Tag Archive for volleyball statistics

Considerations in serve reception ratings

In the article Scoring Serving and Passing Effectiveness I talk about the common usage of a 0-3 type of scale for rating serve reception. In this post, fellow volleyball blogger Hai-Binh Ly discusses how he progressed defining these ratings. Basically, he’s reached the point of using very defined zones to judging a pass’s rating. These are the zones defined within the commonly used DataVolley statistical program. Ly outlines them in his post.

I have my concerns with rigid definitions. Ly mentions some of them with respect to grey areas, but I would focus more on the fact that they fail to account for setter athleticism. Simply stated, a pass that might only be a 1 for a given setter might be a 2 for a quicker one. It could even be a 3. Think about a tight pass that a short setter cannot handle, but a taller one has no problem with.

The thing we have to keep in mind is the underlying idea behind these pass ratings.

The intention was to speak to the probability of earning the sideout. This is what Dr. Jim Coleman had in mind when he developed the rating system. The premise is that a 3-pass results in a sideout some percentage of the time. A 2-pass, on average, sees a team sideout at some other frequency – most likely lower. And so on down the line. From this perspective, a team’s average pass rating indicates its approximate sideout rate.

If pass ratings are going to approximate sideout success rates, then it makes sense to use a more discretionary rating approach. By that I mean rating passes based on the circumstances of the team in question. In other words, what can your setter do with the ball? Rigid definitions for each pass rating do not make sense in that context.

If, however, we want to compare serve reception across teams, or between players, then a more fixed system is more appropriate. In that case, we need a common system of measurement. That removes setter variability from the equation.

So which is best?

As a coach, it depends on your setters. Are they of similar quality? If so, you can use the more discretionary approach. If they are noticeably different, though, you probably have to go with a more rigid system. This is especially true if your passers do not work with each setter basically the same amount of time. It’s the only fair way to compare them.

Statistical analysis in volleyball recruiting

An article about Daryl Morey, the General Manager of the Houston Rockets in the NBA got me thinking about Moneyball for Volleyball. Should I trademark that phrase?

Using statistics in player evaluation

For those who don’t know, the “Moneyball” concept is where a sports organization uses statistical metrics to evaluate potential signings. This is in contrast to the old school eyeball analysis of scouts. The term Moneyball comes from the Michael Lewis book of that title about how baseball’s Oakland A’s used statistical methods to evaluate players and built a highly competitive roster with limited resources. There is also a movie based on the the book staring Brad Pitt. I recommend the book. It provides a bit more insight.

Before going on too far, I should say the Morey article got my attention because of it’s link to behavioral economics. My PhD work was in a closely related field. The article’s focus is largely on the interview process teams use. It’s a long one, so give yourself a block of time to read it.

Anyway, back to the Moneyball idea. Statistics have long been part of volleyball. In recent years it’s gotten a lot more focus thanks to improved applications and data. Joe Trinsey, who worked with the USA women’s team, has been one of the leaders in that regard. Have a listen to the Coach Your Brains Out podcast he’s on (Part 1, Part 2) for a bit of what he’s looked at.

That stuff is all about analyzing our players and teams. And there’s also the scouting element. How are we most effective? What is the other team’s weakness? That sort of stuff.

Stats in volleyball recruiting

What we don’t see much, if anything, about is using stats in the recruiting process. I have no doubt they get used by professional coaches. When I evaluated American players to sign for Svedala I definitely looked at their college stats, though I don’t know how far others take it. One day maybe I will get Mark to talk about it on a Volleyball Coaching Wizards podcast.

But what about college recruiting?

How many college coaches evaluate recruit statistics? My guess is few, if any. I say that in part because of how much time they spend watching video and attending Juniors tournaments. That’s basically the definition of old school scouting as described in Moneyball. The question, though, is whether they could actually go with analytics. I think most will argue that they can’t.

Why? Lack of useful data.

Issues with statistical data in volleyball recruiting

Yes, it is true that lots of high school teams keep stats these days. And much of that information is public. Juniors clubs, though, don’t really publish that information. That’s assuming they even collect it in the first place. My guess is most don’t in any comprehensive fashion. Though a few probably do.

Even if a high school or Juniors team does collect and publish stats, there is the question of reliability. Who is recording the stats and do they know what they’re doing? Even at the college and professional level there are issues regarding the quality and accuracy of the stats we get. Imagine a bunch of junior varsity kids taking them!

