Attackers above the norm

bigatkwar

So I’m procrastinating on my thesis again and wanted to see how attackers stacked up relative to the league – on a per team basis. I’ve measured success via Eff Change in this case, built using league averages. For example, a Big Ten player who attacks in zone 2 after a perfect dig in transition is expected to have an efficiency of 0.274. So if you kill the ball on this perfect dig, your Eff Change is 1.00 minus 0.274 or 0.726. Or if you’re attacking out of zone 4 after a perfect pass (R#) you’re expected to hit 0.335. So if you take that perfect pass and hit into a “good” dig (an output eff of 0.011), you’ve actually dropped your efficiency by 0.324.

All of these individual input to output situations are accounted for – as well as the zone you attack out of. From there I can create an aggregate average Eff Change for each attacker based on these expectations.

Not surprising, the top teams are mostly in the green – and Michigan State holds on tightly to their stereotype of being super terminal in both directions. For the most part, setters, ease up on the dumping…Kelly Hunter…..

pacatkwar

Here’s again what this looks like for the Pac 12. Of course the league averages aren’t the same between the two leagues, so just be aware of that.

I’ll just leave these here and let you come to your own conclusions, but the gist of these visuals is: how well do your hitters handle the situations they’re dealt, relative to the league. As you can see, this is a much quicker way to see who is terminal when they’re supposed to be and who understands swing management when they need to play it safe.

Advertisements

Efficiency Change – Pac 12

Book1.jpg

Here’s just a quick snapshot of Eff Change taken from a late September match between UCLA and USC this season.

I’ve explained the idea behind Eff Change in a couple posts but figured I’d throw it out in a Pac 12 post for fun. The idea plays off the FBSO and Service Error concept that different teams have different strengths/weaknesses. I don’t know enough about the Pac 12 to generalize, but let’s say USC is a stellar offensive team, but they struggle to dig balls. With that idea in mind, getting a stuff block against USC – or perhaps serving them out of their in-system offense might be incredibly valuable, given their offensive strength. But on the flip side, if USC is a team lacking great ball control defenders, then maybe failing to rack up kills against them is pretty bad, seeing how USC already has a lower defensive baseline in this example.

You should go read the original post about Eff Change if you haven’t since it explains in more detail how all of this is calculated…but the idea is that you account for the two teams that are playing – as well as the specific context within the rally (good block touch, perfect dig, poor reception, etc). So you look at both in the input and output contacts. For an attacker, you look at the pass or dig that precedes the attack as the input – and the output is the block touch, kill, error, or dig quality that follows the attack. You then subtract the efficiency of the output from the initial efficiency of the input to create Eff Change. So if you have an input of a poor reception (maybe that input eff is 0.100 since it’s tough to score with that pass quality) but you take that and tool the block for a kill (a kill has an output eff of 1.00 because that team always wins the rally). So you started at 0.100 and ended with 1.00; so that 1-0.1 for an Eff Change of .900. This is a big deal since you started with a weak position and improved the outcome drastically. That’s the gist of Eff Change.

For some of the touches in that above UCLA/USC match, you’ll notice that Eff Change is actually above 1 and below -1 for some touches. So for Frager’s ace, UCLA is actually expected to lose with that input contact more often than not. So UCLA starts with an Input Eff of -0.247 (not surprising as the receiving team is almost always favored) but because she serves an ace (Output Eff of 1.00) her net Eff Change is actually over 1.

Anyway, just some stuff to think about for those who haven’t cruised through all the posts. Let me know what you think / if you have questions. I can talk about this stuff all day.

Blocking Responsibilities – Pac 12

pac12respblocker

Just to preface this post, it’s basically the same as the responsible blocker one I put up earlier using the Big Ten data. That post explains exactly how responsibilities are split between blockers against specific types of sets.

My Pac 12 knowledge is admittedly lacking and I had to google some of these kids. Schoenlein is a senior outside for WSU, Lutz you should know from Stanford (and who is apparently touching 11 feet at the moment), Willow Johnson is a freshman RS for Oregon (and Randy Johnson’s daughter, fun fact).  Woodford is an OH with WSU I believe and Plummer is the Stanford OH you’re likely familiar with.

Just a reminder that since there is no designation of position (OH, MB, RS, S) in DataVolley, I am relying on spacing from the setter. This is compromise that has to be made in order to do the analysis – but inherently fails us when we look at a team like Stanford who has Fitzmorris and Lutz sometimes floating between middle and rightside. Sorry…

What you see on the left is how attackers hit against the blocker in question, whether there is a registered block touch or not. So if you find Willow Johnson (just a cool name so I’m gonna keep using her), you’ll see that when hitters attack her down the line or into the portion of the seam she’s responsible for as a RS blocker, she holds them to one of the lowest Attack Efficiencies in the league.

