I recently watched a match where a coach whose team had been leading 18-11, took a timeout at the loss of the first break point: 18-13. Watching the match, I thought that was probably pulling the trigger too soon as at that point, essentially nothing had happened. Sure enough when his team then lost another three points, he was ‘forced’ to take his second timeout, while still leading 18-16. The team went on to lose the set 23-25, with the coach having perhaps handcuffed himself by taking the first timeout too soon. By coincidence, on the same day I watched a different match in which another coach also took both of his timeouts while his team was relatively comfortably leading. By coincidence his team also saw their entire lead evaporate and he had to stand on the sidelines as a spectator during the most important moments at the end of a set. This time however, the team managed to hang on and win despite having perhaps taken those timeouts too soon. I say ‘perhaps’ because the beauty of the game is that, despite what ‘experts’ may say, noone actually knows what would have happened if something had been done differently. It is all just conjecture. That is what makes it all so much fun. But I digress…
Timeouts are one of the few ways a volleyball coach can impact a game directly. The major reason that timeouts are taken is to change the momentum of the set**. Of course there are some who maintain there is no such thing as momentum in sport (for example) but presuming for a moment that there is, what is the effect of a timeout? One study of the German men’s league studied every match for an entire season. They found that a sideout was exactly 0.4% more likely after a timeout than at any other time. That is, taking a timeout had as close to no effect on the outcome of the next rally as makes no difference. So returning for a moment to our presumption that momentum exists, taking a timeout does nothing to change it.
My experience tells me that reality is not so simple. I believe there is such a thing as momentum. I believe that timeouts can affect momentum. The information from the study should make us think about how we use those timeouts. If the effect of a timeout is minimal then perhaps we would be best served by using them only at pivotal moments. For example, when the score is close at the end of a set and the smallest effect could be magnified. More than once I have found myself with a ‘spare’ timeout at the end of a set that I am sure (well, I would be, wouldn’t I?) was at least partially decisive. One time in a top match we were playing well, although down a couple of points and I didn’t think we needed one a change in momentum or tactics, even at 22-24. We managed to save the set points and I still had one timeout which I was able to use even later in the set and we were able to win. On another occasion, I judged that we were so far behind a timeout would make no difference to the outcome, so consciously decided to keep it to use right at the end of the set to try to disrupt the rhythm of the set with an eye on the following set. Without my help, my team managed to fight back and I still had that timeout to use at 22-22, when it was really useful. Again we won.
Of course, the other interpretation is that the timeouts really do have zero effect and I am only trying to convince myself that I made a difference when I didn’t. Either way, I think it is worth thinking about how we and why and when we use timeouts.
**The second and third reasons is some order are to give new information or tactics to the team and to be seen to be doing something.
did that study show any correlation between serve mistakes and timeouts?
The study was made directly from scoresheets, so they didn’t note how rallies were won and lost.
I would love to know the answer to that question as I have always felt like there are more errors (both serve and otherwise) after timeouts. My current feeling is that there is unlikely to be much of a correlation.
There is a correlation (btw, correlation is not the correct term).
One very obviously reason is: regression to the mean. If a green lighted jump server has an error % of 20% and he already made four serves, it is more likely that he will miss the next serve (under the same conditions). After a timeout (in most of the cases) the server already served a few balls, so it is more likely that he will miss the next serve (compared to the likelihood of his first serve).
That would explain another phenomena: In the first set a specific player has a high point scoring %, when he is serving. The coach changed the starting rotation (that is hopefully not the only reason 😉 ), so this player is going to serve first. Result: This player missed the serve. First obviously reason: Regression to the mean. BUT: I think, that is not the reason (or at least the only reason). I think that a timeout or the change of the starting rotation is changing the expectation, which has an impact on a player’s performance. The task of the coach is to find out, which player is reacting how on the change of expectation.
I think in general, that regression to the mean is an interesting concept. Applied to this context, regression to the mean suggests to me that after not siding out 2 or 3 times, you are ever more likely to sideout, without any intervention. I interpret that as being a vote against using timeouts at all.
What I especially like about regression to the mean is that if you follow it to its logical conclusion, if you have a spiker who scores on his first four attacks, you should immediately sub him off. And a spiker who has made three errors should absolutely be left on at all costs.
It is the right thought process. But what if taking a timeout even increases the chance to sideout, compared to the concept not taking one and hoping for regression to the mean. Also you never know when the regression to the mean will occur. The chances are higher from serve to serve but never 100%.
Regarding the attacker. Because attacking is more complex and has more factors, it is more complicated. But here a short answer: 1.) for the good attacker: yes, that could be right, depending on how he scored the attackes, but the biggest question is what is the mean? Let’s assume his personal hitting efficiency of the last 5 matches is 35%, but today his mean (form of the day) is 65%. Talking about regression to the mean… which one? Obviously you dont know the hitting efficiency of the match before the match is over. In this case, we developed a number/indicator which tells you if an athlete is over or under performing and gives you the probability that you are right.
