Serving Profiles - Part 4

Knockout Queens

Related:
Digging Profiles
Blocking Profiles

Articles in this series:
Part 1 - Introducing The Data
Part 2 - Serving For Aces
Part 3 - Serving In
Part 4 - Knockout Queens

Knockout Queens

In Part 1, I introduced 3 broad profiles of serving teams:

1. Teams who serve a lot of aces, at the cost of missing serves

2. Teams who serve in a lot, at the cost of less service pressure

3. Teams with high service pressure but who don’t necessarily serve a ton of aces

Today we’re going to look at is the Knockout Teams. These are teams are don’t serve a relatively high amount of aces or serve a relatively low amount of errors, but instead manage to be very effective with the balls that are kept in play. We’ll call those “Knockouts” as in “Knocked The Other Team Out-of-System”.

First, let’s look at our leaderboard:

You’ll notice that this is not necessarily organized by Opponent Good Pass (OGP) or KO%. It’s organized by a (hidden over on the side of the chart) factor called “KO Index”, which is basically a measure of how relatively-reliant upon KOs the team was. So Nebraska tops the chart not because they had the lowest OGP or the highest KO, but because their serving was fairly effective overall despite having a very low Ace % and a fairly high Error %.

Winning And Losing

There is correlation between Knockout % and winning. In this sample, there was a 0.12 correlation between KO % and Win %. Meaning, teams with a high KO% had a higher win %… but only slightly. Look at the scatter-plot of Win% as a function of KO%:

It’s not the most convincing chart in the world if you’re trying to show how critical Knockout % is. In contrast, here’s the scatter-plot of Win% as a function of InSys FBK:

The trend there is much more clear.

Let’s look at the other correlations I pulled that specifically relate to serving:

0.15 - Ace %
0.21 - Error %1
0.05 - Opponent Good Pass2
0.33 - Ace:Error Ratio
0.12 - Knockout %

I mentioned this in a previous post, but I’ve been surprised that the correlation for Knockout % was so low. For example, not only is Knockout % correlation lower than Ace:Error ratio, it’s actually lower than both Ace % or Error % by themselves. Let me quote the last article in this series:

If we were truly trying to design a test to correlate OGP% and Win%, we would somehow clone a team and then run multiple seasons with different OGP%. But we can’t do that. And clearly, within a given rally, there’s a significant correlation between pass quality and Sideout %. And Sideout % matters a lot.

So on an individual play basis, knocking the opponent clearly matters. But yet, at least in this recent NCAA season, teams that are good at knocking opponents out-of-system over the course of the season don’t seem to have meaningfully higher Win % than teams that aren’t quite as good. Why? We’ll get to that more in the next article.

A Knockout Serve is, in a vacuum, a good thing. To roughly approximate, at the NCAA level, the Sideout % by pass quality is roughly:

67%: Good Pass (R# or R+)
60%: Medium Pass (R!)
53%: Bad Pass (R-)
33%: Shanked/Overpass (R/)
0%: Ace (R=)

Overall sideout in the NCAA is a bit below 60%. In this sample of top-100 RPI, you could use 60% as a nice round number ballpark.

But here’s where how you calculate Knockout % matters. If you count KO% as essentially the inverse of Opponent Good Pass while accounting for service errors, then you get all the medium (R!) passes counted as Knockout serves. And here’s the deal… the sideout % on a medium pass is essentially right at the average sideout %. This means that a serve that produces a medium pass is kind of a null result… it’s not a bad result for the server but it isn’t a positive result.

Micro v Macro

We do know that Good Pass % correlates to winning. In the Offensive Profiles sample, the correlation between GP% and Win% was 0.47, which is fairly strong. It makes sense then, that the correlation between OGP% and Win% would be similar, right?

Somehow… not.

Again, we know that, on an individual play, the pass quality correlates to Sideout %. But these correlations aren’t operating on a micro level, we’re looking at macro-level correlations to season winning %. Here’s some potential explanations for that discrepancy:

1. There’s enough good servers to go around. NCAA teams have lots of scholarships, plenty of subs, and there’s hundreds of thousands of high school girls playing volleyball. It may be that teams can just get enough good servers on the roster to the point where most teams can field a team capable of serving relatively tough.

2. There’s little correlation between attacking ability and serving ability. In high-level men’s volleyball, you tend to see more of the top attackers also being top spin servers. Float serving doesn’t seem to correlate to attacking ability, and most NCAA women are float servers.

3. NCAA teams probably select their liberos for passing and defense, not serving. This means that the majority of liberos and defensive specialists are getting on the court for their passing ability. Which means that you might see, on the margins, more teams with lower Win % but a strong-serving libero than teams with a lower Win % and a strong-passing libero. (Again, not across the board, but at the margins.)

4. Beyond liberos, passing is probably easier to evaluate and recruit for than serving. It’s probably more predictable from a coaching standpoint to guess who will be a strong passer than who will be a strong server, especially in a club setting with limited court space, etc. A top-10 program (with a high Win %) is more likely to offer a scholarship to a strong outside attacker who is a good passer and mediocre server than to a strong outside attacker who is a mediocre passer and a good server.

When you put these factors together, you can see how, at the macro level, GP% can correlate to Win % but OGP% doesn’t.

Takeaways For Coaches

The important thing to understand about this information is that, from a training perspective, serving is just as important as passing¹ and serving for knockouts is just as important as passing in-system. The data is clear about that. Don’t be confused by the macro-level correlations between team Win %. On a micro level, if you produce 5 extra knockout serves in the next match, it will raise your odds of winning the same as getting 5 extra in-system passes.

(This assumes that X amount of training is going to improve both your serving and passing by the same amount… that question is a bit outside the scope of this article.)

Now, on a macro-level, the other takeaway is that the winningest teams in NCAA women’s volleyball currently prioritize passing over serving. If you’re also an NCAA coach, you might use that information in one of 3 ways:

1. I want to be like the winningest programs. I will also prioritize passing over serving in my recruiting and training.

2. If the most successful, and thus most prestigious, programs are prioritizing passing, there’s going to be relatively more competition for the good-passer-medium-server and relatively less competition for the medium-passer-good-server from a recruiting standpoint. Thus, I’ll prioritize serving over passing, at least in my recruiting.

3. Come on Trinsey, I don’t even have 12 scholarships, much less any NIL money. I’ll take whoever I can get and train them up.

Only you can decide which is the right strategy!

1. Theoretically, it actually has to be at least as important as passing. All teams will attempt more serves than passes so serving has to be at least as important just by volume difference.