Evaluating Blocking: Part 2

In Part 1 of our series on Evaluating Blocking, we noted a few of the challenges that make evaluating blocking difficult:

1. The blocker doesn’t have full control. On most sets, the hitter can beat the blocker with the right shot. It’s possible to make a good block and still have the hitter kill the ball.

2. Many attacks never touch the block. At some levels, over half of attacks will touch the block, but at lower levels (such as high school), more than 3/4 of the attacks will go clean past the block. It’s difficult to evaluate the effectiveness of a block when the blocker doesn’t touch the ball.

3. There are lots of non-terminal blocks. At the NCAA women’s collegiate level, only about half of block touches are stuffs or tools/errors. It’s not always clear whether a “block touch” (that doesn’t result in a point for one team or the other) is a positive or negative play.

4. It can be difficult to separate the performance of the player from the system they play in.

5. The standard box score in the USA, the NCAA box score, is mediocre at best at giving you blocking information.

We talked about the last point, and we saw that adding block errors/tools to the equation gives you better information with which to evaluate your blockers.

Teach players that their number-one job as a blocker is to stuff the ball. But the next most important thing is to not give up an easy point by getting tooled! Understanding both sides of this equation makes blockers better.

But what about point #2? Many attacks never touch the block. In this match, Texas defended 112 BYU attacks. BYU made unforced errors on 9 of those balls, and the Texas block touched 50 of them in some way, which means 53 of 112 BYU attempts, or 47%, were untouched by the Texas block. On the other side of the net, BYU defended 90 Texas attacks. 6 were unforced errors and 37 were touched by the BYU block, which means 52% of Texas attempts were untouched by the BYU block.

What do we do about all of these non-touches?

First, let’s look at how I scored them on the GMS Stats app:

BYU Opponent Sideout Screen, GMS Stats App

Texas Opponent Sideout Screen, GMS Stats App

BYU is on top, and Texas is on the bottom. Both of these screens are displaying our favorite blocking stat: “Stuff to Tool Ratio” as the Blocking Efficiency stat. If you refer to how Volleymetrics graded their blocking in Part 1, you’ll notice two things:

1. There is a bigger gap between Texas and BYU in my app than in the Volleymetrics grading.

2. Both teams have a lower ratio in GMS Stats than in Volleymetrics.

Why is that?

The reason for this is that I assigned a Block Error or Dig Error on every opponent kill. I always grade block and defense in this way, and recommend everybody taking statistics on their team to do the same. So instead of 16 Blocking Errors, I gave Texas 18. Instead of 9 Blocking Errors, I gave BYU 15. In some cases, these Blocking Errors are plays where the block did not touch the ball at all, but I assigned them as the primary responsibility for the kill. For example:

In most grading systems, neither of these would register as actions by the blocker. But clearly, the blockers bear responsibility to stop these attacks. While both of these hits could have been dug, they would require a heroic effort by the defender. My judge is that if you had to decide, “which of these players is primarily responsible for that point being allowed,” you would choose a blocker in both instances. We’ll call this grading system, “primary responsibility grading.”

One of the great things about advanced programs like Volleymetrics or DataVolley is that you can drill down deep into things like this. When I was with the USA National Team, we had a system for evaluating Blocking Efficiency that we called, “Expected Attacking %.” In this stat, we had 6 different grades of block touch, including a tag for, “No Touch,” where the hitter hit a ball hard past a blocker and the blocker failed to get a touch on the ball. In this system, the “expected efficiency” of each touch (except stuffing the ball or getting tooled) is always some fraction of a kill; remember that even good touches that slow the attack down are mishandled by defenders sometimes and even poor deflected touches are dug sometimes as well.

Some other coaches might use a 3 or 4-point scale rating for blocking where they assign each touch a value and give the blocker a grade like 1.6 or 2.2 (there’s some other issues with this particular system that we’ll discuss in part 3). We’ll call systems like this or “Expected Attacking %,” “fractional grading.”

This allowed us to get precise values for how a blocker affected an opponent’s ability to score against her. Unfortunately, most high school and club coaches don’t have dedicated statisticians or the budget for advanced statistics software! That’s why I don’t use a fractional system for blocking in the GMS Stats app. A primary responsibility system is easier to keep track of, because blocking touches are only assigned when the rally ends- either with a kill or stuff block.

Sometimes there are judgment calls:

While this is not a straight-down kill, it’s also a difficult play for the defender. Many coaches would be tempted to just call it a “good hit” by the attacker and leave it at that, but I think you’re missing some information. I decided to assign responsibility to the left-back defender. Let me know what you think in the comments section.

