Thursday, September 17, 2015

Positive vs Negative Feedback or Regression to the Mean

I saw this video the other day. It is in name about whether punishment or reward is more effective. There are examples of anecdotal experiences with one or the other being more effective. I really like science, and this YouTube channel has a nice accessible approach to presenting science. It's also interesting to me as a coach for a couple reasons.

The obvious being that we have all been on teams or have had coaches where punishment was used as a primary method of improving performance. Maybe we are that coach. That coach has their own anecdotal evidence for that method of coaching. Hopefully we have also been in the opposite situation where reward has been used to encourage improved performance. This coach probably has some anecdotal experience, but they also happen to have empirical evidence backing them. In short, we all have anecdotal evidence backing our approach, but some approaches also have empirical evidence. The ones that are empirically backed are better, no matter what our anecdotal evidence says.

I also find it interesting because of the idea of regression to the mean as an explanation for some of that anecdotal evidence. I think we can use this information as coaches. Thinking about volleyball in statistical terms, every player has an average performance level for each skill of the game. The same could be said for team performance. There is going to be some variance in performance, and there is going to be a distribution of performance in each of the skills. Most of the time the attempts to perform the skill are going to be within a certain range. As the player improves, the average changes in a favorable direction, and the standard deviation is going to shrink, meaning less variance. Running this through the idea of the regression to the mean, when a player has an above average skill performance, the next performance will likely be closer to the average performance. It isn't always going to be this way, but over time more often than not, the subsequent attempts will be closer to the average. This is going to happen independent of other factors. If the player has improved significantly, the great performance will not have as big of a regression to the mean as in the past. One of our key objectives in helping players develop is in improving the mean. A great single instance of skill performance by itself is not an indication of improvement. One bad set is not a sign of danger, but multiple bad matches might be. When looking at performance, don't just evaluate the last match. Evaluate the last few.

Similarly, when looking at team performance, it's probably best to have the big epic set win come at the end of the match. I know, this is not something we can just choose to do. Looking back to NCAA men's national championships, there were two consecutive years where the same team won long sets. The first instance has BYU winning the first set 44-42. Even with sets going to 30 points that is a marathon set. They went on to lose the next two sets and eventually the match. Here we have a huge effort to win a set. Regression to the mean suggests there is going to be a let down just because of statistical probability. They played above average and then they played closer to average. Other team motivation, emotional let down, and every other reason sports commentators might mention may or may not be a factor, and they probably weren't a factor. It's just the tendency to get your team average with repeated attempts. The following year BYU won the 5th set 19-17, right after winning the 4th set 32-30. You could even look at Long Beach State's 30-15 1st set win as an above average performance where they will regress to the mean with a simultaneous regression to the mean for BYU's below average 1st set performance. Again, this is likely just a statistical phenomenon. It says nothing about team motivation or player character and toughness.

How can we use this as coaches? Our first concern is to help our players and team develop and structure practices to improve the average performance. Find out what statistical measures are most important indicators of success and measure performance in those stats. We want our players to improve and bring up those averages. I think this is where the competitive cauldron is a valuable tool. Second, when we are in a match we need to keep in mind that there is probably going to be a regression to the mean in performance, both good and bad. This is when coach as a psychologist or therapist comes into play. Someone does something incredible, and they will probably not be able to keep it up. Likewise, when someone commits an error, they will probably do better in subsequent attempts. We need to be our players' biggest cheerleader. Help them keep an even keel and not get too down when they regress to the mean. Help them to weather the storm.

The video mentions Thinking, Fast and Slowby Daniel Kahneman. It's a book about our two systems for thinking that are fast and slow. I'm partway through right now, and I just got to the part about regression to the mean (which mentions the Israeli fighter pilots the video talks about). There is a lot of good information for coaches to consider that should influence how we teach skills and coach players. There is also a lot of good information about cognitive biases and heuristics that will serve useful in evaluating what we look at in our efforts to improve our methods and how we are as coaches.

No comments: