Volleyball Coaching Life-Hard Limits on the Number of Errors

My friend Jim Dietz’ blog post (Dietz 2022) on January 16, prompted the thinking behind this blog post. I had been thinking about this question of putting limits, both maximum and minimum limits on errors ever since the end of the Olympics in 2021, but I needed to let the thoughts to age and sort itself out. In addition, a friend, who coaches in Division 1, posed the challenge to me: what is the maximum amount of service error? Jim’s blog post on approaching the minimum of overall error made me think about it.

I approached the idea from a point scoring standpoint. The initial explanation is basic and putting the discussion and assumptions on common ground, so please be patient.

Volleyball scoring is not very complex. The idea is to score 25 points before our opponent does, with rally scoring; that is, there is a point scored with every ball whether the team scoring the point is serving or receiving serve. The point scoring for the good guys is split between points the good guys score on our own, and the points the bad guys give us as errors they make.

The points earned by a team are:

·       Ace serves

·       Kills

·       Stuff blocks.

The points given to the opponent by a team are errors:

·       Service errors

·       Blocking errors

·       Attack errors

·       Digging errors

·       Passing errors

·       Referee sanctions

A great way to track these numbers is the trendline graphical representation of the point scoring invented by Dan Mickle. https://thecoachesmind.com/trend-line-stats/

The scoring for the entire set boils down to:

·       How we are scoring,

·       How our opponent is scoring,

·       How many errors we make, and

·       How many errors our opponent makes.

There are uncertainties and randomness that are associated with each of the ways the points are scored. We do not know anything about the uncertainties and randomness prior to the set because if we did, they would not be uncertain and random. These uncertainties and randomness are what makes the game interesting and what differentiates playing the game with humans from a computer simulation.

Nassim Nicholas Taleb’s book Antifragile (Taleb 2012), explores a stochastic reality; stochastic being defined as: involving or containing a random variable or process. His idea is any system behavior are rarely certain or deterministic, most man-made system are subject to uncertain and random variations. The book deals with  systems in general, but its focus is on the economic system  because the author was a stock trader. 

In Chapter 7 of the book, titled Naïve Intervention, he explains the idea of iatrogenics, which means “caused by the healer”. The idea is the balancing of the benefits versus the losses caused by a cure: does the benefit of the promised cure outweigh the losses caused by the cure, either intended or unintended?  It speaks directly to the Hippocratic oath: Do No Harm. Iatrogenics is the situation where more harm than good comes from an action, or that more losses than gains result.

When we humans interact with a system, either simple or complex, our urge is to correct those things that don’t immediately conform to what we think is normal. We tweak and  guide the system to a stable result for that moment. This is because we believe in the idea that the aggregation of stable subsystem reactions to a small disorder will necessarily guarantee a stable complete system reaction from larger disorders. Taleb’s contention is that in aggregate, it is necessary for the small results to be undamped, minimally controlled, and random, because minimally controlled, and random instabilities in a small scale will benefit the large-scale complex system by attaining long term stability. Complex system can withstand and survive the undesirable small perturbations to the system, the larger complex system is better able to naturally damp out temporary instabilities in the system; in addition, the larger complex system is better able to adjust and learn from the small perturbations so that when a larger perturbation happens to the larger, more complex system, it is able to withstand that larger perturbation because it had absorbed and adjusted to the smaller perturbations. Indeed, this is the idea behind antifragility. An antifragile system gains from disorder rather than just survive.

This is not to say we should adapt a completely hands off approach: the goal is not to let the system act and react open loop in a laissez faire fashion. The difference is that it is necessary to have a strong understanding of the system dynamics and behavior so that when we do intervene, we can knowledgeably do so without causing harm. What is needed is a robust model of how the system behaves and have some predictive ability before intervening.

This is where the word naïve comes in. Intervening without a model which has consistently demonstrated controllability, observability, and some order of predictability will create a rash of unintended consequences. This would be naïve intervention.

I applied Taleb’s ideas to the case of volleyball scoring by using the scoring modes I had defined earlier.   

