For a while, there have been two things I’ve wanted to try and improve in when using event data. The first is trying to apply some context to the events, looking at things more as sequences rather than just adding or slicing up the events in slightly different ways. The second is looking at defensive actions with data. The only time I tend to use defensive actions is to say someone’s either active or not.
I haven’t made much progress, I haven’t come up with some kind of brand new approach or metric, but I have managed to hack together a few things I’ve found interesting. The problem is that hack together is the most important thing here. I’ve made a lot of assumptions and there’ll be plenty of times where things can slip through the cracks. Watching some clips to test, things seem mostly okay, but I’m not going to claim anything definitive for most of this.