DynamicDistancing

This is what I call a concept I started using when I was developing SandboxFlattener's movement. I was sick of trying to figure out what the best distance was by changing my goal distance and putting it to 1000 rounds against SandboxDT, and I had decided as well that the best distance would be different for different opponents. Since then, MakoHT, FloodMicro, GouldingiHT and FloodMini have included meticulous measures of each robots energy losses and gains at every distance bracket, and they aim to be at the most beneficial one. Note that this recognizes that movement isn't just about avoiding getting hit - it's about finding the most favorable balance between hitting and getting hit. MakoHT currently has my best DynamicDistancing code, if anyone is looking for how to do it. FloodMicro, FloodMini and GouldingiHT also have code for it, but it is not as readable and sometimes not as good.

This is a very interesting way of doing distancing; Cake does a form of distance optimization, but not nearly as sophisticated as in MakoHT, and it doesn't save any data for futur battles. Cake simply tries a certain amount ahead of it's current distance, then that same amount back, then moves to whichever distance was best, and decreases the certain amount by a little bit and repeats. Theoretically, this should allow Cake to slowly slide into a comfortable distance against the opponents. In practice however, it rarely works; one unlucky stride getting stuck against a wall or something causes it to assume the wrong distance, and do stupid stuff like barge towards the enemy and be stuck near it for the rest of the battle. Hence you can get wierd results; try any bot vs cake for say 10 battles of 100 rounds. I guarantee it will end up with wildly different results each battle. MakoHT's DynamicDistancing is much better, what with the distance bracketing and all. Good work :-) -- Vuen

Thanks :-) It seems like more and more robots are implementing this, especially since I put out FloodMicro. -- Kawigi