Dig Percentage & Efficiency Change

dig-error

Dig Percentage is something that comes up as a way to measure defensive efficiency. You might look at two similar passers and see which one of them shanks digs at a lower rate as a means of evaluation. As we can see from the viz above, Wilhite, JWO, Albrecht, and Maloney (3 of whom play for Nebraska, the other happens to be the NPOY) hold the top slots in terms of minimal dig error percentage.

There is certainly the point to be made that defining “dig error” is subjective – and what should count as playable vs. undiggable. But at least to the initial eye test, the results seem reasonable. Just take a second to think it through, when you attack and Sarah Wilhite is on the receiving end, that ball comes up 87% perfect of the time. I know there are club coaches out there who are ecstatic if the ball comes up half the time when it’s hit at their players – so 87% is pretty silly when you really think about it.

But the more you think about it, the next logical question is what’s the quality of these digs? Because if we’re just lipping tough balls and sending back freeballs instead of transition swings, 87% is less impressive. So let’s look at the breakdown:

digbreakdown

Apologies, I know it’s not the simplest to read. But they’re in the same order!

Basically, the greater space the pink, orange, brown, and red take up, the better. For those colors, your team got a swing off the dig/cover. If it’s a D- or Cover-, you overpassed it or did not get a swing. The grey D= and kinda teal Cover= represent dig errors that ended the rally.

Before we all jump on the setters for slacking in the perfect dig department, keep in mind these players typically dig to 10-15 feet on purpose so the libero can use her hands. They’re not aiming for the area of the court that we would usually  refer to as a perfect dig. It’s clearly a lack situational context but we make the most of it.

I think this breakdown helps to the same degree that looking at passer ratings does. Somewhat, but not all that much really. What we actually care about is what those contacts are worth to their respective teams. If a player is 50-50 D# and D+ does it matter if her team hits the exact same off both dig qualities? No. She could be 100% D+ in that example and she should be valued the same. So let’s look at how these major defensive players actually “better the ball.” Hey Owen, you’re welcome. GDH.

dig-eff-change

So now what we’re looking at is the ranking of the average eff change each player brings per dig. Remember that Eff Change is calculated by looking at the context of the situation. For a dig, we look at the quality of the 1st touch for the offensive team, the start zone of the attack, and whether each team is in a 3 hitter or 2 hitter situation (and if there’s a block touch). So if a team hits really well when they get a perfect pass and set the quick in zone3, getting a clean dig on that ball  is a big shift in efficiency. On the other hand, digging a high pipe attack after a poor pass carries much less weight. And again, we look at how well you’re expected to handle a situation and how well all teams playing against your same opponent tend to handle the situation – so we get a better feel for what’s expected.

What’s interesting about the viz above is that the players who add the most value defensively aren’t necessarily the ones who shank the fewest digs. While freaking Sarah Wilhite is just really good at volleyball, the other 3 Nebraska kids we mentioned earlier have fallen outside the top 5 and now other DS/Libero players have entered the mix – the ones we might assume would make the big digs in tough situations.

Personally, I think using Eff Change here helps us ask better questions (dammit Hambly). We see who digs balls that consistently add value to their teams, rather than just asking who keeps the ball in play.

Advertisements

Blocking Responsibilities – Big Ten

One of the things people have never been able to look at in DataVolley is the way to examine blocker responsibilities. The reason this is an issue is that if you don’t have a block contact coded in DataVolley, nobody knows that you were at the net when the attack happened.

The goal of uncovering the responsible blocker is to account for how attackers hit when they attack at an area of court you are “responsible” for blocking. If you’re the OH and your team is getting crushed by slide hitters taking swings down the line, just because you’re not getting tooled doesn’t mean you’re not blowing it. Or just because you’re not blocking balls doesn’t mean you aren’t forcing your opponent into bad swings or easy offspeed shots.

The way I solved this is to look at all players and their distribution around the team’s setter to identify the OH1/OH2, MB1/MB2, and RS. By knowing who is in each “slot” I also know who the front row blockers are if I know which rotation each team is in.