In summary, there’s a “Goldilocks Zone,” for most volleyball statistics where getting the right data can get you 90% of the information you need, but getting that last 10% of information requires massive resources. For NCAA and professional teams, that can be a worthwhile expenditure of resources, but most high school and club coaches just need to get the most bang for the buck!

In my opinion, the Goldilocks Zone for blocking is:

Not Enough Info: Only keeping track of stuff blocks.

Too Much Time/Energy: Using a fractional grading system to capture the value of every blocking action.

Just Right: Track stuff blocks and blocking errors, and use a primary responsibility grading system to capture a few additional non-touches that the block was responsible for.

In Part 3 of our series on Evaluating Blocking, we’ll dive deeper into non-terminal block touches and try to find out how much they really matter.

Evaluating Blocking: Part 1

Two of the most difficult skills in volleyball to evaluate statistically are blocking and digging. We’ll discuss defensive evaluation in another article; today we’ll focus on blocking.

Some of the challenges that make it difficult to stat blocking are the same that make it difficult to coach blocking:

1. The blocker doesn’t have full control. On most sets, the hitter can beat the blocker with the right shot. It’s possible to make a good block and still have the hitter kill the ball.

2. Many attacks never touch the block. At some levels, over half of attacks will touch the block, but at lower levels (such as high school), more than 3/4 of the attacks will go clean past the block. It’s difficult to evaluate the effectiveness of a block when the blocker doesn’t touch the ball.

3. There are lots of non-terminal blocks. At the NCAA women’s collegiate level, only about half of block touches are stuffs or tools/errors. It’s not always clear whether a “block touch” (that doesn’t result in a point for one team or the other) is a positive or negative play.

4. It can be difficult to separate the performance of the player from the system they play in.

5. The standard box score in the USA, the NCAA box score, is mediocre at best at giving you blocking information.

There’s a lot to unpack here, and we’ll need more than one article to really discuss blocking. Let’s start with the last point first.

NCAA tournament box score: BYU vs Texas

We’ll take a look at this NCAA tournament match between BYU and Texas. The blocking information here is contained in the columns BS (“block solo”) and BA (“block assist”). The first thing you realize is that these distinctions are almost meaningless in modern NCAA volleyball, where almost every team is in some sort of help-block system. Between the two teams, there were 24 stuff blocks, and only one was scored as a solo.

By the NCAA blocking criteria:

A block assist (BA) is awarded when two or three players block the ball into the opponent’s court leading directly to a point. Each player blocking receives a block assist, even if only one player actually makes contact with the ball.

That means that on these plays, both blockers received equal credit:

This isn’t to diminish the efforts of the Texas or BYU middle blockers. Their job is to try to get involved in the play as best they can. Middle blockers can and do block balls when they are this late by hitters who hit the ball low into the sharp angle. But on these particular plays, it’s difficult to say that both players should be credited with an equal impact on the play.

It also happens (although not in this match) that middle blockers will stuff a quick attack and one of the wing blockers will take part in the block enough to get credit for a block assist, although they were barely off the ground at hitter contact.

International blocking statistics do a bit better, as they only credit one player with the block.

Serbia box score, 2018 FIVB World Championships Gold Medal Match vs Italy

The FIVB statistics also add an “attempts” category, which is every touch the blocker made in that match.

That gives us a little bit more information (2 out of 4 of Busa’s touches were stuffs, compared to 2 out of 7 for Ognjenovic, that seems to give her an advantage), but both box scores neglect very important information:

Both NCAA and FIVB box scores only provide information about the most positive blocking touch (stuff block) and neglect to include the most negative blocking touch- a tool or blocking error. This would be like recording only kills and aces and not hitting or serving errors. (Whoops, the FIVB box score does that as well? Well… at least it’s consistent!)

Just as hitting efficiency has a higher correlation to winning than just kill %, blocking efficiency has a higher correlation to winning than just recording stuff blocks.

The stat I really like for Blocking Efficiency is “Stuff to Tool Ratio.”

BYU vs Texas Blocking Statistics, courtesy of Volleymetrics.

Although the traditional box scores might be lacking, using more advanced statistical programs like DataVolley or Volleymetrics can give us more insight into what happened on the blocking end.

Quite a few individuals see their blocking statistics change. For Texas, Butler was credited with 4.5 blocks (9 BAs, each worth 0.5 blocks) by the NCAA but 7 by Volleymetrics. On the flip side, Johnson dropped from 3.5 (7 BAs) to 3 and Gabriel and Eggleston dropped from 2 BAs each to 0. BYU also sees some similar changes based on who actually made the block on each play.

Where it really gets interesting is when blocking errors are factored in.