What you see on the right is the Efficiency Change as a result of the block move. From earlier posts on Eff Change you may understand that this gives more credit to a blocker if they take a strong position for the attacker and neutralize it or turn it into a good position for the defense – for example, if the offensive team usually hits .500 on the Go and Willow Johnson gets a stuff block to terminate the rally, her Eff Change will be her output, minus the input. In this case the input is -.500 because your defense loses at that rate against the Go, but the output is 1.00 because your defense always wins when you get a stuff block. This gives WILLOW JOHNSON an Eff Change for this block move of 1.5, which is huge. It also hurts the blocker more if they do something stupid when the offense is already in trouble, like netting against an out of system attack. Don’t do that…

You’ll see from the EffChangeBlock column is that the best blockers add value each time an attacker targets the portion of the court they’re responsible for. This might be via good touches that help your transition offense or it might be just not getting tooled and leaving clean lines for your defenders to fill.

You can use EffChangeBlock to potentially find undervalued blockers. Jenelle Jordan, a senior MB from Cal for example. While attackers hit 0.268 against her, more than double what people hit against the top 3 in that category, she quietly has a positive impact on attacks into her zones. To be fair, these Eff Change stats account for the team you’re on as well as the team you’re playing, and Cal’s numbers are likely to be deflated relative to the rest of the league. So if they’re a subpar digging team, that increases the value of every block touch Jordan gets…

Anyway, just wanted to throw this out there and let people check it out.

Skill Efficiencies by Team

effbyskillTEAM.jpg

This viz is the follow-up to the skill breakdown per player post from a couple days ago; a composite look at how each team fares on a per skill basis. The teams you see are ordered by their finishes in Big Ten conference play in the 2016 season and the data used is from conference matches only. Blocking is calculated via the responsible blocker metric rather than just block touches, meaning there are attempts counted on untouched attacks if the ball went to an area a blocker was responsible for.

The first thing you might notice is that Rutgers isn’t very good. Nor is Northwestern. On average, these teams hurt their Win Efficiencies each time they contact the ball, no matter the skill. The second thing you might notice is that all teams have negative Eff Changes for Reception. I need to drill down into that skill and see why all teams hurt themselves here, even passing powerhouses like Nebraska and Wisconsin. Maybe later this weekend.

Some interesting similarities: all teams have about the same percentage breakdown per skill – meaning that each is expected to perform each skill at about the same frequency, no matter the talent level of the team. This can be incredibly useful to understand if you’re looking for more efficient returns on your time investment in practice. Another interesting trend is the value that attacking adds for 3 of the top 4 teams. While Nebraska is good at everything, the other 3 excel offensively and this is evidenced by the proportion attacking holds in their composite view of added value.

The middle cluster of teams also have similar characteristics. They all perform exceedingly average, being only slightly above/below .000 per skill (minus Reception). Then in the bottom contingent, these teams all lose value on almost all skills – except for Purdue, who, if you’ve watched them, tends to fluctuate on a per match basis.

There are a bunch of different threads you could tug at in this viz, but I’ll let you form your own thoughts while looking at this. If something like this happens to spark a cool thought or helps you see an idea from a different angle, let me know about it and let’s see if we can dig deeper.

 

Attacking Theory

koobsattacking

Hey Kolby.

So I’ve been throwing out the question to some friends about what kinds of stats they’d like to see but have never had access to. Attacking is naturally something that comes up as it’s the most common way to score points and thus a deeper understanding would be beneficial.

Above is my proposed progression for how we should view attacking. These are the 2016 Big Ten attacking stats for individuals with more than 100 swings – and I had to chop out the bottom 3 individuals, (Enners, Duffin, and Fletcher) who you’ll only be upset about if you’re affiliated with Rutgers, because they were skewing my charts. #science

The column on the left is straight attacking efficiency like we’re all accustomed to seeing it. Kills minus errors, divided by attempts. Many of us are also familiar with “KOBE” charts (Kill, Zero, Block, Error) for looking at attacking and the question often arises, what happens with those “zero” attacks – the non-terminal swings? Clearly chipping the ball at the libero is a worse choice than chipping it at the opposing setter, but is there a way to quantify the value of this decision?

When I arrived at Illinois, Hambly introduced me to “True Efficiency,” a way to look at how your opponent attacks in transition after your initial swing. The general idea being: take your normal attack efficiency and subtract (your opponent’s hitting efficiency multiplied by the percentage of times your opponent gets a swing after you attack). For most attackers, this lowers their overall efficiency, unless you happen to be “Magic Mo” Criswell – an outside with the 2014 Illinois team who spatched balls so awkwardly at opponents that her True Efficiency was actually higher than her season Attack Efficiency. This means that when Mo hits at you, you’re likely to hit negative in transition. Pretty crazy when you think about it.

True Efficiency is certainly an improvement over traditional attack efficiency because we can all agree that the outside whose swing management puts the ball on the setter rather than the libero should be rewarded for that choice – since most teams defend better against an out of system high ball than they do an in-system attack.

The next step in this progression is to look at what I’ve been calling Efficiency Change. It’s the idea that you factor in the circumstances an attacker is dealt (perfect pass, poor dig, etc) and you look at what type of output they get you (kill, error, poor dig, perfect cover, etc). As we’ve established previously, each of these inputs and outputs have specific “Win Efficiencies” per team. A perfect dig is worth X to Nebraska and Y to Michigan State and Z to Rutgers. To tug this thread some more, a perfect dig for Nebraska playing Rutgers has a specific value – so does a poor dig for Minnesota against Penn State. These all hold different values. Take it a step further and look at the output from the attack. Getting a kill on a high ball against Nebraska or Minnesota is tough. Getting a kill on a high ball against Rutgers is less challenging. These circumstances should be valued differently.

This is what Efficiency Change seeks to answer. It looks at 4 things (where Team X, Illinois, is attacking against team Y, Nebraska):

1. How well does team X typically do in this situation against all teams? (How well does Illinois hit Go’s in general off a perfect pass?)

2. How well do all teams in this situation do against team Y? (How well do all teams in general hit Go’s off a perfect pass when playing Nebraska?)

3. How well does team Y do against all teams with the output from the attack? (How well does Nebraska do against all teams when they get a perfect dig in trans?)

4.How well do all teams do against team X with the output from the attack? (How well do all teams do with a perfect dig in trans when playing Illinois?)

The first two seek to account for the offensive strength of the attacking team and the defensive strength of the opponent. These determine Input Efficiency. The latter two seek to account for transition strength of the opponent and the defensive strength of the attacking team. These determine Output Efficiency

Once we have our Input and Output Efficiencies, we take (Output Eff – Input Eff) to calculate Efficiency Change. Does our outcome leave us with a better chance to win the possession than the chance we had going in?

How well does an attacker handle the situation they are dealt, in the context of what their contact is worth to both teams. You can chip a ball to the Rutgers libero and still be ok. You can’t do that to Minnesota. That’s why we look at the output efficiency – and we’ve already established that different teams are more or less difficult to attack against. We must account for all these things.

For those that care about the details, I also account for whether each team is in a 3 hitter or 2 hitter situation – as well as the start zone of the attack. Ohio State using Sandbothe out of zone 3 on a perfect pass in a 3 hitter rotation is incredibly tough to slow down. Illinois using Davis on the pipe out of zone 8 off a poor pass in a 2 hitter situation is pretty easy to slow down…mainly because she’s one of our “wee ones” who comes in for RS Naya Crittenden in the backrow. That’s why we account for all these things…not just the blanket overall efficiencies of the teams.

Back to the viz really quickly. Haleigh Washington is a baller. She’s the highest in all 3 categories so whatever metric you subscribe to, you can’t tell me she doesn’t handle her business. But efficiency change allows us to find hitters who contribute the most value, even though they may not carry the heaviest load or score the most points. While Molly Lohman leads Minnesota in both Atk Eff and True Eff, it’s NPOY Sarah Wilhite who has the 4th highest Eff Change per attack in the Big Ten. Lohman does alright, but because she mostly receives sets under great circumstances, she’s expected to perform excellently each time. Looking from this alternate viewpoint, it’s actually Wilhite and her gnarly range that provides the greatest value in each situation she’s dealt.

Eff Change also helps us weed out Purdue’s Blake Mohler and Ashley Evans. Mohler’s numbers look good until you look at her Eff Change – where she actually loses value for her team when she swings – meaning she likely takes great situations for Purdue and doesn’t terminate like she should. Ashley Evans just needs to pump the brakes on the dump. Your front row is like 6’5″ across the board. There are better options…

Anyway, that’s where I’ll leave it for now. Keep nibbling on this food for thought.

One more thing: Should we only recruit middles named Haleigh?

Specialists or Generalists?

 

effbyskill3Yes. I know this looks like some tutti-frutti clown vomit. Thanks.

