[Home]History of Mako/Movement

Robo Home | Changes | Preferences | AllPages


Revision 3 . . May 17, 2003 21:23 EST by PEZ
Revision 2 . . (edit) May 16, 2003 10:47 EST by PEZ
  

Difference (from prior major revision) (no other diffs)

Changed: 1,6c1
I'm not ready, yet, to reveal what kind of theory is behind Mako's current movement. But I can say this:
* When I got the idea I thought that it would be:
** hard to hit for PatternMatching aim
** easy to hit with StatisticalTargeting
* OrcaM is one of the best bots around against it (though I haven't tried ScruchiPu yet because it doesn't work on my PowerBook).
Having given this some thought I now understand why pattern matching works so well against it, but I still don't understand why it flattens the [movement curve]? like it does. The movement leads to predictability on the short term but not on the long term. I should have realised that immediately, but I'm glad I didn't since then I would probably not have taken this path.
I threw out the Mako 1.3 movement system since it was flawed and predictable in the short term.

Changed: 8,10c3,4
However, I now think I know how to get my RandomMovement implementations to produce flatter movement curves so I might soon release a bot which tricks both statistical and pattern matching aim. "Might" being the key word there. =)

-- PEZ
Now Mako sports a kind of RandomMovement. Soon enough we'll see if it's any good.
-- PEZ

Robo Home | Changes | Preferences | AllPages
Search: