Previous post: Serving Series Part 1: Your team needs to miss more serves
Let us imagine a server for whom the only outcomes of their serves are missed serves and aces. Let us call them the “all-or-nothing server.” What is the highest percentage of their serves that can be missed such that they are still playable? What percentage would make them the best server on your team?
Given sufficient data for the team they play on, we could answer these questions for our all-or-nothing server. “Will playing server A over server B make my team more likely to win?” is a question that coaches are or should be invested in asking. In this post, I’ll be talking about what smaller questions make up “Will playing server A over server B make my team more likely to win?”. This will provide some shape for what kinds of statistics we should be gathering and looking at to help us make decisions.
To be able to answer “Will playing server A over server B make my team more likely to win?” we need to be able to answer “What makes a server productive?” “How do we meaningfully measure serving production?” “How do we compare measurements of serving production?”
Servers are productive insofar as they contribute to scoring points. If we are more likely to score a point after the serve lands or after the first contact on the serve than we were at the start of the point, the server has been productive. If we are less likely, the server has hurt us. You can find an in-depth discussion of expected value in volleyball on Chad Gordon’s blog. For our questions about the all-or-nothing server, we just need eV numbers. I think this is more intuitive to think about in win probability terms, so I convert Chad’s eV numbers into win probabilities using “(eV+1)/2=WP.” The lowest ace percentage that the all-or-nothing server is playable with is the win probability of the team’s worst server. If the all-or-nothing server improves on that, they should play. The lowest ace percentage at which they’d be the best on the team is the win probability of the best server. If they have the best win probability, they’re the best.
“How do we meaningfully measure serving production?” is more difficult. Gordon in his blog has access to a lot of data and puts a lot of work into analyzing that data, which one has to do to be able to know if a team is more or less likely to win a point after a given event.
“How do we compare measurements of serving production?” is also difficult. Some amount can be “big number good.” But, we will also need to be mindful of shape of production, which we’ll talk about more when the time comes.
Comments