StatRacket's excellent thread about "Tournament Speed" gave me the impetus to create this thread (please don't merge
. I've wanted to compare the statistics from a Wimbledon tournament from the early-mid '90s with the statistics from a recent Wimbledon but never really had the time. I availed myself of the free time I have on board airplanes and at the airport (I travel very regularly for business) and of course the hi-speed WiFi. The results are very interesting.
First: a little explanation about the chart if you are not familiar with statistical testing. I wanted to go more in-depth with the stats than just "total amount of service points won" because that statistic gives you only limited information. The most important piece of information when comparing how quickly these two tournaments played is "% of points won on 1st serve." So, I plotted every serve (and ipso facto
, return) stat for every match in both the 1994 and 2011 Wimbledons. Unless there is a walkover, each tournament contains 127 total matches and statistics for 127 X 2 = 254 different player performances.
The "Statistically Significant Difference"
column lets you know if there is a real statistical difference between the means of various categories between these two tournaments. The results for a particular statistic, say "% of First serve points won,"
from the 127 matches from Wimbledon 1994 were in one column and the same numbers for the same statistic from Wimbledon 2011 were in another column. A T-Test for two samples was run. The alpha (or P value) used was .05. This means that the test determines whether or not there is something that is causing the differences in means between the two sets of numbers with a 1-.05=.95=95% certainty. So, let's say the "% chance of randomness"
column reads that there's a 2.5%
chance of randomness between the mean 1994 "% of first serve points won"
category and the 2011 "% of first serve points won"
category. This means that there's a 2.5%
chance that something absolutely random is causing these numbers to be different, and that there is a 97.5%
chance that there is some factor or reason that is causing these means to be significantly different. If we use .05 as an alpha, it means we reject our null hypothesis (which is "There is no significant difference between the numbers") if our T critical value (after running the test) is less than .05 and we accept the null hypothesis if the t-critical value is greater than .05. The T-critical value turned out to be .025 which is less than .05 so we reject the null hypothesis that there is no difference. As you can see from the chart, most of the chances for randomness are extremely small. This means that there is SOMETHING or SOME FACTOR that is causing these means to be different. 6.4E-15 means that there is a .0000000000000064% chance that the difference in % of second serve points won was something completely random.
It was significantly easier to win points off of the first serve in 1994. It was significantly easier to hit aces and win second serve points in 2011. Breaks per set and breaks per game and break chances per game were significantly higher in 1994 compared with 2011. Players in 2011 had a significantly higher overall 1st serve % compared with players in 1994.
--Once the first serve ball was in play, it was statistically significantly easier
to win the point in 1994 compared with 2011. Was this because of the lighter balls and "quicker" and lower bouncing courts? Was it because players followed their first serves to net and then closed the point at net? In serve and volley tennis, the first serve dictates the result of the point more than in baseline play (although any style of play is dictated by the first serve, and especially so if it is a really good first serve) since a returner cannot "reset" the point with a return that lands near the baseline (even if it is 'looped' back) thereby negating the overall effect of the 1st serve. In serve and volley tennis, a good serve sets up a player to win the point in the next two or three points as a direct result of the serve and its effect. It was harder to return serves due to inferior racket and string technology. A good serve in 1994 was more difficult to return low at the volleyer's feet and it was difficult to 'stretch' the volleyer with a return. In addition, after the first volley (which could end the point after a particularly effective serve), the return had a more difficult time putting the ball back in play because of the low bounce and the inferior rackets and strings that did not allow a player to pass the volleyer by picking a ball up 8 inches from the ground and putting a lot of pace and good direction on the ball.
--What's amazing about the fact that it was much easier to win a point after connecting on a first serve in 1994 compared with 2011 is that there it was significantly
easier to hit aces in 2011 compared with 1994. In 1994, there were 1300 more
first serve balls in play but 200 fewer
aces hit. You would think the opposite were true. If it's easier to hit aces in 2011, wouldn't it be easier to win a great % of points off of the first serve? Amazingly -- NO. Could it be that players are statistically significantly taller and stronger now? The rackets? It doesn't seem to be the surface because then the % of points won off of the first serve would be higher for 2011 as well but it isn't. In fact, 1994
had a significantly higher number of instances where a player won 90% of the points off of their first serves (15 compared to 4 in 2011)
, and a higher number of players win win 80% of their first serve points even when losing
(19 to 7). In 1994, a player won less than 70%
of their first serve points a total of 67 times out of a possible 254
times and only 6 times by the winner. In 2011
, a player won less than 70%
of their first serve points 81 out of 254 times
and 10 times by the winner.