Finally, there is the question of comparability. What can you ascertain from a given player’s high school stats? What do they really say about that player? We want to gauge how a player will do at our level. I think, however, most college coaches don’t know how high school and/or Juniors translate. Juniors stats are probably a bit better as college coaches very often understand levels of play across the clubs.It can be a lot harder with high school stats. Unless you recruit in a very small area, you struggle to know the caliber of the schools your recruits play against, and more importantly how that compares to a recruit from a different part of the country.

One exception

The exception to the above is transfer prospects. Since those are college players, it is easier to draw a comparison. True, at the junior college level you often have the same statistics issues as you have in high school in terms of quality. It is easier there, though, to know the relative level of play the stats come from. And of course a player transferring within your own level of four-year school play is even more straightforward.

I would say the junior college to four-year college transfer process is most akin to the college-to-professional evaluation process. It provides an opportunity to make better use of statistics.

Are we doing enough?

Those are, I suspect, the reasons college coaches would put forward as to why they don’t use stats in recruiting. Are they valid reasons, though? Should high school and/or Juniors stats get more use? Or should we perhaps base things most heavily on something like the VPI developed by the AVCA?

I am not suggesting we shift completely to an analytic approach. I think most, if not all of us, agree that there is a personality element which must be considered. After all, we’re talking about a sport where one individual’s success is highly dependent on the performance of their teammates. Still, it does seem like some work on what statistics are predictive of success at the next level is worth doing.

Tracking block and defense improvement

One of the things we focused on with the Midwestern State team as the Lone Star Conference season progressed was improvement in our block and defense. Our block timing was poor. That meant not only few blocks, but also few digs. Though we also needed improvement in defensive position and actual digging. We were bottom of the league standings in both categories at one point, I believe.

Per set figures

Recently, I ran some numbers to gauge our progress. I first started with blocks/set and digs/set. Those are the commonly reported figures, so it made sense.

Through the first round of conference matches (10 total), we averaged 1.17 blocks and 11.16 digs per set. Over the course of the first five matches of the second half of the season we averaged 1.57 and 15.47 respectively. That’s pretty good.

Percentages

A coaching friend suggested I look instead at block and dig percentages. Basically, that divides those figures by the total number of non-error attacks (blocked balls excluded from the error count). Since attack numbers can vary from match to match – and five set matches always mess with per set averages – the percentage approach is the better way to go.

For the first half of the season our block percentage was 4.5%. Our dig percentage was 42.1%. That adds up to a total “stop” percentage of 46.6%. For the first five matches of the second half the comparable percentages were 4.9%, 48.3%, and 53.2%. Again, gains across the board.

In each but one of the second half matches our block percentage was higher than against that same team the first time around. The same was true of the dig percentage (different match). Similarly, when looking at the total figure, only one match was worse the second time than the first.

Limitations

While these comparisons tell us the team was more effective in defense the for the first five matches of the second half of the conference season, there is a limit as to how far you can take the analysis. What happens on the other side of the net leading to an attack matters. If you do a better job putting a team in difficulty through tough serves and/or good attacks, you will likely find it easier to block or dig their attacks.

Also, ultimately what you want from your defense is it to generate point scoring. That means it’s worth extending the analysis of something like dig percentage to see how many swings you get from those digs and how efficiently they convert into points.

Is a block a hitting error?

A reader asked me the following question about hitting and blocking statistics.

Is a won block counted as a hitting error for the corresponding hitter?

In U.S. volleyball the answer to that question is usually “Yes.” Elsewhere in the world, I think the answer is “No.”

I say that based on my experience as a coach in Sweden, and also from statistics in European leagues. The common practice there is to break out actual hitting errors from blocked balls. This might just be a function of DataVolley reporting, though.

Which is the right way? That is up to the statistics user.

From the perspective of reporting, the trend is to take a positive view. By that I mean they want to report players earning points rather than players giving up points. In that mindset a block is a positive thing for the defensive player. It is a negative for the hitter.

As coaches, however, we must decide which way to count them. It is about which approach provides the best information for us in the context of our own teams. There is definitely value in splitting errors and blocked balls, which standard NCAA box score reporting does not do.

Personally, I like including blocked balls for hitting efficiency [ (kills-errors)/total attempts ]. There is value in more granular reporting, though.

Kill percentage off perfect pass

The following question came in from a reader:

What percent of kills should we expect on a perfect pass? Serve receive or free balls?

The answer to this is reliant very much on level of play. High school girls probably do not score at the same rate as college men, for example. Unfortunately, the mailer didn’t tell me what level they are at.

I honestly don’t have a specific answer in any case. I reached out to Mark Lebedew from At Home on the Court to see what he had to say, and he told me in the men’s PlusLiga in Poland (the top professional division) it’s a 62% kills rate, with a 47% hitting efficiency. This struck me as low, but that just goes to show that personal impressions aren’t always (or even often?) right. 🙂

Mark went on to say the PlusLiga sideout rate off perfect passes is 72%.

I’m curious to hear what folks with good figures say about kill % and sideout rates at their level. If you have any data, please share via a comment below.

Possible paths for volleyball research

The subject of the influence of a coach’s decisions on match outcomes is now a talking point in coaching circles. That wasn’t always the case in the past. For many years the assumption was that coaching interventions (timeouts, subs, etc.) without doubt influence outcomes. This is the coaching mythology. That mythology is being challenged by the research.

Examples of this come from Mark Lebedew. He did a basic study based on the question of whether timeouts in any way influence the likelihood of the server missing their serve. In other words, are servers more likely to miss after a timeout. This is believed by many coaches. It very likely is influenced by confirmation bias, though.

A while back Mark also wrote about some research into whether timeouts impact the next point. That piece was was based on some findings from basketball which suggest they are actually counterproductive. Not content to stop there, Mark followed up with additional posts here, here, here, here, here, and here. The timeout subject was also tackled by a researcher in a presentation at the 2016 AVCA convention.

This research is definitely a good start. That’s all it is for the moment, though. I’d like to go down some other research paths with respect to volleyball. What do you think? What question(s) do you have that can be addressed by analyzing available data?

Your mandate and situation influences your coaching approach

A post by Coach Rey explores the anti-Moneyball idea with respect to Kansas City Royals manager Ned Yost. It references a NY Times article on the subject which talks about how Yost doesn’t operate based on analytics. This is something for which he was regularly criticized. The general thrust of both the blog post and the article seems to be that you can win without relying on the stats.

Here’s my own takeaway, though.

Yost specifically talks about making long-term decisions with respect to player development. He wasn’t trying to win every game. His mandate in both of his most recent two jobs was to develop young teams. When that is your priority, you make different types of decisions than you do when trying to win the largest possible number of games.

The same sort of thing applies in the type of situation some of us are in as coaches. It’s a case of how we play at season’s end being more important than how we play today. Mark Lebedew in his time at BR Volleys had the luxury of being able to mix up his lineup from match to match in German league play. He knew his team was stronger than others, and it was all about the playoffs. That allowed him to spread playing time and develop they younger guys.

I was in a similar sort of situation at Svedala in that every team made the playoffs. Yes, there’s an advantage to finishing higher in the regular season standings. Yes, we also wanted to qualify for the mid-season Gran Prix by being top 4 at the halfway point. The big objective was going after the league championship, though, so I could take a somewhat more developmental than “win now” attitude early in the season.

Obviously, not everyone has that luxury. When I was coaching at Brown there was not conference tournament. It was just the regular season schedule. When I was at Exeter for the first two years we needed to finish in the top three in our league to reach Championships and were VERY motivated to not finish 3rd to avoid having a first round playoff against one of the winners of the other leagues. In cases like that, winning now is very important.

What about you? What sort of situation do you operate in?

Coaching Log – Nov 2, 2015

This is an entry in my volleyball coaching log for 2015-16.

There was only one other match besides our last week, with Örebro beating RIG. No surprise there. That pushed them to the top of the Eliteserie table on 8 points to our 7. They’ve played 4 matches, though, to our 3. There were no other Oresundliga matches, so Engelholm’s win over us sees them to of the table there on 7 points from 3 matches. Everyone else only has 2 matches, aside from Gislaved (our opponent this week) who is 0-3 thus far.

I spent a lot of time in the latter part of the prior week working on getting hold of and analyzing the video from our match. Technology issues really slowed things down. I basically ended up manually going through every attack for both teams and coding the rotation, the pass/dig, the type of set, the type of attack, and the outcome. I really wanted to take a look at our set distribution and effectiveness.

Two things really jumped out. First, the disparity between when we pass well and when we don’t. Obvious, you expect to be more effective with better passes. When we passed or dug a 3 (see the rating system) we had a 55% kill rate with only 7% errors or blocked. When our pass/dig was a 2 we had a 45% kills rate with only 5% errors/blocks. When we passed/dug a 1 our kills were only 8% and the error/block rate was 20%. I’d like to see us push the 3s up maybe 10 points and get the 2s up to 50%. As for the 3s, I think we need to flip the percentages around, more or less.

The other thing that jumped out was how few middle quicks we ran – just 7 total. Part of that was a match-up thing for that match as we ran more slides, but even still those numbers need to be much higher. I also wasn’t happy with our overall numbers for right side attacks.

