View source for Talk:Mirage

From Robowiki
Jump to navigation Jump to search

Contents

Thread titleRepliesLast modified
Congrats on making the top 10!514:26, 20 March 2024
Interesting gun210:54, 9 February 2024

Congrats on making the top 10!

It didn't take you long to hit your goal after all!

D414 (talk)22:39, 18 March 2024

You do not have permission to edit this page, for the following reasons:

  • The action you have requested is limited to users in the group: Users.
  • You must confirm your email address before editing pages. Please set and validate your email address through your user preferences.

You can view and copy the source of this page.

Return to Thread:Talk:Mirage/Congrats on making the top 10!/reply.

I'm seriously impressed with the top ten 10 in the general melee rumble with a mini. I suspect that Mirage is going to be as difficult to dethrone from the mini melee throne as Wallaby has been from the micro one. I try not to think about how badly BeepBoop is crushing everybody in the general 1v1 ...

D414 (talk)13:08, 19 March 2024
 

Thanks! It did make it easier that the mini melee rumble hasn't seen active competition in a long time!

--Kev (talk)14:24, 20 March 2024
 

Congrats! Looking forward to a new general melee king to dethrone ScalarR ;) It's really impressive how a mini bot can do.

Xor (talk)12:37, 20 March 2024

Ha, I don't have plans to work on that for a while but maybe one day! I am curious if single-tick could outperform classic play-it-forward for megabot melee guns.

--Kev (talk)14:26, 20 March 2024
 
 

Interesting gun

Sounds like your gun is effectively a single tick pattern matcher, with a fixed match length of one and multiple choice?

At first I was surprised that could be effective but I suppose in melee the next state of most robots only depends on the current state (of the battlefield) so it makes some sense.

I guess in a mega bot it could be expanded to consider other elements of the battlefield state to get more accurate results?

D414 (talk)11:27, 8 February 2024

Yes, exactly! To do multiple choice, it samples the next tick at (weighted) random rather than picking the most frequent one as in classic single tick. Definitely using more input features could help, but I think velocity/deltaHeading/acceleration are the most important features and the play-it-forward sort of implicitly captures wall features by rejecting movement trajectories that go out of bounds.

--Kev (talk)04:21, 9 February 2024
 

Your gun does have an excellent performance, and it is really amazing that you were able to fit it in a mini.

GrubbmGait (talk)10:54, 9 February 2024