The immediate flaw that we have to deal with: I am making the assumption that the order is: S, OH1, MB2, RS, OH2, MB1. Some teams, like Stanford/Minnesota/Michigan/etc like to mess with this order and end up having middles blocking right, rights in the middle, etc. Or if you’re Purdue, you just have Cuttino doing whatever she wants and chasing the biggest hitter. The reason this is an issue however, is that against a Go set to the OH hitter, I make the assumption that the player in the RS “position” is the pin blocker in this situation. Because I only know the lineup and not the physical location of the blocker, we have to live with some error (until we get player tracking like SportsVU). So until we better this solution, players like Hannah Tapp, Danielle Cuttino, Audriana Fitzmorris, etc will be victims of this inadequacy. Sorry.

blockingresp

Here’s how I have the attack codes split up by blocker. I categorized attacks outside of zone 3 in 3 ways: at the pin, outside in, and straight in. At the pin are the top two courts, outside in are the next two (32s, inside sets to the RS, etc), and straight in is the third set (back 1, 3). The bottom set is for attacks in zone 3 or 8. I did my best to align blockers to sets they’re most likely to block – and have shaded in the area behind them as I saw fit.

The shading of the court is certainly open to discussion and can be altered. DataVolley has 9 zones and 4 subzones meaning there are 36 boxes to designate. Boxes in red are the responsibility of the pin blocker. Boxes in blue are the middle blocker.

If we look at blockers who were responsible for at least 150 attacks during the 2016 Big Ten season, here’s what we get:

blockatkeff

The first column is the actual attack efficiency when people hit at the responsible blocker or into a zone she is responsible for. Without being too biased, “Sporty J” Poulter absolutely handles her business. Carlini is alright too…

Personally I’m a little surprised to see so many pin blockers in the top echelon, seeing how I gave them a good chunk of court to be responsible for on almost all attacks.

*to clarify, regardless of where the ball lands, if there is a block touch recorded in the DataVolley file, that person making the touch is the responsible blocker. This way, if a good pin blocker dives huge into the angle and stuffs the ball, that pin blocker gets credit – not the middle who had better things to do…

The column on the right is the Eff Change for the responsible blocker. This means that getting stuffs and good touches when the opponent is in a more advantageous starting position nets the blocker more love. This also means that netting when your opponent is busy spatching a high ball out of bounds loses you a lot of love.

Would be interesting to break down where most of the Eff Change for the blockers comes from. Is it stuff blocks against in-system attacks? Not messing things up when your opponent is in trouble? Getting a ton of block touches in general? I’ll have to drill down on that one.

Intro to Efficiency Change

wineff

*if you haven’t, go read Output Eff first or you’ll be lost.

Here are the same two rallies we’ve been looking at for Output and Input Eff. The only difference is that we’ve calculated Eff Change via (Output Eff – Input Eff).

The idea here is to look at the situation you were dealt and the result you got based on your contact. So for Skjodt’s first pass (R+), this touch is just a little better than our expectation and she raises her team’s Win Efficiency by around 0.050. Later in the rally, Jenna Lerg’s perfect handling of a freeball (F#) sets Michigan up nicely to win the rally – however when Mahlke attacks, Gillis get a great block touch, turning a bad situation for Wisconsin into a good one (worth .400 to Wisconsin). To terminate the rally, Gillis takes MacDonald’s D+ and gets a kill. The reason that Gillis’s swing isn’t worth 1.00 is because Wisconsin on a D+ already sets up Wisconsin nicely to win the point (eff = 0.150). So because Gillis starts from this advantageous standpoint, her kill is worth 1.00 – .150, leaving her with a value for her kill of 0.850.

To pull on this same thread, Haggerty’s setting error to end the second rally actually costs Wisconsin more than a point because she took a good situation for Wisconsin (D+) and immediately lost the rally for them – leaving her with an Eff Change of her setting error of -1.150. Moral of the story, get out of Carlini’s way.