Monday
We had a pair of guest players in training. They both play at RIG, but were on a school break. One of them is from Svedala and has played in various teams with the Swedish players. The other is from Ystad, which is the home town of one our players, so again known to most of the groups. That gave us an even 12.

Ball-handling was a major focus for this session. After warm-ups we did Continuous Cross-Court Digging. I instructed the hitters to pick up the power of the attacks to increase the challenge. From there we shifted to some serving and passing. I began with the setters and middles (4 players) serving to everyone else on two sides – so 3 passers and a target, with a passer rotating to target after two good passes.

To continue working on middle quicks I then shifted the setters and MBs to the net. I left two passers in on each side covering half the court, with the other two players serving. Again, a passer rotated out to go serve after two good passes. At the tail end I added in a pipe attack.

I then shifted that into Speedball Winners with fixed setters and MBs. This time the setters were back row to work on the timing with their transitions. We finished with Bingo-Bango-Bongo. I mainly did that to work on Rotation 1 when we’re in transition after reception (so the OH in 2 and OPP in 4), but we did all the rotations.

Tuesday
I met with the team for about 15 minutes before training to talk about my analysis of the video from our last match. We talked about looking to get kills more out of our good passes and digs (mainly about speeding up the offense) and in better dealing with things when we’re out of system. In the case of the latter the focus was on bettering the ball and getting our back row attackers more active. We also talked a bit about improving our block/defense, in particular with regards to how the defenders in 6 play.

Just one of the RIG players in training this session, plus our part-time MB. After warm-ups and pre-hab exercises I had the player do some partner 2-touch pepper over the net to warm-up their shoulders. That was followed by a variation of passing triplets. In this case, to work on proper platform angles I had them dig/pass down balls coming at them down the line (1 to 5, 5 to 1) to properly redirect them toward target in Zone 2.5.

My American OH had an idea of a drill to do to work on block positioning and penetration, so I let her run that next. Basically, it was liberos hitting from boxes at the pins (they tried from the ground, but they were having to hit upwards too much) against a 3-person block with the MB closing to double and the off blocking moving into defense. The blockers got a point for a stuff block or a touch that could be played by one of the 3 involved. They went until they got 50 points.

From there I broke the group in half to work on offense. The OHs and Liberos were on one court with one setter. I had one of each serving, with the other two OHs and the other Libero in serve reception working on faster sets. The OHs rotated on 5 good kills and at the same time the Liberos swapped.

On the other court I had the MBs and OPPs with the other setter. They worked on middle and right side attacks with the OPPs passing coach-tossed balls. There were 3 MBs, so one was the attacker for 5 good balls, while one was the opposing middle and the other the opposing OH for blocking purposes. After the OHs on the other court each got 1 round of 5 good sets, I had the setters switch courts. The MB/OPP group came up with some interesting play options that are worth developing further.

We finished up with 22 v 22 to get both serve receive and transition work in. The games were quite competitive, so we only got through rotations 1, 4, and 6. I had the MBs rotate so one of them was playing defense each round on the team with only one OH.

Wednesday
This was a very technically oriented session with just the core group. After warming up I had them do some positional digging in groups. Basically, on a rotating basis each player dug a ball from 2 and one from 4 (hitters on boxes), with a group target of net good digs (-1 for overpasses). We went through positions 5, 1, and 6. Then we did combined digging of cross-court balls for defenders in 1 and 2 and 5 and 6 to work on seams and tips.

I then split the team to have the setters and MBs working on the timing of 1s. The others did serving and passing. Two players passed half the court. There was a target and the other three players serving. The passing pair needed to get to +20. They got 2 points for a 3 pass, 1 point for a 2 pass, -1 for an ace or overpass.

The last exercise was a 5 v 5 game. One side had an OH, a Setter, a Libero and both MBs. The Setter was back row, leaving one MB to block in 2. The other side had one OH and a RS hitter front row, with a Setter, Libero, and OH back row. I coach served every ball to the no-MB side. The focus was on faster balls to the pin hitters, plus pipes. The receiving team got a point if they got a kill or caused the defensive team to dig 1 ball, otherwise the defensive team got the point. They played out the rallies, but I scored only based on the first attack. We played 3 games to 10 to rotate the OHs through, also moving the setters and liberos around.

The players said they liked the training – all the technical work. Honestly, though, a big part of its motivation was that they worked really hard the night before and had weight training immediately prior to practice this night. If we’d pushed things with another high intensity game-play session they probably wouldn’t have lasted very long – at least at the level of focus and effort I’d want to see.

Thursday
Our Saturday opposition (Gislaved) played a match on Tuesday night. After getting it downloaded, I created a trimmed version (no timeouts, no between set breaks, etc.) and then focused on just the 2nd set when Gislaved was nearest the camera. I made some notations in the video, then posted it up for the players to watch (with a link to see the full one if they wanted).

This time I went back to focusing on how we can attack their block/defense. For the last match I isolated the swings of their major attackers, but this time around I wanted to focus on what we’re doing. Basically, a return to the approach I was taking before.

Friday
No no returning guest players, meaning only had 9 in training (see Other stuff for the reason it wasn’t 10). After warm-up and pre-hab and some ball-handling I had the players do back court winners 3s to work on both back row attacking and defending against it as we’d likely see plenty of it on Saturday. I then ran a servers vs passers game with the OHs and RS going against the Setters, MBs, and Libero.

Basically, the rest of the session was spent running through the rotations in a 5 v 4 fashion. On the 4 side I had the Setter front row along with a MB and OH, and an OH in 6. On the 5 side the Setter was back row with the Libero and 3 front row attackers. The OH on the 4 side served to start each sequence. After that rally ended, I gave the 4 side a free ball. Rallies were played out where possible, but the scoring was based based on the first attack (kill or + attack = point for the receiving team, otherwise point for the defending side, except for a covered block which was no point for either side). We played games to 10, flipping the Setter and RS player on the 5 side, and switching the back row OH from the 4 side with the OH from the 5 side. The teams were set so they matched rotation personnel.

The offense run during that last phase was quite aggressive as we started working the OHs in some different faster sets.

Saturday
Match Day started with team photos at 10:00. We did them once in September for league website purposes, but this was actually the official shooting with all the right numbers, etc. That was followed by our hour of serve & pass at 11:00, then team lunch following. The match was at 15:00, with Gislaved in town for the first of our 4 matches against them this regular season. You may recall they were the first team we played in pre-season.

As seems to be the case so often, we really got on top of the other team early in the match. Our serving pressure really put Gislaved off their game – even forcing their best hitter to be subbed out in both sets 1 and 2 – resulting in 25-15 and 25-20 scorelines. Set 3 was one of those weird ones that happens at times. They got on top of us early and we never could claw back. It ended up 15-25. The fourth set was pretty tight, with some back and forth momentum swings. We were able to pull it out 25-22 in the end, though.

The funny thing is our passing was markedly better in sets 3 and 4 than in the first two. We sided out better in the weaker passing sets than in the better ones, though. We struggled with block positioning when we played this team the first time and did so again once more. Yes, we got 9 blocks, but that’s actually a bit below our average.

Thoughts and observations
Overall the energy and intensity level was good. We did some interesting new things on offense in terms of diversifying the attack and going faster. I still think our defense could be much better in certain respects, though being better in the block is part of that.

Other stuff
One of my liberos spoke with me before training on Tuesday. She’s going to be starting a new job in the not-too-distant future. The result is that she probably won’t be staying with the team. In fact, on Friday at about midday she let the team know via the Facebook group that she wouldn’t be training that night or attending Saturday’s match. Surprise! That sees us down to 9 for the core squad, which is tough. I need to look at options for getting more alternative bodies involved.

Old school analysis

In modern volleyball there are any number of applications and technological tools that can be applied to coaching. DataVolley, for example, is widely used for statistical and video analysis. There is also VolleyMetrics, for those who want to outsource this time of work. These things cost money, though, and not everyone has that – or the staff to put it to use. That means sometimes you have to revert to old school methods.

That’s what I found myself doing once when I coached Svedala in Sweden. I wanted to do a thorough analysis of a recent match. It was a disappointing loss. We were up 2-0 and had leads late in both Sets 3 and 4. Not having a better option at hand, I went through the video. I wrote out the rotation, pass/dig rating, set, attack type, and result for every non-serve ball played over the net by us and the opposition. I then plugged that into a spreadsheet so I could break it down.

Yeah. Fun stuff! :-/

We did have official match stats, though they are of dubious value – especially the first version of them. We also used SoloStats on the bench during matches to track serve reception stats and examine rotation performance. I needed, though, to be able to drill down and cross-section things to get a better handle on what we’re doing well and what not so well.

I hoped we could eventually come up with a better way to do all this, but it was a small club with a limited budget. More or less, I had to go that route for all our matches where I wanted more detailed analysis.

What were my findings, you ask?

The big one was we needed to be much better out of system. We got kills only 8% of the time and it was either errors or blocks on 20% of the plays. And therein is the value of statistics – even very basic or hand tallied ones. They give you ideas for things to look at more closely or to spend more time on.