In player development, we have two tools for increasing performance: 1) improving capacity and 2) changing shape of production. Improving capacity is improving skills. Changing shape of production is changing how an athlete deploys a fixed set of skills so that they can be more productive even without improving any skills. Each athletes has a set of outcomes that they generate when they perform a skill. These outcomes can be grouped into categories that make a shape. For servers, the possible outcomes are the results of a serve, an ace, an error, and different sorts of passes. Shape of production for servers is the share that each possible outcome of a serve has of the total. Returning to the simplified chart of outcomes from “What makes a server productive?”, a server’s shape of production is the percentage breakdown for misses, perfect passes, 3 passes, 2 passes, 1 passes, and aces. This shows the server’s body of work and provides an explanation not just of how much they produce, but also the more nuanced topology of that production.
To further understand shape, let’s consider two archetypes of servers. A location server and a power server might work out to be similarly productive, but they will have very different shapes to their production. A location server will be productive by generating a lot of 1s and 2s, and hardly ever missing, but they will have costly reps where they give up a perfect pass or a 3 pass. A power server will be productive by generating aces, 1s and 2s, and rarely giving up a perfect or 3 pass, but they will have costly reps where they miss serves. Two servers, one of each archetype, could be similar in their degree of production while generating that production in very different ways.
When coaching an athlete, changing the shape of their serving production could look like asking an athlete to be more or less aggressive. When they make that change, we expect the share of serves that result in each possible outcome to change as well.
I’d like to introduce some vocabulary: terminal serves versus in-play serves. Terminal serves are aces and errors, the event immediately after the serve is the point ending. In-play serves are overpasses, okay passes, and good passes, the event immediately after the serve is something that continues the play.
As an athlete becomes more aggressive, we would expect the share of terminal serves among their total serves to increase. As aggression increases, (aces+errors)/total attempts becomes larger. Because they are mutually exclusive, as aggression increases, we would expect the share of in-play serves among total serves to decrease. As aggression increases, (overpasses+okay passes+good passes)/total attempts becomes smaller. Among in-play serves, we would also expect the balance to shift in favor of overpasses and okay passes and away from good passes as aggression increases.
What creates difficulty is knowing how a change in aggression will affect the magnitude of these changes. We can be reasonably sure that those assumptions are right directionally, but the magnitude is a mystery. Once we have data that can make us reasonably sure of the expected outcomes and their relative probabilities for a given approach, making decisions on what approach to use and when should be simple.
Let’s think back to the second part of this series and our all-or-nothing server. Let’s change the thought experiment slightly. Instead of picking between the all-or-nothing server and another DS, what if we were picking between two serving approaches? For example, what if a server’s 10/10 aggression approach yielded 50% aces and 50% misses, while their 1/10 aggression approach yielded a 2 pass every time. That server is going back to serve, in the 5th set, down 14-13, conference title on the line. You can only tell them to serve a 10 or a 1, what do you pick?
The aggregate numbers say that you should tell them to serve a 10. The 10 serve has a 50% win probability, the 1 serve has a ~42% win probability. An approach where half of the serves will be errors is a scary thing to use when a serving error ends the season. I don’t think that there is a coach who wouldn’t agonize over the decision and who wouldn’t regret the decision if they told their athlete to be aggressive and they missed their serve to end the season. I myself have this very year taken some off my serve when serving down 1 on a game point because I didn’t want to end the tournament on a missed serve. But, I really do think (in this admittedly extreme and contrived thought experiment) that it’s the right decision to take that approach.
In future posts, I’ll talk about how to choose between serving approaches that are more nuanced than the picture I just provided here. I also mentioned earlier that decisions become easy once we have data we can be confident in to make decisions. In future posts, I’ll propose some ways to gather data we can be confident in so that when the results of our decisions make us feel bad, we can at least be secure in the knowledge that the feeling bad is caused by cognitive bias, and not having made a mistake.
Comments