At first glance, one of the stories of this match was the blocking advantage Texas had. They outblocked BYU 15 to 8! (NCAA box score actually had Texas with 16 blocks, but it’s likely they counted a ball BYU attacked into the net as a block- that happens sometimes.)

However, look at the blocking error column. This includes the blocker being in the net, as well as the hitter tooling the block. Texas had 16 and BYU had 9. Now things don’t look so lopsided. In terms of a raw +/-, both teams gave up 1 more point via blocking error than they scored via stuff block.

We could also measure this in a ratio or efficiency rating, by dividing stuffs by tools. In this case, Texas was 15/16 = 0.94 and BYU was 8/9 = 0.89. I call this ratio blocking efficiency, stuff-to-tool ratio, or just, “Stuff to Tool.” This is a stat that few people use that automatically upgrades your understanding of blocking effectiveness, both on the individual and team level.

It’s reasonable to assume that stuff blocking correlates well with blocking efficiency, just as kill % correlates well with hitting efficiency. But just as some hitters are high kill and high error, some blockers are both high stuff and high tool. This statistic also helps us coach. Teach players that their number-one job as a blocker is to stuff the ball. But the next most important thing is to not give up an easy point by getting tooled! Understanding both sides of this equation makes blockers better.

This brings us back to another reason NOT to use block solos and block assists. Using the NCAA logic for assigning block assists to any blocker involved in the block, whether they touched the ball or not, we’d have to assign block errors to any block involved in the block, whether they touched the ball or not. This strikes most people as absurd, so why are we assigning block assists to both players when only one made the block?

To summarize:

1. Block assists are almost meaningless. The FIVB method of assigning the block only to the player who blocked the ball gives more accurate information.

2. Both NCAA and FIVB statistics neglect at least half the story by only listing stuffs and not blocking errors.

3. Blocking efficiency (stuff-to-tool ratio) gives us a much clearer picture of a team or individual’s blocking effectiveness than just stuff blocks.

In Part 2, we’ll look at how the GMS Stats App analyzes blocking, why it’s different than Volleymetrics, and what to do about all those pesky non-terminal block touches!

What Is Point Differential And Why Does It Matter? - Part 2

In our earlier post on Point Differential, we talked about different ways of expressing the score:

• Total number of points (ex: 104 to 103)

• Point Differential (ex: 50.2% to 49.8%)

• Sideout Differential (ex: 56% to 52%)

While each of these has value, Sideout Differential probably tells us the most information, because, in addition to telling us the margin of victory (or defeat), it tells us a little bit about how that victory (or defeat) occurred. A 70% to 68% Sideout Differential tells us something different than a 50% to 48% Sideout Differential. But what else can Sideout (or Point) Differential tell us?

In terms of Sideout/Point Differential, there’s 6 scenarios that can happen:

1. We have a large positive differential and win the match.

2. We have a large negative differential and lose the match.

3. We have a small positive differential and win the match.

4. We have a small positive differential and lose the match.

5. We have a small negative differential and win the match.

6. We have a small negative differential and lose the match.

(There’s also the scenarios where the differential is zero and we either win or lose, but those are similar enough to scenarios 3 through 6 to lump them in together.)

Scenarios 1 and 2 aren’t too interesting. Well… maybe interesting in some ways, but in terms of statistical analysis, not so much. You were either much better or much worse than your opponent that day and there was a gap in fundamental skills. Small tactical adjustments aren’t going to erase a 7-point lead.

Scenarios 3 through 6 on the other hand, are a lot more interesting to us from a statistical perspective, and these scenarios are ones where little things can tip the balance of the match one way or the other.

These four scenarios recently played out in the NCAA Women’s Volleyball Final 4. Let’s take a look, using our handy GMS Stats app to produce the analysis.

In our post breaking down the Final Four match between Illinois and Nebraska, we saw scenarios 4 and 5 play out, depending on which side you were rooting for.

Illinois Point Differential Screen, GMS Stats app

Illinois outscored Nebraska, but lost the match. Or, if you’re a Nebraska fan, Nebraska got outscored, but won the match. This is uncommon, but it does happen. In a 5-set match, the team that scores more points only wins about 78% of the time.(*) On the flip side, we have the Stanford - Nebraska Championship Match:

Stanford Point Differential Screen, GMS Stats app

On average, the margin of victory in a 5-set match is about 5 points, so these two matches were close even by 5-set standards.

In the post on that match, we looked at some of the rotational matchups and how they played out. A lot of times, these matchups can be the difference in a close match. Rotational order, matchups, and clutch play are the three things I like to look at in close margins of victory.

