How Does 135 Differ from 0-4, 5-8, 9+?

The first answer is that rather than having short, medium and long rallies, we have the server’s rally length (135), the receiver’s rally length (246) and long rallies (7+). That helps us understand patterns of play a lot better.

There’s another crucial difference: Any “big rally length data” that you see on TV or reported by other analysts reports on balls in court only. That’s why, for example, there’s a zero rally length. It’s a double fault. There is an upside to this: it means that computers, TV networks and other analysts can use thousands of points to prove something like “most points in tennis are zero to four shots”. We’re not debating that.

The problem is how it’s applied. Consider the points in the videos below. 

They are from the same match, Stefanos Tsitsipas v Rafael Nadal. Note the patterns are almost identical and repeated often in both men’s and women’s tennis. The problem is Stefanos Tsitsipas missed the forehand on shot 3 in one of the rallies, but made the forehand on the other.

The third rally in the video (from the same match again) is Rafa hitting a clean winner from a Tsitsipas 1st serve. When the courtside computer measures rally length, that 2 shot rally is the same as the second rally. That’s where we disagree.

In 135 tennis analytics, the first two points have the same rally length: 3

In 0-4 terms, the one where Tsitsipas made the forehand was a 3 shot rally. The one where he missed was a 2, because only shots that go in are counted. 

The first implication is that in 0-4 rally length, we’re not comparing apples with apples, because clearly those two points were the same pattern. Further, in 0-4 rally terms, the point with the error is treated the same as the third point in the sequence ie Rafa hitting a winner on return.

The second implication is that the only way to measure how well a player is executing his/her 3 shot patterns is by using 135!

Unforced v Forced Errors

It should be apparent by now that we’re measuring unforced errors in our 135 rally length data. This is vital because it means that progress can be measured. For example, if a player is winning less than 50% of their 3 shot rallies, they have work to do. Typically, pros win around 65% of 3s in a winning match. We’ve consulted to juniors whose 3s are between 10%-30%.

That’s why 135 measures progress. If a player is winning 20% of their 3s in January, but improves to 40% by June, we have progress and improvement, which we can continue to measure.

So, now we know how rally length is measured and applied by 135, the question becomes: “what is the difference between a forced and an unforced error”? After all, if we’re measuring improvement, we’re going to need to be able to tell the difference.

Don’t worry, it’s nowhere near as daunting as it sounds.