### Attacking is the most important skill.

### But we only evaluate 50% of it.

##### *** if you already understand the basics, *skip to the bottom for the advanced stuff* ***

Of course these exact terminal/non-terminal splits vary by gender, level, etc.

But the overarching theme is a constant.

We are **ignoring** about half of the attacks.

So let's look at those other 50%, they can't actually all be worth nothing?

You're the coach.

Bobby takes 10 swings in a match.

**4 kills, 1 error, and tips the other 5, playing conservatively**

Billy also gets 10 attempts.

**4 kills, 1 error, crushes 5 and forces OOS attacks**

#### Both guys are hitting 0.300 right?

False. Black bear.

We know instinctively that forcing OOS attacks or creating good opportunities for your own team are better than tipping easy balls for the defense.

But how much better? 2x? 3x? 10x?

#### Welcome back... to Expected Value

To be clear, these are the same chart, expressed two ways.

Left - in terms of **Attack Efficiency**

Right - in terms of **Expected Possession Value**

- Translated:
**how likely are you to win the point**, given this attack

**It doesn't look like much**

But making good decisions on non-scoring attacks can have a **0.100 point difference, per attack!**

So back to Bobby and Billy.

Reminder

- Bobby = kills, 1 error, and tips the other 5, playing conservatively
- Billy = 4 kills, 1 error, crushes 5 and forces OOS attacks

Bobby: (4 * 1.000) + (1 * -1.000) + (5 * -0.057) = **0.272**

Billy: (4 * 1.000) + (1 * -1.000) + (5 * 0.047) = **0.324**

A **50-point split** like that could certainly be the difference between on the court vs. on the bench.

### But this is the easy version.

If you're using DataVolley or VolleyMetrics, you'll already be familiar with these attack qualities and recognize that A+ and A- correspond to In-Play (good) and In-Play (poor).

#### But what if we used all the data we had, not just A+ and A- ???

What if we knew:

exactly where the set came from,

if it was in-system, OOS, or an overpass,

how many blockers you faced,

and the quality of the team you're playing?

#### Well shoot, we'd be probably be able to design even better Expected Value metrics...

### So now that we understand the basics,

### here's how Expected Value really works:

**Expected Points Added = Result - Expectation**

Every "state" of a rally has a likelihood that either team will score.

So what we really mean is: take the Expected Value after - Expected Value before

Imagine a rally.

Starts with a **tough serve** that forces an out of system set from deep in zone 6.

There are **3 blockers** waiting for the outside hitter on the incoming high-ball.

The outside takes this **OOS set** and chips it high into the block, **recycling it.**

The libero covers his outside, putting up a **perfect dig** to zone 3 for the setter.

The middle gets on a route, but the setter plays it overhead to the opposite for a **1-on-1 attack**.

The opposite easily **kills the ball** in this 1-on-1 situation.

### Who's the hero in this rally?

Let's take a look.

**40%** = that's how likely you are to win, given an OOS set from deep in zone 6

- we use: XY coordinates of
**set's origin**+**out of system**+**3 blockers**+**quality of opponent**

**70% **(+30) = how likely you are to win, given you recycle off the block into a perfect dig

- we use: where the setter will
*now*set from, etc. etc.

**75% **(+5) = how likely you are to win, given the 1-on-1 situation for the opposite

- we use: set origin + in-system + 1 blocker + quality of opponent

**100% **(+25) = how likely you are to win, given it's a kill....

- we use: nothing, it's just a kill. always 100%.

Expected Points Added is the value you create as you move from one "state" to another.

The **outside** hitter here adds 30% of a point by **taking a terrible situation and creating a good one**

The **opposite** hitter adds only 25% of a point for killing the 1-on-1 set, because he is **already expected to perform well** in that situation

We do this for every. single. attack.

(to be fair, the computer does it...)

So when we talk about Expected Points Added per attack, we simply average across every attack for that player.

Boom.

**Expected Points Added (EPA).**

Data-savvy folks, want to pull metrics like this? Check out volleyR

https://www.volleydork.com/post/volleyr-expected-value-for-volleyball-automatically

^ you only need better technical skills than your grandparents, I promise.

Quick Reminder: how to convert between Expected Value and Efficiency

https://www.volleydork.com/post/efficiency-vs-expected-value