A college buddy of mine, Teddy Niemira, posed a question revolving around developing players’ all around game or merely enhancing strengths as a better use of a coach’s time. Do you want someone who is pretty good at everything or elite at one thing? I didn’t take this in exactly the direction Teddy wanted, but that’s ok, it’s my blog.

What you see above is a breakdown of the top 50 players as defined by their Efficiency Change per contact. Players with fewer than 500 touches have been removed and for context, Carlini had the most at 3411 and names that fell under 500 were the likes of: Mahlke, Goehner, Brashear, Stackhouse, Halm, and Wenz.

On the left, you see the how each player does on a per-skill basis in terms of Eff Change. Naturally, setters aren’t great at serve receive, Cuttino isn’t a good setter, and so on. This column is only how well players do on all skills – regardless of how often they actually use these skills.

In the middle, you see the percentage of each skill as a portion of all the player’s touches. Keep in mind the responsible blocker post for earlier, meaning that even attacks that are not touched by the blocker register as an Eff Change for the responsible blocker in the situation – and this these numbers are not perfect for individuals who don’t block in their “traditional” locations (middles blocking right, etc – looking at you Mr. McCutcheon).

On the right is a combination of the two. The column takes into account how much value the player adds per contact, multiplied by the frequency at which they make these contacts (the middle column).

Personally, I think it’s easy to get lost in this viz and forget the question. So while there’s a lot of cool info in there, I’ll try address what I’m supposed to be addressing: would we rather have specialists or generalists?

To answer this, I want to start with the Touch% per skill. I want to know if the role the athlete is playing is by it’s nature, a specialist role or a generalist role. If you direct your attention to #3 on the list, Faye Adelaja, you’ll see that Attacking and Blocking make up 92% of her responsibilities – and Atkinson, just below her, is called upon in the same fashion. That may be no surprise if you watched Purdue this season. They are a team composed primarily of specialists. Their front row is there to hit and block balls. Their backrow is there to pass and defend.

Outside hitters on the other hand, tend to have a multitude of skills with decently even usage percentages – same with liberos. Take a look at Foecke, JWO, and Wilhite. These players have their hand in just about all the skills, and with that in mind, we would want them to good at each of the skills, which they mostly are. Players like Detering with Penn State, we want to be elite at the skills they focus on. As we see from Detering’s breakdown, she adds a ton of value per set, her most common skill by far. This is what we’d want from a specialist. Generalists like Foecke, JWO, and Wilhite add value per skill in multiple areas, though not necessarily to the same degree to specialists like Washington.

My initial takeaway here is that for Middles, Setters, and Rightsides – be elite at the skills you’re supposed to be elite at: attack/block for the hitters, setting for setters. For position players with their fingers in everything, be ok with adding less value per contact, but add value for all skills. I’d like to break down these graphics by position to see how outsides lineup against middles skill-wise. Or how teams in general do with all these skills and how that relates to their finish in the Big Ten – or nationally. And down the rabbit hole we go…

Service Error and Eff Change

effchange-team

*Check out the Eff Change posts if you haven’t. Output, Input, Eff Change

Returning to the very first question – How aggressively do we want to serve?

Here is what team service error% looks like graphed against the average Eff Change per serve in each match. Again we see an inverse relationship between the two variables, suggesting that an increase in SE% tends to result in less effective serving.

To be clear, this is not an absolute. Michigan State arguably had the best serving performance with around 0.175 improvement in Win Efficiency per serve, while also missing more than 5% of their serves. But as a general trend, teams with better serving performances also tended to have lower error from the endline.

effchange-player

This trend becomes even more pronounced when you drill down into the player specific view of each match. After removing servers with fewer than 5 serves in the match, the above results echo what we’ve already seen. Making fewer errors is associated with typically better Eff Changes.

One of the reasons this may feel strange is that we’ve become accustomed to service error. Again, this isn’t to say that those who make errors from the line can’t have good serving performances – just look at those Bailey and Kranda matches at the top – but as an overarching trend: more risk, less reward.

A logical next step may be to break this down by type of serve. Does the increase in risk by hitting a jump float make it less valuable than a “flean” standing float that achieves similar outcomes (hey Karch). Is hitting a full jump spin serve worth the risk? And how does this change when you look at the men’s collegiate game?

Or is it more important to look at who you’re serving against? Adding 5% more error against Michigan State add more value than adding 5% more error against Nebraska. If Nebraska is going to pass and attack similarly, then why add the errors? But if that 5% is what drops MSU from hitting .350 in serve receive to hitting .150 because they’re not running quicks from off the net or something – then maybe that’s where the value lies, in the specific matchups. And eventually…in the specific rotations?

Food for thought. Nibble away…