I have one question for Walt Hickey about his Civil War article. Why did you take the user input data (which has beautiful intrinsic Rock-Scissors-Paper features) and linearise it? Seriously, this is basically the equivalent of going into a Pokemon game and getting rid of types. WHY?
Anyway, so I ran some analysis on the raw data (rather than the linearised data) and, well, Captain America still does really badly.
Captain America Win Rate 20.4%
So I thought I’d be a mad genius and ask the question (that Walt also asked) what if Cap got to pick the battles?.
Strategy One: Our best, their worst
Every round, just pick Cap’s best player, and take Iron Man’s worst…
Write and run some sweet code
Captain America Win Rate 19.9%
Oh no. No no. Oh wow. Maybe… Maybe I wrote the code the wrong way around. Let’s reverse the teams. That’ll do it.
Strategy Two: Our worst, their best
Edit and fix that sweet code
Captain America Win Rate 16.8%
Anyway then I decided to mad beast mode it up and give up on those awful strategies.
So, like, the biggest problem with running your best against their worst until your team wins is that by the time you get to their best player, you’re down to, like, Falcon.
The worst.
Anyway, so, you want to avoid that at all costs. So how about getting rid of him early? Why not run your worst (on average) player against the person they have the best chance of defeating? They’re gonna be crap right at the end, let’s use them as cannon fodder up in front.
Onemillionsimulationslater
Captain America Win Rate 22.8%
Short of doing an actual optimisation, and writing actual code, this is where I’m going to leave it. It’s a Friday night, beer O’Clock!