Volleyball Analytics: Some Ideas

I’ve always been interested in sports statistics.  Collecting and collating cricket statistics is literally amongst my first memories and was very definitely my first hobby.  My interest in sporting numbers has maintained and over time I have also developed an interest in the even more interesting how’s and why’s behind the numbers.  There are lots of places one can investigate the how’s and why’s of sporting numbers.  Books like Scorecasting, Basketball on Paper and Wages of Wins are really interesting to follow as an intellectual exercise even if I don’t always understand completely (or care about) the details of the different sports.  On the internet, Wages of Wins has a blog, and writers like Bill Barnwell and Zach Low at Grantland.com write about the NFL and NBA respectively using a heavily analytics based approach.

Volleyball, to my knowledge at least, lags a long, long way behind other sports.  One volleyball analytics blog, post semiregularly but the focus is almost solely on spiking efficiency in US college volleyball.  Which brings me to the point of volleyball statistics themselves.  So many of them are superficial or meaningless or subjective or some combination of the those.  For example, spiking percentage (kills as a percentage of attempts) provides some useful information as scoring points is the principle objective of the game.  Likewise, spiking error percentage (errors as a percentage of attempts) provides information on points given up by the spiker.  Spiking efficiency ((kills – errors) / attempts) on the other hand, simplifies those two numbers into one less meaningful one that overvalues minimisation of errors.  None of these spiking numbers takes into account the up to 50% of spike attempts that do not end the rally.  Blocking statistics are even more useless.  The standard measure of blocking is blocks per set, but this figure doesn’t take into account number of attempts, number of opportunities or any number of other things.  And so we can go through all the areas of volleyball.

Partly ‘inspired’ by reading the listed books and blogs and articles, I’ve tried to find different kinds of numbers that could lead to different (new?, better?) ways of understanding the game.  Using Data Volley statistical software I can manipulate numbers over multiple games in a lot of different directions.  A couple of things that I’ve come up with are intriguing, at least to me.

The basic area that I’ve been looking at is how particular actions impact on the likelihood of winning a point.  For example, our percentage of winning the rally when we receive (sideout %) is around 69%.  However, when we have perfect or positive reception (not perfect but all options available) that percentage rises to 76%.  That isn’t our attack percentage after reception, but the percentage of all winning rallies in all ways.  This is all logical and expected.  However when I started to look at free balls (a ball crossing coming across the net without an attack) I came across something that I thought was odd.  After perfect or positive free ball reception, our rally winning percentage is only 63%.  Conventional wisdom says that a free ball is the easiest situation in volleyball and should lead to a very high percentage of points.  But the reality is that, controlling for the quality of the first contact, we are significantly less successful than after service reception.  I’m not completely sure why that should be the case.  For the purposes of comparison, our opponents are also better after reception but not by as much (68%-63%).

On a similar note, I’ve been trying to thing of ways to include those 30-50% of attacks that aren’t accounted for in ‘traditional’ attacking statistics.  Using a similar approach, I came up with a likelihood of winning a rally if a player attacks during that rally.  For outside hitters the range is between 54% and 71%, while the middles are all between 73% and 77%.  It might be interesting to note that the range of this number is much smaller than the range of spiking efficiency (54-77 versus 12-50).  But then again, it might not.  I’m not really sure.  It seems to tell a different story about spiking, but I’m not sure if it’s a better story.

I’d be interested in your thoughts.

Advertisements

17 thoughts on “Volleyball Analytics: Some Ideas

  1. jones

    maybe 2 reasons:

    1. the team is less organized while receiving free balls (in most cases)
    2. the psychological pressure is a lot higher, because everyone expects the team to score

    Like

    Reply
  2. Alexis Lebedew

    My first thoughts, regarding the freeballs are:

    1 – regardless of the fact that it is an ‘easy’ ball, it generally occurs out of system. Compare this to service reception, which is always completely in system. Everyone knows exactly what their role is (particularly on a positive or better pass). With freeballs there requires transition from a less than comfortable location (perhaps), and generally a different communication system, or, at least, a communication system which is hampered by the fact that the ball is already in the air (compared to siding out).

