good as his or her second service”

Stating that a player is only as good as his or her second service is simply a way of saying that the second serve is a better measure of a player’s quality than the first serve. Clearly, servers are more vulnerable in a second serve situation, which might suggest that the players who manage to win second serve points more often are going to also tend to be the players who are winning more matches overall. On the other hand, players with a high percentage of good first serves might get away with being as or even less effective on second serve points. So it isn’t immediately obvious whether we should defend myth 13 or not.

When K&M investigated the question in Analyzing Wimbledon they use several approaches and each one frustratingly comes up with a different answer. The differences boil down to how K&M measure player quality. When they use seeding to account for quality, they observe a stronger relationship between first serve performance than second serve performance. However, when they correlate first and second serve performance with the difference in player rankings they find the opposite relationship for he men–second serve being more different for an equal difference in player rankings than the first serve–and no difference between the first and second serve correlations with ranking for the women’s game. Again, all of these results are based on performance at Wimbledon only.

Revisiting Myth 13 for the ATP

With the discrepancies in their results, K&M throw up their hands and conclude that their isn’t strong evidence either for or against the superiority of the second serve as an indicator of player quality. This is one of a number of questions that have been quite sensitive to how one goes about actually measuring quality.

As in the past, when it comes to accounting for quality, my preference is to use player Elo ratings over official rankings, as Elo does a better job of giving player’s appropriate credit for big wins as well as appropriately penalizes them for surprising loses. If myth 13 is right, we would expect a stronger correlation between the difference in the Elo ratings and their second serve points won than their first serve points won. What do the past 5 years of ATP and WTA match data show?

Figure 1 summarizes the differences between the player Elo (y-axis) and first and second serve percentage of points won for the ATP. The top row shows the relationship for all matches between 2010 and 2015, while the bottom row only considers matches at Wimbledon in order to be more comparable with K&M’s analysis. Looking first at all matches, we see that both aspects of serve have a positive correlation with player quality differences, as measured by Elo, but there is a somewhat stronger relationship for first serve. In fact, the correlation with the first serve is 0.35 (95% CI = 0.33-0.37) compared to 0.31 for the second serve (95% CI = 0.29-0.34).

Thus, first serve appears to have a slight edge over second serve in indicating player quality, turning myth 13 on its head.

But at Wimbledon, the story is different. Both serves appear to have a statistically equal relationship, first serve having a correlation of 0.36 (95% CI = 0.29-0.43) and second serve 0.39 (95% CI = 0.32-0.45). There is even some suggestion that second serve is more important than first on grass, which would be consistent with K&M’s conclusions.

Revisiting Myth 13 for the WTA

What about the women’s game where first serve is generally less dominant overall? When we run the same Elo-based analysis for the WTA we find even stronger evidence that the first serve is a stronger indicator of quality across all matches. Figure 2 below shows a correlation of 0.35 (95% CI = 0.32-0.37) with first service performance and quality compared to a correlation of just 0.26 (95% CI = 0.24-0.28) for second service.

Unlike the men’s game, this relationship is virtually the same at Wimbledon. So, all together, this suggests that first serve is generally the more important factor in differentiating winners and losers in professional tennis but second serve becomes more important when surfaces are fast and the overall ability on first service is high.

All of this is likely to be moot for the finals of this year’s Roland Garros, where the players who will walk away with the trophies are most likely to be the ones who can best weather multiple days of consecutive matches and strike best with a rain-weary, heavy ball.