This is kind of a gnarly way to look at the game – and individual rallies. It takes a minute to let these ideas sink in, but let them marinate for a little and see what you think.

Input Win Efficiency

inputeff

*if you haven’t, go read Output Eff first or you’ll be lost.

Here is the other half of the puzzle. Input Efficiency.

Input Eff works the same way as Output Eff except we look at the previous contact to determine the situation you were dealt. The idea here is that attacking after a poor pass is not the same as attacking after a perfect pass. Or that digging a great quick attack is much easier if there is a great block touch to slow it down first – it’s much tougher to dig if there’s no block touch.

So here again we look at both sides.

1. How well does your team making the contact do against all teams when the Input Contact is X. How well do you attack when your InputContact is R#, D+, F!, etc.

2.How well do all teams do against your opponent when the input contact is X.

For serving/receiving there really is no Input Contact so we just use Point Scoring Eff or FBSO eff in general when you go back to serve or your team is in serve receive respectively.

If we look at the viz we see that Gillis has two block touches in that first rally. The associated input efficiencies come from the first touch in those possessions where Michigan passes a R+ and then Lerg passes a F#. Because Wisconsin is more likely to lose following an opponent F# (-0.410), that is the Input Eff for Gillis’s block touch following an opponent F#. Wisconsin is less likely to lose the rally following an opponent R+ (-.290) and this is the Input Eff for Gillis on the first block touch.

If these all needs way more explanation, someone let me know. Please.

In the next post, we’ll combine Input and Output Eff to determine how players add or subtract value on their respective touches, while accounting for the circumstances they are dealt. How well do OHs do when given a poor pass to hit off of? How well do defenders do when digging after an opponent perfect pass vs. terrible pass? We’ll look at these things as we introduce Efficiency Change.

Output Win Efficiency

outputeff

Here is the extension of the FBSO eff post – but applied to all types of touches.

For every contact, we find the resulting Win Efficiency. So for Skjodt’s R+ contacts, Michigan’s FBSO eff here is around .300.  Naturally this flips for Bates’ S- where -0.300 is Wisconsin’s “Point Score” efficiency on that reception to attack by Michigan. This naturally makes sense as the receiving team is typically more likely to win the point on the first possession than is the serving team.

You can glance around and see the logical output efficiences: A# is a kill by Gillis, B= is a Kieffer-Wright getting tooled, and Haggerty blows it trying to set the second ball and loses the point (E=). And it makes sense that these contacts have output eff’s of 1.00 or -1.00; the teams making these contacts will always win or always lose the point on the current possession when the touch is terminal. These are the simple ones.

The interesting ones are the non-terminal touches. Things like receptions, digs, sets, block touches, etc. What I’ve done is found the the Win Eff for each of these. With Jenna Lerg’s perfect freeball pass (F#) to start the possession, Michigan’s Win Eff is around 0.500 when attacking off a F#.. When Haggerty digs Mahlke with a D#, Wisconsin’s Win Eff off a D# is a little above 0.300.

For non-terminal attacks, I find the Output Contact. Possibilities are block touches, dig qualities, and covered balls back to the attacking team. So if we look at that very first attack by Mahlke, the output contact is the B- by Gillis. To calculate the output eff, we look at Michigan’s Win Eff when defending an opponent whose first touch is a B-, which is around 0.050.

Now is the cool part.

An inquisitive mind might argue that a perfect pass against Nebraska might be worth more than a perfect pass against Rutgers (sorry Rutgers). Or that getting a kill on a high ball after a poor pass against Rolfzen twins has more value than a high ball against Rutger’s block/defense. Moral of the story, we need to account for both sides of the equation.

We need to acknowledge that your opponent has an effect on your performance. Illinois may hit .200 on the season, but that number is expected to be higher against worse teams – and lower against the top contenders. To account for this, we average the two sides.

