A Better Way to Select All-Americans (Men's 2023)

The NCAA tournament is upon us!!

So too, are the AVCA All-American awards, and I figured this would be a good time to put in my two cents.

What follows is my earnest, best effort to capture how we have historically viewed the game, while pairing those metrics with their modern day equivalents as derived from Expected Value.

So let's look at each skill, from both an old and new perspective.


Old - Opponent Passer Rating

How well, on a scale of 0-3, does the other team pass when you serve?

We think of this as a proxy for good serving,

but really it completely misrepresents the value of getting aces

(more here)

New - Opponent Eff vs. Expected

Do you lower your opponent's ability to sideout?

We like this because it's directly related scoring points

More specifically, stopping your opponents from scoring points

This is improved even further by taking into consideration the reception quality of the team you're playing.

So if you are serving at a super weak passing team and you get them OOS a lot, well, that's much less impressive than if you are scoring aces against Hawai'i or Penn State for example.

This way, when can isolate the value added, above expectated, for every single serve.


Old - Passer Rating

It's great for estimating how often you pass well.

But it basically ignores getting aced.

(still more here)

Which if you're likely to sideout at 70%, getting aced means you lost 0.7 points on a single touch

Hugely undervalued problem.

New - Efficiency vs. Expected

Two main improvements here: using efficiency and adjusting for the specific server

By using expected value, we get the full range of potential outcomes and value them at exactly what they are worth.

A perfect pass isn't a 3, it's worth X hitting efficiency by your subsequent attack.

If you get aced, well shoot, that's the same as hitting -1.000, so we value it that way.

This provides a much more accurate account of the value being created by the passer

The second is adjusting for each server.

Some servers are legitmately tougher than others, we have a full season's worth of data to prove this.

Therefore when passing vs. UCI's Henno (easily the top server this year), we shouldn't expect nearly as valuable of a pass as compared to when someone's grandma gets out there with a standing float.

We might expect 55% expected sideout vs. Henno, but closer to 70% vs. my grandma's serve.

This way, we give proper credit to passer handling tougher servers, while avoiding inflating those who only pass the easy stuff.


Old - Hitting Efficiency

It's actually not horrible since it relates to points. But it lacks context.

Maybe 0.250 is great for an Outside?

But what if every single swing was from a perfect pass, 1-on-1 situation?

Still great?

What if I hit 0.250 but every set came from beyond the endline

and I'm hitting against triple blocks?

Now it's great.

What if I hit 0.250 vs. Long Beach's block and defense? 

Pretty good.

What if I hit 0.250 in my grandma's church league?


Context matters.

New - Hitting Eff vs. Expected

Exactly what it sounds like, we include a bunch of factors to create a better expectation.

Strength of opponent...

Where the set comes from + whether you MB was on a route...

Number of blockers you're facing...

If it's an overpass...

We combine all these factors and create an expectation for every single attack.

But wait! There's more!

You know how only 50% of attacks are terminal?


Well we also account for all the non-terminal attacks!

They don't all have a value of zero

If you don't score, but create an overpass swing for your team, that's valuable!

If you don't score, but force an OOS swing from your opponent, that's valuable too!

If you don't score, but recycle the ball and get an in-system attack for your team, that's awesome!!

If you tip casually to their libero... that is not awesome - and you helped the opponent, so we ding you for that.

By using Eff vs. Expected and better understanding the non-terminal swings,

we can do a much better job evaluating attackers and their decision making.


Old - Blocks per Set

Pretty self-explanatory.

I don't think most folks would say this is a great metric.

Mainly because unlike the others, it lacks a "per opportunity" aspect

We only know the numerator here, but we don't know if you had 5 chances for blocks or 500

Also, not all great blocks are stuffs

If their middle is about to bounce a quick and you get a slow-down to basically turns it into a freeball for your team, that's got a ton of value.

With blocks per set, we lose both context and "per opportunity" components of the equation.