·       How we are scoring,

·       How our opponent is scoring,

·       How many errors we make, and

·       How many errors our opponent makes.

This is our system. There are no other ways of scoring as far as I know. The four modes of scoring are interrelated because a change in one affects the other.

Increased errors from our team means that the pressure on the other team’s earned scoring mode is decreased, which results in our team needing to score more in a timelier manner, or our team needing to put more pressure on the opponents to make them make more errors. Every point shifts the emphasis on each of the four modes. The game flow is affected with every point, and the impact of each of the four scoring mode shifts with every point. Of course, digging into that kind of granularity is an overkill, but most coaches have an internal accounting of these four modes from point to point, if not numerically then instinctively.

Looking at the components of the four scoring modes:

·       Ace serves

·       Kills

·       Stuff blocks.

·       Service errors

·       Blocking errors

·       Kill errors

·       Digging errors

·       Passing errors

·       Referee sanctions

We see that none of these factors are sure things, as we can see in the descriptive statistics teams take during sets and matches. The numbers never duplicate itself completely and absolutely, there are always variations; there are and always will be uncertainties and randomness built into all of the scoring components, which are aggregated into the four point scoring modes. Another key issue is that there is an element of time involved: we need to get to 25 points before the other team can get to 25.

As coaches, we are always asking our players to minimize our errors and maximize our opponent’s errors, whether it is overall errors, categorical errors like service errors, or individual errors. It is natural and logical to do this. But, by stating that request generically does not affect the uncertainties or randomness inherent in the scoring modes.

It is however, a small jump logically to go from speaking in general terms — minimize or maximize,  to speaking in specific terms — only make 8 errors in a set or make X number of service errors. This subtle shift is to make the scoring components that were allowed to be variable to become deterministic. In the real world, randomness is distributed throughout  all the factors which make up the results, in this case, the randomness is distributed to four scoring modes. But, if we made one of the four modes non-random by setting it to a specific number, the randomness that was distributed to four modes are now concentrated on three modes. In addition, the causal relationship between the four scoring modes plays a role in making the concentrated randomness impact each of the random scoring modes more significantly.

For example: if a team is told to limit their total errors to 9 errors: service, blocking, kill, digging and passing. If a team can accomplish this by will, the emphasis is shifted to the opponent scoring. If the randomness is a positive one for the opponent, if a team that was not able to generate an offense which is capable of scoring 16 points on their own went on a streak where they were scoring well, each of those points score affect the final outcome more impactfully than if a hard limit was not placed on the errors committed. In this case, a positive variation for the opponent translates to us either hoping and forcing the opponents to make more mistakes and/or hoping that we can score more to overcome the opponent’s positive variation.

Another way to think about this is to recognize that there is randomness with our errors but that the allowable error limit includes those random events that can happen. The question is then: what is a realistic error limit to set which includes the randomness that cannot be predicted?

The opposite is also true. If we wanted our servers to be more aggressive in serving the ball, we can tell the servers to have a maximum number of service errors in mind as they served. This deterministically shifts the burden from our service errors to all the other errors: blocking, kill, digging and passing. It also shifts the burden to keeping the opponent from scoring more, i.e. playing better defense; to scoring more than our opponents on a play by play basis, i.e. we need to stay in synch with our opponent’s scoring; and to force our opponents to make more errors. These are all metrics that any beginning coach can recite, but the impact of each three options become more important because we have conceded a set allotment of error points.

This is not to say that coaches won’t be tempted to set maximum and minimum performance goals for their teams. The question to ask is whether we have a predictable error model for each individual player and for the individuals interacting as an aggregate to confidently know that we are NOT naively intervening in the set? Do we know whether our intervention will result in unintended consequences which could be a case of iatrogenic? Do we have a way to ameliorate the situation if the variations work against us?

Works Cited

Dietz, Jim. "Error #9?: The breakpoint of a set." Good, Bad, I'm the Guy with the Blog. January 16, 2022. https://thinkingbeyondthebox2018.wordpress.com/2022/01/16/error-9-the-breakpoint-of-a-set/ (accessed January 17, 2022).

Taleb, Nassim Nicholas. Antifragile. New York City: Random House, 2012.

 

 

Volleyball Coaching Life-Counting

 

I'm counting out time,
Got the whole thing down by numbers.
All those numbers!
Give me guidance!
O Lord I need that now.

“Counting Out Time”—Genesis

Counting is an instinct for humans. Counting played an essential and useful purpose in our daily survival and evolution, so we humans have persisted in quantifying everything that we do.

In our zeal to applying the numbers we collect, it is also natural that we occasionally misapply the results from the counting; sometimes the misapplication is harmless, but sometimes the misapplication works towards our disadvantage.

Thanks to Malcolm Gladwell, the 10,000-hour rule is cited repeatedly as a heuristic for attaining expertise. (Gladwell 2011)  Unfortunately for Gladwell, the person who did the research, Anders Ericsson, very publicly refuted Gladwell’s interpretation of his research, which led to misapplication of the 10,000-hour number. (Ericsson 2020)

Yet people continue to misunderstand and misapply Ericsson’s conclusions, taking Gladwell’s attempt to popularize as original research.

Thinking about the continuing reliance on the 10,000-hour rule made me think about how we use counting in our coaching. For example: we give our players numerical goals as a quality control measure. For example, we can have them pass 20 balls, or we can specify 20 good passes, the good passes caveat is our way to introduce quality into the exercise. We assume that the players can make the connection between counting good passes and how to make good passes. Of course, that connection should never be assumed, but that is why coaches get paid big bucks, right?

As I started thinking in terms of system as I understood it, it occurred to me that most of our counting involve taking snapshots of a continuous flow of action; it is a snapshot taken at a specific time in a specific location. Sports — as with all of our reality that we experience— is a continuous chain of actions. We use Markov Chains, to be exact, to model sports as a chain of discrete events, even though it all looks like a continuous flow of events (Wung, Stats for Spikes-Markov Chains 2021). We count and make the assumption about the counting that we do because we need to simplify the reality so that we can understand what is happening in front of us. We do this by freezing the action at the time when a countable event happens. The drawbacks of that approach are that we cannot  possibly count every single touch all the time so that we can create meaningful sample space of statistics; counting discrete events creates an incomplete picture and does not capture  all the intangible qualities that characterizes the continuous reality. Yet, since counting discrete events is the only reasonable thing that we can do to capture reality, we continue to insist on counting. Most people understand this difference, but we have become so intent on the counting that we forget the reason for our counting. The tail is truly wagging the dog.

The motivating reason for Ericsson’s research, as well as the reason for Gladwell to write his book Outliers, (Gladwell 2011) is to distill the process by which people who do what they do well —much better than anyone else in the history of what they do—into a concise formula, another example of our brain’s need for formulas. Indeed, the problem with Gladwell’s writing is that he sometimes over distills and oversimplifies the academic studies in his zeal to explain the complex research results to the layman. Distilling and simplifying the complicate process of achieving excellence is what Gladwell does very well; indeed, better than most other popularizers. The point of his work is to show how someone can attain mastery— a word that is fraught with nuances— of a craft, a skill, a game, or a sport. It is indeed difficult to describe mastery, but it is easier to describe mastery than it is to describe how to attain mastery. The purpose to his writing is to point out the salient features of complex process that experts employ to achieve mastery. Part of the problem with distilling to the essential nature of the mastery process, however, is continuity. The path to mastery is a continuous process, an iterative journey full of failures, adjustments to those failures, and ultimately success. It is not a path that can be streamlined into a recipe or a formula. This is why the 10,000-hour rule is not only an erroneous and harmful misapplication, but it also does not directly address the bits connecting the numbers that we count to the process of how to master the subject that we wish to master; it only measures the byproduct of the mastering process, the number of hours of practice.  Some have amended the idea of the 10,000-hour rule by making the rule the 10,000-hours of Deliberate Practice. It is an improvement, but it still does not address the essence of How to do, it just addresses What to Do.