--The % second serve points won in 2011 was much, much higher than in 1994.
Players aren't serving and volleying and thus the returner isn't winning points right off of the return or the next shot. At some point, players decided that it was suicide to come in on their second serves. Was racket and string technology making it easier to return serve? Was the grass playing differently? If the surface was "faster" and lower bouncing, perhaps the return had more value, especially if you got it at the feet of the volleyer. There are no available statistics for tournaments played before 1991, but from eyeballing the stats of 1991, 1992, 1993, and the rest of the "serve and volley years," 44%-46%
seemed the norm for points won on the second serve
. Perhaps the available technology and the bounce convinced the players that they HAD to come in off of their second serves, but staying back might not have produced results worse than only 43.8%
of second serve points won, which was the mean for all players in 1994. Baseliners like Agassi and Courier did well on second serves. Agassi won 51.7% of his second serve points in 1992 and 52.5% of the 2nd serve points in 1999. Courier won 51.4% of second serve points in 1993. Both players' %s were very much above the mean, but perhaps they had skills that few others had. Sampras did very well on the second serve winning 50%, 55%, 53%, 56%, 54%, 50%, 57% of second serve points in his title years, all above the means for those respective tournaments. It would have been interesting to see Lendl in 1990 stay back more because Agassi did it in 1992 and Courier in 1993, only 2 and 3 years later and they were very successful on second serves. Then again, weather and rain and how the court was playing may have made it difficult for Lendl to do so, but 1992 was a wet and rainy Wimbledon.
--The one stat that I didn't enter was the double fault, and I wish that I had. There was a statistically significant amount more double faults in 1994 than in 2011. Players needed more on the second serve to volley, maybe they felt there was a great risk-reward ration (see observation about winning points easier off of the first serve in 1994), and so on. If the double faults in 1994 were dropped down to the 2011 levels, the % of points won on the 2nd serve for 1994 would dramatically rise. There were 60 instances of players winning 50% or more of their second serve points in 1994. There were 134 such instances in 2011, more than half of all possible instances.
A lot of these observations can be read in table form:
factor in determining the breaks per set, breaks per game, and break chances per game
was the significantly LOWER
first serve % in 1994. Had the first serve %s been the same, the breaks per set, breaks per game, and breaks chances per game numbers from 1994 would have been almost identical to 2011's numbers or maybe even less
. Is it easier to hit a first serve now due to racket technology and string technology? McEnroe often mentions that it used to be considered a good day to hit 55% of your first serves in when he was playing but now that's not good at all. If you plot the median/mean first serve %s from 1984, 1990, 1995, 2000, and today, the average first serve % is probably statistically significantly higher today than in any of those years except for maybe 2000. Then again, as mentioned previously, maybe players went for more on the first serves knowing there was a greater chance of reward. The statistics show that once the first ball was in play, it was easier to win the point in 1994. Then again, this could be the result of the aforementioned effect of a player closing the net after a first serve an thereby taking direct advantage of the first serve's effect.
There is a large difference between the % of first serve points won and % of second serve points won for 1994 Wimbledon. The gap is smaller for 2011 Wimbledon. The gap is even smaller, as one would expect, at clay tournaments. The value and expected reward of the first serve was higher at Wimbledon 1994. Here is final chart that shows some the differences between % of points won off of first and second serves.
EDITED TO ADD A LOOK AT THE HEIGHTS OF THE PLAYERS:
To add another wrinkle into the conversation, the mean heights of the players entered in Wimbledon 1994 and Wimbledon 2011 are almost identical. It could be supposed that the mean height in 1994 was somewhat skewed by the fact that grass specialists (who tended to be taller players) were in the draw in place of players who skipped the tournament although most of the top 100 was in the field.