    2 – we actually practice siding out a lot more than freeball offence. Which makes a lot of sense as it happens much more often.

    Like

    Reply
    1. markleb Post author

      If you consider that most wash drills have free balls as the extra balls, and how many actual free balls there are in a match (not much more than two per set in my league), free balls are the single game situation that we practice the most.
      Except that the way we practice is completely different from the way that they occur during a match.

      Like

      Reply
  3. Patrick Aparicio

    Personally, I’m not sure if a team should be less organized in a free ball situation. Liberos taking large responsibility should theoretically free up spikers to take proper approaches?
    Could it be that the majority of free ball situations happen in the few long rallys that we have in men’s volleyball…spikers are a bit more tired, resulting in errors or losing the edge versus block/defense?

    Like

    Reply
    1. jones

      Even when the libero receives the free ball, I think that organization can be one reason for not being as successful as after service reception, because I guess that many free balls are rebounds from the block, i.e. there is very few time to reorganize.

      Like

      Reply
      1. markleb Post author

        In this context I am only including actual free balls, ie balls dug or set over by the opponent.

        Like

  4. Hugh Nguyen

    Great post Mark! I think we underestimate the randomness of free balls. I remember you mentioning at the FIVB Setting Seminar that receiving the free ball can be quite different whether the opponent plays it from in front or behind the attack line. Do we make free-balls realistic in practice drills that reflect the variability of a match? Receiving serve (while I’m not convinced is easier) is more predictable in the sense that how to deal with serves from all the different angles, scouting etc means we’re certainly more prepared for them.

    Like

    Reply
  5. royvball

    I’m not really sure, but I think about these things:

    1) Transition – Like others above, my first inclination is to think that the transitioning phase may be part of the reason to blame.

    2) Hitter selection – Are you able to see in the matches you analyzed… if the setter tends to set outsides more in a free ball situation? If so, it may bring free-ball win percentages lower, based on your stats of whether outsides or middles hit in a rally (I know it sounds counter-intuitive, but hell, free-ball win percentages being lower are also counter-intuitive.)

    3) Blocking – Do free ball situations allow opposing blockers/defenders to be more ready (slightly more balanced, patient and reading) to defend the attack? For example, are opponents’ blocking stats generally more positive when they give a free-ball vs. service reception?

    Like

    Reply
    1. royvball

      More thoughts:

      4) Setter position – What happens to your stats when you break it down for positive free-balls when the setter is in front-row vs back-row? Is there a noticeable difference? If so, might there be a coincidence that most of the free-balls you analyzed consist of one more than the other?

      Like

      Reply
  6. gonzo

    Do your stat on freeball take into consideration the opposition? I would imagine a free ball action would be ran potentially less fast than a normal reception, leading to an action that is more “readable” by the opponent? in general i think one good angle would be not only to focus on the team stats as if they were playin gon their own but to stats in relation to how the opposite team is reacting. for instance in the example i was mentioning one tell-tale would be to look how those freeballs points were lost. if you see more blocked point then maybe it is a problem of the other side reading your game.

    Like

    Reply
  7. Pingback: Free Balls : The ‘Answer’ « At Home On The Court

    1. Dave Smythe

      Maybe the conversion stats differ because of he nature of the action, which for the attackers, they have a moment of relief, space and time, whilst the defenders go directly into must save mode at all costs.

      Like

      Reply
  8. Martin

    Although this article has been posted a while ago, I’m right now also trying to find answers to questions like attack efficiency. What do you think about the much used passing efficiency (0 for an ace, 1 for a poor pass, 2 for a good one, and 3 for a perfect one)? Interestingly a similar number is not used in attacking.
    In the end the problem is always the same: we have a lot of numbers, that reflect the reality better, or we want to simplify things by putting it all into one number, loosing detail.
    Any news which numbers you are using now?

    Like

    Reply
  9. Chris

    Here are my suggestions for attack statistics:
    1. Percentage of points won from an attack within the next 3/4/5… touches of the ball.
    2. Average dig quality of the opponent
    3. Break down attack statistics with respect to the block/defense situation: attack efficiency (or any other attack-related statistics) against single/double/triple block
    4. Break down those stats with respect to set-quality

    Like

    Reply

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s