Best Liberos in NCAA Women's Volleyball (2022)

(my apologies to Elena Scott and the others I left off at first - the original intent of the post was to be Top Passers, so I set the threshold at 300 receptions. this was latter amended to Top Liberos/DS, but I forgot to change the threshold to be >300 rec + dig as the post describes - the correct list is attached)

Defining value for a libero is honestly tricky - and to be clear, I don't think I have it 100% nailed down.

But I'll walk through what I have so far, and explain how I got there.

Looking at two things: Reception & Defense

- How well do they receive, relative to expectation?

- How well do they defend, relative to expectation?


We need to solve 2 parts of the equation.

How hard is what they're doing? How are they performing, relative to that level?

But since serving is such an individual thing

...and can vary wildly within a single team, we need to account for this

So we look at the individual server to determine our expectations.

Specifically, what is the Expected Sideout % that each server forces, on average?

We don't use the "actual" sideout here, because if you play for Texas, even a medium pass has a better chance of scoring than if you played for another team.

So we use, Expected...Expected Sideout. This way, we can try to quantify the "strength of the server" and use that to set the expectation for our passer.

Actual Expected Sideout refers to the value of the reception to an average team in our dataset.

Again, we use Expected Sideout so that medium passers with good offenses aren't infalted - and great passers with bad attackers aren't penalized.

Expected Sideout over Expectation = Expected Expected Sideout - Actual Expected Sideout

This means, relative to the expectation of facing this specific server, how much can we expect your pass to help the team sideout.

Lexi Rodriguez here has an Expected SO over Expectation of 4%

Meaning that every time she passes, if you normally expect 66% expected sideout, she gets you up to 70% expected (or +80pts in terms of attack efficiency)

(^ for example: this would be taking you from 0.250 up to 0.330 attack eff) - a big deal!!!

You can see more about converting between Expected Value and Efficiency here:


For the defensive side of the ball, we don't specifically account for individual attackers.

This is because players can attack from a wide variety of situations and slicing again by specific attackers shrinks the sample size quickly.

But by limiting our data to the top conferences, we introduce fewer ~weaker~ attackers into the expectations.

To start, we take the Expected Value of the offensive situation an attacker has.

1 on 1, perfect pass = high Expected Attacker Eff (will be tough for L to defend)

Bad pass, triple block = low Expected Attacker Eff (should be easy to defend)

Overpass attack = super high Expected Attacker Eff (real real hard to dig)

We take all varieties of offensive situations into account...

and then look at how much value the Libero's touch has

The worse the dig, the more value for the Attacker.

If it's a kill that is touched by the Libero, then the Actual Attacker Eff = 1.000

This can harm some liberos who are flying around, trying to touch / poach everything.

If the libero successfully digs it, the quality of the dig determines the Actual Attacker Eff.

Meaning a poor dig may still be beneficial to the original attacker's team.

A perfect dig is naturally detrimental to the original attacker's team.

Liberos who do not touch the ball are not credited/debited in this equation.
It's just a limitation of the data we have.
But maybe computer vision can solve this?

Anyway, then as usual, we subtract one from the other.

Stopping Power = Actual Attacker Eff - Expected Attacker Eff

I'm using the term Stopping Power here as the impact the Libero has on the Attacker's Eff.

The better the libero, the more she will decrease the Attacker's Efficiency.

In the chart, we combine all of the libero's reception and digs, looking at average value added per touch.

This is how we ranked everyone.

I'd like to improve on this metric moving forward, but this is where my mind went after brainstorming for a while. I think it's an improvement for sure. But certainly still has its assumptions and therefore its shortcomings.