New - Opponent Eff vs. Expected

This one is trickier than the rest.

First, the hard stuff.

If a blocker touches the ball, he is by default the "responsible blocker"

But if the attack is untouched by the blocker, we need to know whose "zone" it landed in, so we can determine which blocker had jurisdiction.

So if it's a Go ball to the outside hitter and he hits down the line, then it's the rightside blocker's responsibility.

If the ball is attacked crosscourt into zone 5, then it's the MB's responsibility.

For each attack code (Go, Red, Bic, Front 1, 3, etc.) we have called out which blocker is responsible for which slice of the court.

While this isn't 100% perfect, it's much better than simply ignoring this part of the puzzle.

If your setter is blocking and never touches the ball - but the attacker is bouncing balls straight down the line...

is that really the fault of the defender?

The same with the middle - if the opponent is just hitting quicks straight ahead and your middle isn't touching anything...

is that really the fault of the backcourt defense?

So now we have established the need for a responsible blocker for most attacks

(if the outside hits a ball crazy stupid sharp to zone 4C, there's no blocker responsibility on stuff like that)

And so now it's time for the easy stuff.

Using the same methodology as before, we can evaluate the expectation for the blocker.

(Strength of attacking team, number of blockers, where the set came from, etc.)

And then using the same non-terminal evaluation, we know the value of the result.

Opponent had a 1 on 1 and you got a stuff block? Huge value add.

Opponent had an OOS set into a triple block and then you netted? Tough... you get dinged here.

You make a great block move and your opponent is forced to tip and the backcourt plays it perfectly? We get some credit here too, even though you didn't touch the ball.

This is final piece.

if you block the ball, you are credited with 100% of the resulting value.

If you do not touch the ball, but it's in your "zone" - you are credited with 50% of the resulting value.

So if that setter from earlier doesn't touch the ball and their attacker bounces it down the line, you lose 50% of the value as the blocker.

Ex: Perfect pass, 1-on-1 attack = let's say defense has 30% chance to win the point vs. this situation

They bounce it down the line, around the setter's block = 0% chance to win the point

0% - 30% = -30%

We say that this is 50% the fault of the blocker, so the setter gets dinged -0.15 points for this.

The same if the attacker was forced to tip and now the blocking team has a perfect dig.

Maybe you went from 30% chance to win (pre-attack) and now up to 70% chance to win in transition.

That's +40, but we give the blocker half credit, so the blocker gets +0.20 points for this play in his zone.

Anyway, that was a lot.

But basically we add up all the value created by the blocker

And divide by the number of opportunities he had to make a play on the ball (when in his "zone")

And we compare this to the Expectation

Boom. Opponent Eff vs. Expected.


Old - Digs per Set

This one is pretty silly in terms of stats.

If your team is bad and cannot score, you will naturally have many chances to dig.

As your number of chances increases, so too will your total digs.

A player who digs 60% off of 10 chances per set = 6 digs/set

A player who digs 20% off of 36 chances per set = 6 digs/set

This metric really doesn't make much sense because it allow for no context at all.

New - Eff vs. Expected

This is basically the flip-side of attacking.

If an attacker has a 75% chance to score, then by default, the defender has a 25% chance.

When a defender scoops up a would-be bounce from Merrick McHenry (looking at you, Theoren Brouillette)

That takes the defensive team from a 25% chance up whatever the value of the dig is.

If the dig is perfect, maybe that's worth 65% to the defense,

so we credit the defender with 65 - 25 = +40 chance to win the rally

a huge boost in value!!

We take the total value created from digs, divided by the total attempts:

boom, Eff vs. Expected

*One fault with the defensive metrics is that is does not account for range.

A player may have a larger range, touch more balls, and in doing so, hurt his overall "efficiency"

So that's how we put together these lists for All-Americans this season.

Taking direction from metrics of the past, we forge forward into the world of Expected Value.