Notable Info
- It turns out the WHR system that we recently rolled out does not properly handle draws — for every draw, it was adding a win AND loss. Since I lumped close decisions in with draws based on the research I did w/ the Glicko system, this caused a pretty large ratings issue. I started noticing strange patterns of fighters in draws and close decisions who both lost considerable points directly after the fight.
Further research is required to determine if this is a systemic problem or an implementation problem. In the interim, I updated the WHR ratings and tossed out all draws, along with no contests and disqualifications. Before, we were also treating split & majority decisions as draws in the WHR ratings, but we will now treat those as wins (for now). Furthermore, I increased the accuracy as the ratings now run until the convergence is below 0.00001 (vs. 1.0). Technically, this paints a more “accurate” picture.