top of page


If you watch enough tennis on TV, or even if you only watch it in January, you get used to terms like winners, unforced errors, aces and points won at the net.

As an Analyst and Strategy expert to pro and college players and educator to developing juniors, I often get asked “when you’re talking to pros, how often do you talk about the stats they show us on TV?”. The answer: not much, if at all.

Why? Largely because the bulk of the “TV stats” are all about what happened at the end of a point, without any regard for how the players reached the end of the point.

When we talk TV stats, it’s all point ending.

It’s better than it used to be

In recent years, rally length ranges have been added to some of the TV stats. Most junior players have seen rally length broken up into 3 ranges: 0-4, 5-8 and 9+. Once they start competitive play, most developing players would know that those rally lengths occur around 70%, 20% and 10% of the time in pro tennis.

The implication being that players and coaches should concentrate on developing skills in shorter rallies rather than the traditional “rally til you drop” skills so often seen on the practice court.

Even with this rally length range data, however, there’s a problem. First of all, they’re ranges that measure shots in court only ie a howling unforced error isn’t counted as a shot (part of the reason a double fault is a “zero” shot rally).

Secondly, and far more importantly, how can a junior player develop his or her own game based on those numbers?

The short answer is, that they can’t, or at least it’s very narrow and limiting. It’s what we call “big data”. That is, saying that “70% of all pro tennis rallies finish on shots 0-4” doesn’t actually tell us how Thanasi Kokkinakis plays the sport versus how Diego Schwartzman plays; or how does Simona Halep’s serve strategy differ from Iga Swiatek?

All it says is “here’s the range based on tens of thousands of shots and, by the way, unforced errors don’t count because the ball didn’t go in”.

The Story of a Point

If we ask a child to read a book, we don’t tell them to turn to the last page and read “THE END”. Yet that’s what TV stats do. The top 3 pieces of information delivered on TV are:

  1. What was the final shot type?

  2. Was the final shot a winner or an unforced error?

  3. Who hit the final shot?

That’s 3 pieces of information about the end of the point only! Nothing about how the players got to the end, or whether it was a 3-shot rally or a 53-shot rally. What’s worse is that forced errors are hardly ever mentioned, so we’re not actually measuring every point in a match any way.

I digress.

Instead, as an Analyst, I want to know the following:

  1. Who was the server?

  2. How long was the rally?

  3. Who won the point?

So, like the story that we want the child to read, we have:

  1. A beginning – the server

  2. A story in the middle – the rally length.

  3. An ending – the player who won the point.

It’s from this starting point that all other information can be added, like was the server serving to Deuce or Ad for example.

How we should look at Analytics to help develop players

The server and the point winner are the easy ones in our story. Rally length is a little more detailed but bear with me.

The bulk of points in tennis finish on shot 1, shot 3 or shot 5 ie the first 3 shots played by the server in any given point. It’s no coincidence that the analytics system we established to measure player performance is called “135 Tennis Analytics”.

Then there are the returner’s first 3 shots: 2-4-6, and lastly, there’s the long rallies, which go 7 or more shots, collectively called 7+. I won’t go into the frequencies, win rates etc here. That’s for another day.

The important thing to realise is that, if we talk in server’s shots and returner’s shots, we’re talking about a sequence of shots made by a player, NOT both players. 5 is often related to 3, which is related to 1, for example. We’re not grouping in 0-4 which, because it’s a range, has nothing to do with who hit the shot.


The last thing that must be realised is that we need to measure the difference between unforced errors and forced errors. This is where a lot of people start to panic. But you shouldn’t. There are actually fewer points than you think where analysts differ over whether an error was forced or unforced.

What’s important is that, in the opinion of this analyst, the difference in those 50/50 ones is subjective. What’s unforced to Novak might be forced to a 10-year-old is one example.

Last of all – Profiling

Armed with 135, 246 and 7+ data about a player with whom I’m working, I can tell them where their game is going well and where it needs improvement, and it’s based on how the individual plays.

Whether it’s a pro player, college player or developing junior just starting out on a competitive journey, if you can profile based on 135, 246 and 7+ rally lengths, and measure forced and unforced errors, you have a method by which you can measure player performance. At the moment, all we have is a player rating and a score, both with plenty of ambiguity, and both based on match results.

As one of the coaches with whom I work closely likes to say, “what we can measure, we can improve”. If you can’t measure what’s going well and what needs improvement, how are you going to fix it?

Author’s Bio: Nicholas Scott is the Co-founder of 135 Tennis Analytics and co-developer of the 135 Tennis Analytics system. Nick works with Thanasi Kokkinakis and his coach, Todd Langman, as well as other ATP and WTA pros, college and high school tennis teams and coaches, and developing players of all ages. For more information, visit


bottom of page