So. Made my first Shiny web app using R and feeling pretty good right now. It’s not sexy by any means – and really only displays 4 different plots that are pre-made, but it’s still super cool compared to the absolutely nothing I had built before this.
You can check it out by clicking here: Harker Effect Size Shiny App!
Within the app, we are looking at how Attack Efficiency Change affects winning or losing the set or the match.
The visual you see is just a distribution of each player’s attack efficiency change in won sets/matches (teal) or lost sets/matches (red). Some players have very wide distributions indicating that they hit at a wide variety of Eff Change numbers in won sets or a wide variety in lost sets. The taller the peak of the curve, the more sets/matches which fall into that range of Eff Change.
The app allows you to look at the data at the set level, the match level – as well as hard-driven or offspeed attacks.
Behind the scenes we’re looking at the stat, Cohen’s d, to measure of how each player’s contribution fluctuates between won and lost sets or matches. Cohen’s d and effect size are used to compare means using standard deviations. I’m not an expert, but my layman’s understanding is that if Billy for example hits on average 0.300 with a standard deviation of 0.100, then if he hits 0.350 in won sets and 0.250 in lost sets, he would have a cohen’s d of -1 because he hits exactly one standard deviation lower in lost sets compared to won sets. Traditionally, anything larger than 0.8 would constitute a large effect size.
I’m still working on adding the specific stats to the web app, but where there are large shifts between teal and red curves for a specific player, that indicates a larger effect size for that player with those contexts.
If you look at overall attack efficiency change per set, regardless of hard-driven or offspeed, you get the following result:
Of course this doesn’t prove causation or anything like that – but maybe getting Billy Fan going is really the key difference between won and lost sets. You’ll notice that Jarrett also plays a large role here.
The funny thing is that Jeffrey Kwan, really the glue guy on the team, has an attacking performance that doesn’t seem to matter in wins or losses – meaning he hits about the same either way. This might suggest that teams should spend more time slowing down Billy or Jarrett as a means of beating Harker at the set-level.