2.) the bad hitter. again the theoretically the conclusion is right but i would more disagree and here are even more factors involved compared to the good hitter.
I didn’t say I agreed with that idea. For one thing, I am very confident that any professional coach who tried that would very quickly be out of a job, even if he won. It is so far out of the realms of the ‘normal’ he would be considered borderline insane.
Your points are perfectly valid.
For the record, I am not entirely sure I ‘buy’ the ‘no such thing as a hot hand’ theory.
oh yeah. totally, looking for a coach who has the balls. going to watch your match in january. Let’s see what you got. 😉
I think there is a ‘hot hand’ In my opinion the basketball study didnt take the importance of points into account. There are some studies in tennis which used a nice concept to evaluate the importance of point. The point which “matter”! they correlated the performance of athletes which fan engagement and other things to find out the impact of home court advantage. Same in volleyball. I belief you could support the theory of the hot hand when you look at the important point.
In an important match (unfortunaltely, I cannot remember which one, might have been in last year’s Champions League) I saw Vladimir Alekno taking two timeouts immediately one after another, i.e. without a single ball being played in between. Alekno coached the away-team (or for some other reason, the spectators were not on his team’s side for the most part) and the set was really close. The crowed kept cheering throughout the first timeout, but when the coach took the second, suddenly the gymnasium went completely quiet. It’s just one instance, of course. However, I myself took it as a rememberable lesson, that indeed there is such a thing as momentum in volleyball, and that the coach has not many, yet some means to influence that momentum. Almost needless to say, that the lesson became even more impressive to me due to the fact that Aleknos team won that very set.
Anton, I totally agree that whatever explains the typical outcomes of rallies after timeouts (assuming for a moment that there really is a phenomenon to be explained), that explanation cannot be due to statistical propabilities for the most part. The effect you mention -regression to the mean – , cannot be part such an explanation at all. For as far as I understand what you’re saying, your argument seems to be a case of the so called gambler’s fallacy. Classically it goes like this: On a roullette-table there are as many black fields as there are red fields. Therefore (regression to the mean) the total number of cases, where the ball lands on red, equals the number of cases, in which it lands on black in the long run. Now, suppose you have observed four reds in a row (or four succesful services), what is the probability that the ball will come to rest on a red field (that the next serve will be succesful) in the next round? It’s the same propability as in every round before, because the treis are independent, i.e. the roullete-table has no memory of it’s past outcomes! Of course, a volleyball-player does have a (literal) memory of previous tries, but if this memory does have an effect at all, I presume it would point into the other direction: A player, who remembers succesful tries for the next try, will be more succesful. That’s what I take to be meant by expressions like “he went hot” or “she plays in the zone”.
I agree with Berti that none of the events we are talking about are truly independent events. But it should be noted that there have been many studies now on the ‘hot hand’ that ‘prove’ that it doesn’t exist. The most famous of those studies is this one…
Click to access gilovich%20%281985%29%20the%20hot%20hand%20in%20basketball.%20on%20the%20misperception%20of%20random%20sequences.pdf
And great Alekno story…
very nice study. But isn’t the argument that the events are not independent an argument against the gambler’s fallacy theory…
I think the argument that events are not independent is an argument against a lot of statistics in sport 🙂
There are a lot of nice statistiks which can help you to predict dependent events…. 😉
now, this gets even more interesting…
why should the likelyhood for a player to make a mistake increase after 1, 2 or 3 good serves? is he taking more risk?
anyway i would think that rythm and flow are two important expressions here. that’s why experience teams start mopping the floor everywere after one or two good serves of the opponent.
i could also imagine that some of the pressure on players serving after timeouts is created by coaches by overemphasizing “make no mistake after the timeout”, which I manytimes hear, and don’t like at all.
I think all of these things are factors. As Anton mentions, the expectations of what should happen after a timeout affect many things. As does the break in concentration, the change in psychological stress and any number of other things.
Berti, thanks for your input.
I am totally aware of the gambler’s fallacy. If i toss a coin the chance is 50/50, no matter if i just tossed 4 heads in a row, BUT the chance to toss 5 heads in a row is 3% (to toss a head before the 5th toss is 50/50). If I toss a coin infinite, the proportion of head and tail approaches 50/50, but the difference of head and tail doesnt systematically decrease to 0. (This is what gambler don’t know). And i guess every person would still bet on head after i tossed 5 times in a row a tail.
