As a Davis Cup captain and a high-performance coach, a lot of players ask me how they could improve their shots or some other aspects of their game. Recently, more tennis enthusiasts and people who want to know more about tennis have been asking me whether I could share my analysis, so I decided to write a blog. This way I can present my views on tennis since these insights could be interesting for a wide range of tennis fans. In my first post, I will try to answer the following question:
“What is more important for the first serve, the speed or the accuracy?”
The motivation for this analysis was a player approaching me and saying that he has a great first serve, but feels that he could make it even better. Since approximately 70% of points end within the first four shots, the importance of improving the first serve even marginally is clear.
The analysis in this post is (mostly) based on Hawk-Eye data obtained for his four matches from his recent tournament on a hard court. My focus was therefore to improve his first serve effectiveness. But what is (the result of) an effective serve?
For example, Carlos Moya described that the idea behind changing Nadal’s serve was based on the fact that his first serve did not maintain speed after the ball made contact with the court. An improved technique made his serve more effective. (For my player, this was not the problem, because the ball rotation on his serve is different.)
But no matter the approach, we want the first serve to result in a weak return or (even better) no return at all. But what is a weak return? For this analysis, we will assume that the slower the return, the weaker it is. This is a fair assumption, since (as seen below) the return speed nicely correlates with the server winning the point.
We now need to figure out what kind of serve we should serve if we want a slow return or no return at all. Do we focus more on accuracy or more on speed? Let’s analyze the data and see what it tells us.
To get a feel of our player’s first serve let’s look at where the serves land if they cross the net and what are the serve speeds for serves in various directions. All opponents were right-handed.
We can see that the green and blue dots are good serves while the red and orange are the missed ones. For the speed of the serve, we can see an average speed and maximum speed for each direction. Since he is right-handed, the deuce wide serve is slower due to adding a lot of slice rotation.
But how does the serve speed influence the speed of the return? If we just draw speed of the serve vs. the speed of the return, that doesn’t tell us much, because speed is not equally important for all serves and all directions.
Instead, let’s focus first on the return speed based on the serve accuracy. Later, we will also include the speed of the serve with the combination of serve locations.
Serve accuracy vs. return speed
To begin with, let’s look at how return speed is connected with serve location. For each serve, I drew a bubble with its size indicating the speed of the return. Aces and serves where the opponent just touched the ball, have the return speed equal to zero and I marked these serves with red dots.
We can see straight away that serves close to the line result in slower returns or no returns at all. To show this in numbers, I divided each service box into five zones as shown below. I divided the wide and the T serve into two zones, one for very accurate serves (VW and VT) and the other for less accurate ones. For the direction down the T, a serve is marked as very accurate if the ball landed at most 40 cm from the line. For the wide serve, I drew a diagonal line at approximately 45 degrees to distinguish between serves. For each zone, I calculated the average serve speed as well as the average and median return speed.
For each direction, the difference in average serve speed between accurate and inaccurate serves is very similar. But the difference in return speed between accurate and inaccurate serves is quite significant.
Combining serve speed and serve accuracy
The following figures combine the serve accuracy and the serve speed with the return speed for wide and T serves for both sides. Again, the size of the dot indicates the speed of the return.
We can see several interesting things on these diagrams for our player’s serve:
- Accurate serves near the lines result in slow(er) returns no matter the speed of the serve.
- If serves to the forehand are inaccurate, the speed of the return is the highest, no matter the speed of the serve.
- It is hard to see a significant correlation between faster serve and slower return for either serve.
- Wide serves on deuce side are not very accurate. To put this accuracy into context I calculated how many good wide serves on deuce side were actually very accurate. For our player only 42% of these serves were accurate. For his opponents, the number of accurate serves for this direction was 70%. This is quite a difference.
- Further analysis shows our player favors the T serve on deuce side with 62% of serves hit down the T. The reason for that may be in the inaccuracy of wide serve, since inaccurate serves result in good (forehand) returns. But the problem with an unbalanced serving pattern is that an opponent can adapt and focus his return more on serves hit down the T. By improving his wide serve, he would also make his serves down the T more dangerous.
Comparing with opponents
Next, I wanted to compare different serves and also to compare his serve speed vs. serve accuracy to other players. In order to do that, I fitted the lines that describe the data points best (minimized the squared error). The first presented graphs are for our player.
No matter the serve direction and the serve speed, the return speed does not change much. We can see that faster serves typically result in slower returns, but the correlation is low.
Focusing on serve accuracy, we see that lines here are more vertical. Especially if the serves fall closer to the line than 60 cm, the accuracy and return speeds are greatly correlated.
The graphs for his opponents look slightly different.
We can see the average return speeds are slower and the correlation between serve speed and return speed is greater. The lines for serve accuracy vs. return speed have very similar shapes to our player’s and the return speeds are again higher, which indicates that our player’s serves result in slower returns than the average of his opponents. One reason for this can be his quick service motion with a low ball toss that is very hard to read and thus to predict the serve direction.
As we have shown here, analyzing serve speed and serve accuracy is player-specific. For our player, the advice was to focus more on serve accuracy than serve speed. But again, this is true for him and not necessarily for other players.
But this kind of analysis is just the beginning. With access to more Hawk-Eye data, many more great insights can be revealed. Through analytics, the understanding of the tennis game will improve in the same way that it happened with other sports where using data science to analyze your and your opponents’ game is now necessary to stay competitive.
If you liked the article, give it as many claps as you like (up to 50). This will increase the chances for this article to reach wider tennis and (sport) analytics communities.