The danger with looking at one match is that there’s a lot of randomness involved. Is that rotation in which you gave up a late-game run really a bad rotation, or did you just catch a hot server at the wrong time? This is why we like to take a multi-match view of things. In the GMS Stats app, we built that into the Wizard feature:

Point Differential Wizard Screen, GMS Stats app

This is the Wizard screen I pulled from a 5-match sample. I was coaching a professional team and wanted to see how our performance translated into wins and losses. For a single match, this translation is simple: if you outscore your opponent, you expect to win the match. If you get outscored, you expect to lose.

But things get trickier over multiple matches. If your Differential is 2% over a half-season, what do you expect your record to be? How about 5% or 10%? Fortunately, the Wizard feature does these calculations for us!

In this 5-match sample, we had a small but meaningful 2% edge over our opponents and that resulted in a 3-2 match record. Intuitively, we probably sense that this is about right. Maybe, we figure, we might be able to go 4-1, but 5-0 feels like a stretch. On the flip side, since we’ve outscored our opponents, we probably feel like 2-3 would be a letdown and 1-4 would indicate that something is really wrong.

Over time, this Point Differential-to-record translation can give us coaching insight. We can now imagine the previously mentioned scenarios reducing down to:

1. We are outscoring our record.

2. We are underscoring our record.

If we were 5-0 with a differential of only 2%, this would tell us some things as coaches. The first thing it would tell us: don’t get too cocky just yet, we’re probably getting lucky! It would also indicate that, while the results are good now, we should still be looking to make some changes if we want to keep the winning streak alive. There’s always the temptation to not make changes when the team is winning, but in this scenario, changes might still be warranted.

If we were 2-3 or 1-4 with a positive point differential, it tells us something different. First of all, it tells us that we’re not doing quite as bad as we think! So, while some changes are probably necessary, they might not need to be as drastic as the record might indicate. It also tells us that we’re probably losing some close games, and we have an opportunity to close them out better.

In this case, we look at our three Close Game Factors:

1. Rotation Order

2. Matchups and Scouting

3. Clutch Play

More on these factors in our next post!

(*) 5-set win % and margin of victory data provided by Volleytalk legend The Bofa on the Sofa.

What Is Point Differential And Why Does It Matter? - Part 1

In a previous blog post we showed the Point Differential screen of the GMS Stats App.

Nebraska Point Differential Screen; 2018 National Semifinal vs Illinois

But what is Point Differential? Why does it matter? How does it affect me as a coach?

Point Differential is simply the difference between how many points I score, and how many points my opponent scores. It’s a fancy way of saying, “the score.” Most of us intuitively grasp the following two equations:

Scoring More Points Than The Other Team = “Good”

Scoring Fewer Points Than The Other Team = “Bad”

“Great, Joe, score more points than the other team. I can see why the National Team hired you.”

Calm down, unnecessarily sarcastic imaginary reader.

This screen from the app actually shows Sideout Differential, which gives us a little bit more information. This screen shows Nebraska (“your”) Sideout and Illinois (“Opponent”) Sideout. What’s the difference between Sideout Differential and Point Differential? It’s really just a matter of perspective.

Quick Volleyball Stats 101 Lesson:

A “Sideout” is any time the other team starts the rally with a serve and we win the point, whether they miss the serve, or we win on the first chance to attack, or it’s a long rally that we win in the end. If the rally started with the opponent serving, and we win the point, that’s a Sideout. (So an Opponent Sideout is any time we start the rally with the serve and the opponent wins the point.)

“Sideout %” is the number of chances we had to sideout (also: the number of times the opponent serves) divided by the number of times we actually sideout. So if the other team serves 100 times, and we sideout 54 times, our Sideout % is 54%.

So Sideout Differential is just a way of looking at Point Differential from a different perspective. You cannot have a better Sideout % than your opponent, but a worse Point Differential. And you cannot have a better Point Differential than your opponent, but a worse Sideout %. So why use Sideout Differential instead of Point Differential?

The best reason is that Sideout % is a reliable indicator of level of play. In U12 volleyball, the serving team has an overwhelming advantage. Sideout % is well under 50%. At the professional level, the receiving team now has the advantage. Both teams will Sideout well above 50% in most matches. So imagine two different matches:

Match 1: Sideout 41%, Opponent Sideout 40%

Match 2: Sideout 61%, Opponent Sideout 60%

In both these matches, the Differential was 1%, but the style was different. Match 1 was a defensive battle where both teams were going on runs. Match 2 was more of a sideout battle, with the offenses at an advantage over the defenses. Knowing the differential as well as the overall sideout rate between the two teams allows you to get an idea of the overall level of play, as well as how your team fared.

In Part 2, we look at more uses of Point or Sideout Differential and go from looking at one match to multiple matches.