Every year there are a few things with these rankings that have me scratching my head, wondering how it happens. This week I see one of those. I'd like to be able to work out how Follett is ranked above Benjamin, a team that beat them by 30+ earlier in the season. Granted, The Toy doesn't look at who beat who. It just runs numbers. This is just the kind of thing that baffles me and costs me hours of staring at this algorithm, and ratings, and just hours trying to work out how in the world that happened.
Just for the record, I am rarely able to work these things out. It's just fun, albeit frustrating, for me to try to figure these things out so I can better answer questions as they come up. In this case, though, my only answer is this thing is way too complex to figure out.
I will say, though, that for the last two weeks, it's been at over 80% accuracy overall and, for a computer system with around 300 teams, I'd say that's pretty darn good!
