User:AW/generalTargetingThoughts

I have a few random thoughts on targeting that I thought I would share:

Kernel density estimations as used in k Nearest Neighbor targeting methods should be used in VCS guns. In guess factor targeting, it is pointless to shoot at GF's 1 and -1. This is because there will be 18 pixels of enemy robot between 1 or -1 and 0. So you could shoot at, for example 0.97 and it is possible that shooting there will hit them when they are not at GF 1 and it will also hit them if they are.

This brings up another point, at close ranges, some authors have their guns shoot at a point that the enemy must end up in. However, this could be easily accomplished, and hypothetically targeting at all ranges improved, if after the kernel density estimation, the angles within one bot width or one-half bot width were averaged together and this angle was used as the firing angle. ie. if we had bins 24 and 26, with 50 and 51 entries in them respectively, all other bins were empty, and the enemy bot covered 4 bins (half a bot width was 2 bins), then the best bin to target would be 25, not 26.