1. How does your team making the contact usually do in this situation against all teams

2. How well do all teams do against your opponent when all teams are in this situation

So if you usually hit .500 on quicks to your middle, but against Penn State teams usually only hit .300 on quicks to your middle – then your “expected Output Eff” would be 0.400.  We account for the contact team’s strengths/weaknesses – but also account for how your opponent affects different situations as well.

We do this based on the entire season’s worth of data so that these expectations aren’t biased by single match highs or lows. So that perfect freeball by Lerg is actually the average of how well Michigan does with a F# and how well all teams do with F# against Wisconsin.

I would argue that this is a better way to look at the value of each touch as it relates to your ability to win or lose the possession.

Service Error and FBSO

fbso-team

As a follow-up, here is a more advanced way to evaluate serving. FBSO is of course indicative of your opponent’s ability to win the point on the opening possession – but what many overlook is losing on the opening possession. What you see graphed along the y-axis is “FBSO efficiency”, meaning we subtract lost points on the 1st ball from those that were won – in the same way hitting efficiency is calculated rather than just Kill%. This is why some FBSO eff’s are negative. The labels you see are the serving teams.

Compared with Passer Rating, this viz shows the opposite relationship to service error%. While there are certainly team performances where opponent FBSO eff stays low even as team SE creeps towards 10%, the overarching trend suggests lower service error is associated with lower Opp. FBSO eff.

fbsoplayer

We see this trend again when drill down into player-specific data. Each circle here is a single match performance by the player and the resulting FBSO eff by their opponent only when receiving against the player. Players with fewer than 5 serves in the match were filtered out.

The idea of looking at the efficiency of your opponent winning the point on the first ball makes more sense than solely looking at passer rating – mainly because Nebraska probably hits better on a good pass than Rutgers does on a perfect pass. Therefore getting Nebraska to pass a 2.0 on a 4pt scale might mean they still win the point as often as Rutgers does while passing perfectly. So in that sense, if they both FBSO at .500, your “tougher” serving against Nebraska really hasn’t help you win points (and that’s what we care about).

**This general concept of the efficiency of winning the point is incredibly important moving forward. In this example, we’re looking at how the level of service error relates to your opponent winning or losing on the first possession. We could look at FBSO eff by reception quality – or transition eff by dig quality – or by type of set – or by where you attack from – or really anything!

I’ll refer to this moving forward as Win Efficiency, just meaning your (won-lost)/attempts with regard to a specific contact. But this is how we’re going to define value for each touch; how it affects your ability to win/lose the rally.

Service Error and Passer Ratings

passrateteam

How aggressively do we want to serve?

Tough, but not too aggressive? Keep it in and make your opponent work? Screw it and let it rip? Probably somewhere in the middle…? Serve tough until enough parents in the audience complain?? (Side note, listen to Becks talk serving if you never have)

The first step coaches usually take is to look at the Ace:Error ratio from the service line. Most will tell you they’re happy with 2 errors per ace, or something similar. Cool, but this doesn’t really provide a “why” – it’s just a way to glance at how “tough” you’re serving, but fails to explain anything of value.

The next logical step is what I’ve shown above. Service error percentage against your opponent’s passer rating. The idea here is to lower their pass rating while keeping your error percentage to a minimum. And to clarify, the teams labeled above are the serving teams in the particular match and their opponents are rated on a 4 point passing scale.

The above viz shows team performance in every individual match in the 2016 Big Ten season. And in general, the cloud of data slopes downward, indicating that the more service error you’re willing to risk, the more you can expect your opponent’s pass rating to drop.

But this leads us to the next step. Iowa, Ohio State, and Indiana have the 3 lowest single-match opponent passer ratings on the season. But are their respective 8, 11, and 27% error levels worth it?? The issue with looking at passer rating is that, by default, it ignores service errors – and so finding the appropriate tipping point becomes inherently difficult.

So what’s better than looking at Passer Rating as a means to measure the value of a serve? Service Error & FBSO, Service Error & Eff Change

p.s. I’ve attached a player-specific version of the same viz below. I’ve filtered out any players who served fewer than 5 balls in a given match, but the trend is the same.

passratingplayers