Can Stats Capture Player Development?

Recording Developmental Actions in Volleyball Stats

At the UK Volleyball Coaching Symposium, renowned US coach Jim Stone asked me a question. It was whether volleyball stats showed any actions that were the right thing to do developmentally even if the result is negative. My answer was, unfortunately, no. The stats only record the outcome of the actions rather than any intentions of doing the right thing developmentally. This question immediately got me thinking about the possibility of tracking developmental actions.

Why Track Developmental Actions?

It would be interesting to have this kind of data. One example is my emphasis on aggressive attacking (it’s my fifth team value). I want players to jump and attack high-ball and likely poor sets and not passively dig over free balls. While the current stats record free ball attempts, it would be useful to know how many of those free ball attempts could have been attacked. I am sure there are other situations like this that you would want to track that aren’t in the traditional stats.

How to Track Developmental Actions

I’ve already been using custom codes in DataVolley to record specific actions like incorrect defensive positioning or poor setting choices. These codes are not for counting errors but for identifying areas for improvement in video feedback sessions. However, I could use the same system to quantify these actions for a more comprehensive view.

What Developmental Actions Should We Track?

This is a big question because this list is exhaustive. Just think about all the feedback you have given players in your coaching career. You can probably list hundreds of developmental actions to include. However, do you want to include them all? You also run into issues with definition, e.g. I want my players to be aggressive in attack even with poor sets but when is a poor set hittable and when is it unhittable? There’s also a practical issue with what custom codes you will use. If you identified 100 developmental actions you want to track, you are simply going to use the custom codes of “001” to “100”. That is going to be a nightmare to remember them all.

A Simplified Approach

To address these challenges, I propose a simpler approach: coding developmental actions as either positive or negative. It could be for any reason. This avoids overly specific definitions and allows for flexibility. By calculating a ratio of positive to negative actions (i.e., an efficiency ratio), we can assess a player’s overall performance while considering their developmental efforts.

Below is an example of a player’s attacking stats. The Hitting % isn’t great at 0% but the Developmental Actions Efficiency Ratio is good at 4.0. I can build a narrative that the player’s attacking isn’t efficient right now but they are doing the right things. Of course, you would need to watch those attempts on video as to why it was positive. It could be that the player is super aggressive on every set, hence many errors.

Limitations

This system has three major limitations. First, the stats are vaguely defined, leading to objectivity issues. The same situation could be interpreted and coded differently depending on various factors, such as the coder’s mood, biases, and individual judgment. I know that when I sometimes see the custom code for video analysis months after I entered it, I forget why I entered it in the first place. If there’s too much subjectivity in the stats then it won’t be trusted or utilised in any decision-making process.

Second, there’s coding inconsistency, resulting in reliability issues. I may code positive and negative developmental actions for being aggressive in attack because it’s an area for improvement. However, I may not code these actions for a libero who has to be outside or a setter, as that’s not their primary role. Additionally, I might not code these actions in the future because the team is now sufficiently aggressive and we have moved on to focusing on other aspects of the game. Again, if the stats are unreliable then they won’t be trusted or utilised in any decision-making process.

Third, I can only code to actions where the ball was contacted and not to actions where the ball wasn’t contacted. For example, one of my team values is hussle, i.e. going for every ball. So player #1 shanks a pass and players #2 and #3 do not chase down the ball. This will be recorded as player #1’s reception error. However, how do I record that players #2 and #3 did not chase down the shanked pass? If I add the code to player #1’s passing error then that negative hussle is attributed to player #1 and not #2 and #3. I can’t add another main code to players #2 and #3, e.g. setting or defence error, because it would mess up the main stats.

Final Thoughts

I like the idea of tracking developmental actions in volleyball. It can provide a richer understanding of player performance beyond the final outcome. By combining traditional stats with developmental action data, coaches can make more informed decisions and help players reach their full potential.

However, a lot more thought is needed to make it work to overcome the issues of objectivity, reliability, and non-ball contact actions. I have started to code some of these things with my American School of London team so I will see what that experience was like.