Or Why Per Set Statistics are Bullsh*t

Pop quiz.

Question 1: How many points is a set in NCAA Division 1 Volleyball?

If you said twenty-five, you are wrong. If you said twenty-five except in a fifth set when it is only fifteen, you are wrong.  If you said it depends, then congratulations you have won the game.

Question 2: Name each team that corresponds to the primary or secondary color referenced in the following table.

Hint: These figures are the average points per set for each Big Ten team last season.

Answers are at the bottom.

This presents a significant issue when it comes to doing rate statistics.   A four kill per set player at 41.12 points per set would average more than 4.3 kps if they had the same kill rate and played for the last team in the table averaging 44.29 points per set.  A quick look at the current KPS table on NCAA Stats says that’s the difference between being 57th and 34th.  A significant difference.  Additionally, because the gap will widen linearly as the initial numbers grow, the result is that the players at the very top of the chart can be misrepresented to the highest degree.  That’s simply unacceptable when trying to use a statistic to formulate any significant statistical argument.

And that’s not even the worst example I could come up with.

Per Set statistics are meaningless without additional context.  The context that a person would need to supply to make those statistics worthwhile is tedious and time consuming to track and calculate.   This results in volleyball fans and the media continually relying on and relaying statistics that in reality mean very little.

It needs to change, and that’s the purpose of this article and a couple of pages that I’ve now added to the site.  While per set statistics are exceptionally flawed, points are not.  In fact, points applied in the right way can be exceptionally accurate when calculating a rate statistic.

Here are the top ten players on the current(9/15/2018) stats.ncaa.org KPS leaders table.

Here is the same list, but instead of per set, the rates are per point in each team’s games.

Kills Per 100 Points

There are differences and there can be extreme differences. In fact, with many of the common statistics volleyball fans and the media use, using points is inherently flawed.  When measuring kill rates for players, total points in a match has an inherent flaw.  Specifically, during any match there will inevitably be points wher a player who plays for all rotations will still have no chance to get a kill, namely aces and service errors.

AN: There are caveats beyond this as well. Not every player plays six rotations and there are points beyond aces and service errors during which a player would also by definition not have a chance to get a kill. Rotational errors are the specific issue in this case because I did not have the foresight to track them efficiently in my database and will need to do some significant redesign before I can efficiently account for those points.

Beyond using total points for a rate, simple subtraction can be used measure the rate at which a player gets a kill for every point in play.

This is the table for the statistic named Kills per Point-In-Play.  Original name, I know.

Kills Per 100 Points In Play

It’s not even the same list as the original per point list.  While this approach is not currently perfect, it’s still much better than using per set because the denominator of the statistic being calculated means the same thing across all players.  The context per point statistics provide is important because without it volleyball statistics are very nearly meaningless.  An instance of “Don’t let the perfect be the enemy of the good.” if you will.  Making even incremental progress toward better understanding and knowledge is important.

Measuring virtually every statistic by points played improves it drastically, but Aces and Service Errors per set make the least sense as a statistic.  It is entirely possible to play a set in which a primary server will not serve during the set.  In fact, in fifth sets, it actually isn”t all that rare.  Luckily, NCAA Play-By-Play pages happen to track exactly who serves each point, a fortunate thing in this instance because those pages can be used to get per serve rates for aces, service errors, and service points.

AN: This(A primary server not getting the opportunity to serve.) actually happened to Lauren Stivrins in Set 3 of the Nebraska vs Missouri State less than three hours after I saved my latest draft of this article. Yeah, that happened.

These statistics and more have been made available on each team page as well as Division and Conference leaders.  The leaders are available using the “View Complete Advanced Statistics Leaders” link on each division or conference page.

And that friends, is why per set statistics are bullsh*t.

Answer Key

Top to Bottom: Rutgers, Penn State, Nebraska, Wisconsin, Indiana, Illinois, Northwestern, Iowa, Maryland, Minnesota, Purdue, Michigan State, Ohio State, Michigan

The Preface

We have data now, but we have no context.  That makes a lot of organizing extremely difficult at best and a fool’s errand at worst.  So, I’ve given a lot of thought to some of these teams but very little to others.  Alas, sometimes you just have to throw things at the board and see what sticks.

The List

1. Stanford
2. Wisconsin
3. Minnesota
4. Texas
5. BYU
6. Florida
7. USC
8. Russ Rose
9. Baylor
10. Nebraska
11. Illinois
12. UNI
13. Creighton
14. Washington
15. Pittsburgh
16. Kentucky
17. Oregon
18. Washington State
19. San Diego
20. Florida State
21. Louisville
22. Michigan
23. Colorado State
24. Purdue
25. Marquette

The Justification

I don’t have much today.  There are simply no connections between the pods in our directional graph.  Some teams did stuff that I thought probably deserved some movement and got moved.  Some teams did very little to prove or disprove my initial assumptions and stayed put. I was in Lincoln at VERT(Longer article to come.) and didn’t see much beyond the four teams that were there.  I spent most of the rest of the weekend traveling to and from Lincoln. The official post on VolleyTalk courtesy of user vbprisoner can be found here.  The AVCA’s vastly inferior poll can be found here.

Nebraska’s Place in the Fray

Tough to tell right now.  The first match for Big Red was pretty different than the clean volleyball we’ve come to expect from the Huskers.  However, with all the new players and the losses sustained by graduation and transfers, it wasn’t a totally unexpected start.  Statistically, it wasn’t even that different from Nebraska’s appearance in the VERT Challenge in 2017.  Obviously, I’m not expecting this team to win a National Championship.  That is an unfair expectation at any time, and doubly unfair when half of the team is new to the program.  It does; however, illustrate the incredible transition a team can make from opening day to tournament time.

That’s why they play the games.  That’s why we watch the games.

The second week of volleyball season is in the books and this is who lit up the ViPR Win Probability Added sheet.

All-Around Division 1

Heather Hook of UNI takes the crown for the highest base score for WPA.  She scored an impressive 2.2865 in UNI’s four set victory over USC.  Getting the job done all over the court with 50 assists, 8 kills on 16 swings, 10 digs, 4 block assists and a service ace.  I suspect a partridge in a pear tree may have been watching that performance.  It was a performance which also takes the award for the highest accumulated score per set.

→ Southern California vs. UNI (2017-09-01)

Specialists of Division 1

The highest per set attack score was produced by Green Bay’s Lydia DeWeese.  Scoring eighteen true kills on thirty-one attempts with only two hitting errors in a competitive three set loss to Butler.  All three sets were won with two point margins and the third set scored 56 total points.

→ Butler vs. Green Bay (2017-09-02)

Minnesota’s own Samantha Seliger Swenson takes home the award for the highest setter score.  Samantha captured the award by slightly more than a thousandth per set, just edging Duquesne’s Dani Suiter.  A comfortable three set sweep of Tennessee provided the stat line for Gopher setter.

Game Not Available

Swenson accumulated forty-nine assists in three sets during a performance game that saw Minnesota hit 0.396.  More than a solid day at the office, but Swenson also added three kills, two blocks, and seven digs.

We come to the most fickle of skills the block.  The domain of those later described as having a sense of the moment, or a penchant for delivering in the clutch.  Lauren Frilling of Xavier takes home the prize this week.   Her eight block performance scored 0.4963 on the stat sheet.  Two solo blocks and four block assists in a four set loss to Miami of Ohio.

→ Miami (OH) vs. Xavier (2017-08-29)

Aces and service points rule the serving score and Ivana Blazevic of Maryland Eastern Shore certainly acquired plenty of both.  During Eastern Shore’s three set sweep of St. Francis Brooklyn, Blazevic served twenty-seven times and her team scored on twenty-two of them.  Eight times the ball went over and did not come back.  Ivana added thirty-six assists to her serving exhibition.

Game Not Available

Emily Lopes of CSU Bakersfield accumulated thirty-two digs in a four set match against Valparaiso.  Averaging eight digs per set and doubling the total of the next person on the team placed Emily as the top performer in the Roadrunners victory.

→ CSU Bakersfield vs. Valparaiso (2017-09-01)

That’s it for the week’s Top Performances.  Congratulations to all of the standouts.

