I have an idea of advanced two layer statistics, and I'm going to start a new bot. An interesting fact is that targeting and movement will use completely the same idea, probably the same statistics class. The main idea is that (probably) every bot has spikes in his movement and firing, or it has to flatten it's graph. This algorithm should adopt to such behavior (of course it would handle the spikes like simple statistical bot). hope it will work :)
First try to implement this idea. Simple one dimension statistics by now. Testing results overview: ~11% hit rate against Shadow and SandboxDT. Not so bad. Using more dimensions and fixing most of bugs should increase hit rate.
Any comments, etc. are welcome
testing targeting would be better against known weaker bots first - good guns can effect the score against bad movers - against shadow, a gun firing randomly betwen a GuessFactor of -0.2 and 0.9 will probably yield similar results. -- Paul
I have tested it a little agains random movement bot. If robot's movement is not flat, my bot should fire like simple statistics bot, but it's hit rate was definetly higher than my statist bot (that's strange a little). By the way, fixing bugs I raised hit rate aigainst SandBox to 12.5-13% -- asm