Insight: Win probabilities can be misleading. Do not use them without context to determine whether a tactic is effective.
You played a hard-fought match. Your coach charted it, and the two of you sit down to pore over the serve percentages, break points saved, and rally lengths. But of particular interest are the various win probabilities. The goal, after all, is to win, and those probabilities tell you which tactics, from wide serves to approach shots, are winning tactics. Or do they?
As it turns out, not always. Sometimes, win probabilities can be misleading and even lead to the wrong advice being given. For example, say you won 50% of your baseline points. Not bad. But at net, you won 70% of the time! In that case, you’d better start crashing the net! However, a deeper look at the data reveals something else. When you came in, you did so on short, easy balls. But you only get a limited number of those, and you exhausted them already. Therefore, if you start coming in more often, you won’t keep winning 70% of the time. You might not even win 50% of the time.
This is where causal inference comes in. It’s the way to draw conclusions about why the data are the way they are. The way to determine whether you should come in or stay back. The way to turn numbers into tactical advice you can take on court. But how?
Let’s take a journey down memory lane to the 2016 Wimbledon Men’s Final. Andy Murray the crafty counter-puncher vs. Milos Raonic the titanic server. A hip injury nearly forced Murray into retirement, but now he’s mounting an inspirational comeback. So what better match to study, especially given the wonderful contrast of styles, than the last Slam final he won? 10 July 2016. The day Sir Andy stood atop the tennis world.