Returning to the more basic volleyball example. Do we really want our players to pass 20 perfect passes in a drill? Or do we want our players to take the opponent’s wicked serve and convert it into points for our side? I want the latter than the former. How does that happen?

It takes the convergence of the discrete parts of the playing experience to achieve this; a discretization that we coaches impose on the act of passing so that our players can internalize the disparate parts of the integrated action. We cannot, however, make the player learn how by throwing them to the wolves by immediately exposing them to the reality. We cannot coach by  hoping that the players can understand the action of passing by facing what they will be facing in game play. Some players can and they do this naturally, but for the rest of us, it takes years of experience, analysis, ability to self-analyze, and autodidactic learning to achieve. Unfortunately, most of the players will get frustrated and quit the game before that happens.

The act of teaching skills by parts and then teaching the players to make connections between the  disparate parts of a skill and then integrating the skill back into an integrated whole has become unfashionable. Even though teaching the whole prematurely has its own deleterious results.

One of the effects by focusing on the 20 perfect passes is that we are forgetting and neglecting the result we desire. This is not to say that we should stop counting, but we should emphasize that the counting has a monitoring function rather than a goal setting function. The conversation should be: 20 perfect passes in three minutes or passing 2.4 in the last drill; rather than using the counting results as goals: the goal is to get 20 passes in two minutes, or the goal is to pass 2.4 in this drill.

Rather than saying that 10-000 hours of deliberate practice is what will guarantee mastery, we need to say: continuous and focused deliberate practice is what is necessary to achieve mastery, do not let the amount of hours set your limit, take as many hours as you need.

Goodhart’s Law is once again at play here:

Goodhart’s Law: “When a measure becomes a target, it ceases to be a good measure.” (Wung, Stats for Spikes-Using Statistics as Goals 2021)

We need to stop using the intermediate measure from the process of learning as a target, we need to use it as a check on how we are doing.  The difference is nuanced but there is a clear difference.

Players will often make the same assumption, they will ask about what the goal is, it is only natural. This also a chance to include the growth mindset into the discussion.

Another problem with using the intermediate measures as the target is that reality is nuanced and depends on the context of the situation. In the case of the 10,000-hour rule, the context has to do with the attributes we are born with, physically, emotionally, mentally, and intellectually; as well as the circumstances surrounding our attempt at mastery. This is not to say that talent is a constraint that cannot be overcome. This is saying that not everyone requires the same time and effort to achieve mastery.

In terms of the simple passing example, the context has to do with the serves that is being passed, the skills of the server, the skill level and physicality of the team, the practice environment, and so on. Passing a 2.4 against the USA Gold Medal winning WNT is different that passing 2.4 against the players on your own team.

The point of this article is to remind ourselves that we must never lose sight of the prize, to focus our sights on the real result rather than on the byproducts of the process, whether it is to achieve mastery or to achieve better first touch while playing, rather than achieving 10,000-hours of practice or 20 good passes in a drill. This is a common problem, losing sight of our end goal. As Daniel Kahneman, the author of Thinking: Fast and Slow (Kahneman 2013) states: When forced with a difficult question, we often answer the easier one instead, usually without noticing.

Works Cited

Ericsson, Anders. "Anders Ericsson: Dismantling the 10,000 Hour Rule." The Good Life Project. 2020. https://www.goodlifeproject.com/podcast/anders-ericsson/ (accessed January 17, 2022).

Gladwell, Malcolm. Outliers: The Story of Success. Back Bay Books, 2011.

Kahneman, Daniel. Thinking Fast and Slow. NYC: Farrar, Straus and Giroux, 2013.

Wung, Peter. "Stats for Spikes-Markov Chains." Musings and Ruminations . March 21, 2021. https://polymathtobe.blogspot.com/2021/03/stats-for-spikes-markov-chains.html (accessed January 17, 2022).

—. "Stats for Spikes-Using Statistics as Goals." Musings and Ruminations . March 6, 2021. https://polymathtobe.blogspot.com/2021/03/stats-for-spikes-use-of-statistics-as.html (accessed January 17, 2022).