Notable players of the week are a part of sports.  The people who did the most or the strangest things.  Well, the strangest might be hard to see in statistics, but the most we can find.  So, the question of today’s post is, “Who had the highest ratings by ViPR this week?”  A question we can most definitely answer.

All-Around Division 1

The player with the most eye-popping line of WPA for the week is Lindsey Ruddins from UC Santa Barbara.  I don’t even want to spoil it before putting the game summary.

→ Florida St. vs. UC Santa Barbara (2017-08-26)

2.8903!

The stat line is no less impressive.  Thirty-kills on eighty-three swings with only eight hitting errors and not a single kill came by way of an opposition block error.  In addition, Lindsey put up 18 digs which accounted for almost a point of her score.  That’s a day at any office.

That’s impressive, but it was also a five set match.  How about the person who scored the highest WPA per set in a match.  That award goes to Indiana State’s Laura Gross in their three set win over the Big East’s DePaul.

→ DePaul vs. Indiana St. (2017-08-25)

A three set WPA score of 1.8990 translates to 0.633 added for each set played.  The line is again impressive with nineteen kills on thirty-five attacks and five errors with eighteen digs added for good measure.  A stellar night to be sure.

Those are the all-around players this week, but let’s see some specialists.

Specialists of Division 1

Amanda Carroll of Florida Gulf Coast takes home the prize for the highest attack score per set with a twenty-three kill performance.  She took forty-one swings and committed only two hitting errors.  Take a look at the game summary.

→ FGCU vs. IUPUI (2017-08-26)

No, your eyes do not deceive you.  Amanda scored more than a point higher than anyone else in the match.  Her attack score of 1.4179 was more than double the next player’s attack score.  A dominating night for the lady from FGCU.

Brooke Short of Louisiana Tech takes home the award for the highest setting score per set in a four set loss to North Dakota.  Despite the loss, Brooke put up forty-seven assists in four sets and added twenty digs, six kills, and an ace just for good measure.

→ Louisiana Tech vs. North Dakota (2017-08-26)

In terms of Win Probability Added, blocking is a notoriously fickle skill.  A player may have ten blocks in a game, but if those blocks don’t come in high leverage situations, the players block score won’t pop off the page.  So whose block score does pop off the page?  That would be Marshall’s own Addisyn Rowe.  It’s extremely rare for a block score to be the impetus of a player making the Top Performances list, but take a gander at the summary for UMKC’s sweep of Marshall.

→ Marshall vs. UMKC (2017-08-26)

Addisyn converted three solo blocks and two block assists into 0.4819 of Win Probability Added.  That’s spectacular.  Timely blocks make huge changes to the game.

Serving is another skill that is both difficult to grade and fickle besides.  Aces are easy to grade but like blocks must be timely as well.  What about consistently good serves.  Long serving runs don’t usually score that well either because they put the game out of reach and after the first couple of points usually don’t score that highly for the server.  That means timely aces and consistent serving are the only two ways to really hit it big on the service line.  Well, long runs may not always score that well, but they certainly help.

→ Penn St. vs. UT Martin (2017-08-25)

Penn State’s own Bryanna Weiskircher scored twenty-four Service Points on twenty-eight Serves against an overmatched Tennessee-Martin team.  Each time she put the ball in the air, there was an 85.7% chance it would land on UT Martin’s side of the court.

How about those pesky back row players? How made the other team cry out in frustration by always getting there when an attack went anywhere near them.  The award for the highest digging score was just too good to spoil.

→ Chicago St. vs. Evansville (2017-08-25)

No, that isn’t a typo.  At least I don’t think it is.  Forty digs.  In a four set match.  Forty.  Lauryn Cruz of Chicago State takes home the prize for the most bonkers stat line this weekend in a four set loss to Evansville. Add to that eighteen kills on fifty-eight swings.  Not all was well for Lauryn with ten hitting errors, a service error, and a reception error but forty digs were enough to put Lauryn on the list this week.

As a fan of a blue blood, I feel compelled to mention that digs tend to favor teams that aren’t facing elite hitters, but that’s a token argument at best.

Well done to all of this weekends standouts.

AN: Exact grades may change from the time of writing.  Each game is analyzed every time new ratings are calculated.