Back to serving: Let’s say a server has an error% of 20%. The chance that he makes 5 serves in a row is 32.7%, The chance that he makes serve number 5 after he already made 4, is still 80%. (@Martin, hope this answered your question)
I did that in practice: All player had to serve 10 serves in practice over 8 weeks. (Always at the same point in practice). I simply scouted if they made the serve or not (i added a target and pace of the serve, but not important for the discussion). The average error% for the serve number 1,2,3…,10 is almost the same, but if you isolate the cases where a player made 3 in a row and look at the error% of his 4th serve, it looks different. (I added artificial a “memory”). You can use similar mathematical model to analyze a players form of the day or at what point (sample size) you can trust your ingame scouting/numbers.
As we all agreed, it is not the only explanation. I still think, that expectations have an impact. Not only that, I could show that the results are different along personality types.
It is certainly an interesting point about expectation. This is one of the first things I learnt. Whenever I tried to manipulate my starting rotation to take advantage / hide strengths and weaknesses, what always happened is that my strong / weak rotations changed. There are lots of different contextual things going on that are not obvious.
I truly believe that expectations are playing a big role. We need to teach our athletes to deal with different expectations.
The biggest problem with manipulating the starting rotation after the first set is, that the sample size is to small. By chance alone it is likely that your normally weakest rotation is the strongest after the first set and the other way around. A lot of people have a tendency to generalize small sample sizes. Here you can also run an analysis to figure out how likely it is if your team is over or under performing in a certain rotation.
In my opinion it is even more important to look for the right matchup against a specific opponent.
In everything we do in volleyball we come up with the problem of small sample sizes, because a match has so few actions of different types.
I go with matchups, taking into account my own strengths and weaknesses.
yeah, pregame scouting with a big sample size is important.
I’ll save an answer for that until another time 😉
interstingly enough are is the fact, that a team that is leading 8-x or 16-x is very likely to win the set. so the question is, if it is not important to make sure you keep a lead once you got it. from that perspective a timeout at 18-13 seems alright.
however, after all this discussion about numbers, expectations and propabilities, I, as a coach need to decide whether to take a timeout or not. Knowing that of course teams are different, I would still be interested what are the factors you are using as a basis for this decision?
I for myself tried to start a list of a couple of factors here:
1.) a certain feeling for momentum (this is not really a hard fact, but for me very important. many times I think this is related to the body language of my players on the court)
2.) there are things we do tactically that I want to change
3.) There is a very strong server on the other side, who I know is likely to make a mistake after an interruption (I try to remeber this /take notes about this)
please add, I would be interested!
You are right, the point in the end is to help your team be successful. I think point 2 and 3 are very strong reasons to take timeouts. When I was starting in professional volleyball I never took timeouts for reason 3 thinking that at this level it is a waste of time. I was wrong. It works at every level.
I guess point #1 is the main point of the post and the main reason we all take timeouts. We are all looking for ways to judge momentum. I am sure the statistics can help us in some ways, but ‘momentum’ is such a weird thing it is difficult to define it, let alone come up with any rules to use for it.
I read the Jonah Lehrer book ‘How We Decide’ which explains that different types of decisions are better reached with cognitive reasoning and others are better reached by emotional reasoning. I suspect that timeouts are probably better coming from an emotional level. Since then I have tried to be open to my emotions during a game and to take timeouts when I ‘feel’ it. I think this has been helpful for me.
The reasons for taking timeouts listed by Martin are valid with respect to the aim of maximizing chances to win a particular match. And yes, there are some such matches, where wining is everything.
However, I think that this is not a very lasting perspective and that it is us coaches, who have the responsibility of thinking beyond the mere result of the match in front of us. There are might be other aims connected with playing a certain match: preparation for another -more important -match, developement of ceritain players, tactics, the team (ok, very broadly described, feel free to narrow), mastering other challenges than reaching 25 points before the opponent does in general.
With these other-than-winning-this-match-aims in mind, I think there might be a bunch of different reasons for taking timeouts, e.g.:
– breaking the momentum of your own team (for developemental reasons)
– unsettle the referee by taking timeout in the very last moment or the “third” timeout (not very sportsmen-like behaviour, therefore not something I usually consider, but I have seen these things in minor leagues quite often), alongside with influencing the crowd, see the Alekno-story: maybe belongs to the point of winning matches
– give new challenges to the team against overwhelmingly strong /weak teams
– improve your own coaching abilities
– Ask players for their perception of opponent, teamspirit, general situation
Another, quite general thought: These days one reads very often, that good, old “Never change a winning team” is not valid, since it implies that you either never change your team at all or that you already ARE in trouble, when you start considering changes, which is obviously too late and unneccesary. So I think the same applies to timeouts: You better take them before you are in trouble. Actually, it might have been this consideration that was behind the moves of the coaches you descriped in your original blog entry, Mark. Of course, then the question becomes: How can I possibly know that I’m going to be in trouble before I’m a actually in trouble? Here indeed matchups, scouting and so on might do a valuable job.
PS Glad to see that we’re all in the same boat regarding the contents of statistics, Anton. Thanks for clarifying your thought to me!