Digging Profiles - Part 2
Don't let the ball hit the floor
Related:
Serving Profiles
Blocking Profiles
Articles in this series:
Part 1 - Introducing The Data
Part 2 - Keepy Uppy
Part 3 - The Creator Economy
Part 4 - Conversions
Keepy Uppy
My brother, a supportive family member who is completely uninterested in the nuances of volleyball, once described Olympic volleyball as, “the world’s most intense game of don’t-let-the-balloon-hit-the-floor.” And you know what, he’s not wrong.
There’s a lot more nuance to studying team defense, but let’s start with the teams who were the best at keeping the ball from hitting the floor. In particular, I’m using some statistical techniques to tease out the teams who were relatively better at keeping the ball in play, as compared to creating swings and/or converting those transition opportunities.
Fitting into the 3 broad profiles we discussed in Part 1:
Teams that excel at keeping the ball off the floor
Teams that are good at creating opportunities on those digs
Teams that are good at converting those opportunities
We’ll call these 10 teams the “Keepy Uppy” teams. The strongest part of their defense was just keeping the ball off the floor. As always when we do these profiles, we see a range of team success among teams in the profile, so let’s dive into the numbers a bit more.
Winning And Losing
There’s definitely a correlation between digging the ball and Winning and Losing. In this sample, the correlation between Dig% and Win% was 0.48, which is fairly high. For example, the correlation between Ace:Error Ratio was 0.33.
Let’s look at the other correlations I pulled that specifically relate to defense/transition:
0.48 - Dig%
0.52 - Create%
0.50 - Convert%
0.35 - Digs Created%
0.26 - Creates Converted%
The 3 that we’re going to focus on the most are Dig%, DigsCreated, and CreatesConverted.
(This is because Create%, as Volleymetrics calculates it, contains Dig% as part of its formula, and so does Convert%, so you’re sort of double-counting. By using the derived formulas, we avoid double-counting and untangle some of the individual elements of defense/transition a bit more.)
I was surprised to see Dig% with such a high correlation! Just from a mathematical sense, multiplication tends to be driven by the smaller number1 so I thought Dig% might show a lower correlation just based off that. That wasn’t the case.
Does Dig% = Defense?
Here’s where you have to start to untangle a few things. For example, take this dig:
This is unquestionably “a dig”, but most coaches would probably assign an equal or greater contribution to the block tough than the defender. It doesn’t take anything away from the digger, but when you start looking at team correlations, quality of block is likely going to correlate with Dig%. An ideal analysis would untangle the Dig% from the quality of block, and when I was with the NT, I would do that. But since it’s not in the public data, I won’t do that here.
I think the Dig% I’m using here is a decent enough proxy that it still gives us some information, but it’s something to keep in mind.
Another consideration is that not all kills are assigned to be a Dig Error. And Volleymetrics can be a bit inconsistent in how they grade. For example, one of these was scored as a Dig Error and one was not.
In my ideal scenario, I match every Kill up with either a Dig Error or Block Error. That way, you have a more consistent analysis. I chose these two examples because they are so similar, but one of the issues with how Dig Errors are typically graded only when a player shanks a dig, meaning: there was contact with the ball and a dig wasn’t made.
But isn’t a big part of defense the ability to get to the ball in the first place? Baseball struggled with this for a long time, including awarding Gold Gloves to Derek Jeter when advanced metrics graded him as actually being one of the worst defensive shortstops in the league. He was sure-handed (ie, he rarely shanked the ball) but with very limited range. And our coaching intuition tells us there must be volleyball defenders like that as well. (And the inverse, for that matter.)
For these reasons, I prefer a Dig% metric that encompasses every ball hit in the court2 regardless of whether it was touched or not.
But alas, we have the data that we have, and this data is still predictive, as the correlation shows. In subsequent articles, we’ll unpack the data around the other two aspects of dig/transition.
Ex: 8 x 2 = 16. Add 1 to 8 and you get 9 x 2 = 18. But add 1 to 2 and you get 8 x 3 = 24. Usually dragging up the smaller number makes the product bigger. Correlations don’t exactly work that way, but my mathematical intuition told me that optimizing the CreatesConverted (the smallest raw number) would have driven up your ability to win transition rallies. Interesting that this wasn’t the case.
If a ball was “bounced” and a defender did not have any reasonable chance to dig the ball, I assign the error to a blocker instead.