The only way I could think to improve it is to take more attributes into account, but then you’re taking from a smaller pool of data per attribute and probably harming the model. In theory with enough data points you could make it a much better model. But you’d need stats from likely every...