I know this has been asked before but are there any good bots available for offline testing?
I’ve been uploading 3-4 different bots and slowly slowly fixing tiny glitches, wish could optimize the initial defense list automatically. Wish could test things offline against a colleciton of decent bots.
From other games I know playing against my own bot gives very specific/overfitted attack strategies that never work in production against others.
Well the issue with that is, that we would need the code of those bots and anyone could just upload them to the actual competition (or a slightly modefied version).
So the answer is no, at the moment (as far as I know) there aren’t any bots for offline testing.
But you could probably make your own in about 5 minutes and just use that (just make some maze, pingcannon or something along those lines)
You can also make a bot that try to copy the behavior of some other bot using a replay file.
This way you should be able to simulate most static bots (won’t work too good with adaptive or randomized bots).
Replay parsing is a good bet. Just grab the ones you want you’ll get a fairly good representation of it, perfect of it’s static. Even dynamic ones would be worthwhile if you’re having significant issues against a particular type of a issue or crash bugs on the live server you want debug messages for, so I’d recommend going this route.
The quickest option would be making a couple of your own bots to test with, though. A lot of the maze ones require all of ~4 lines for attack logic and just a few modularized chunks of static defense logic.
That makes sense, having 3-4 different simple maze/ping cannon bots would at least help optimize scoring multiliers & help find bugs.
Sending out scramblers at the right time easily blocks ping rushes, corner guns need some proactive defense, but currently figuring out how to optimize against dumb-dumb-tree ‘scrambler cannon’
Long term it would seem my semi static maze bot couldn’t survive long at the top as other player would optimize against it. Either need to scramble 3-4 different static-y strategies or else go completely dynamic.
Having one static design certainly does make it vulnerable to algos that change (dumb-dumb-tree is mine). In replicating it’s strategy you could try just doing it after a set number of turns (for a testing algo). The challenge here is that it only does it if it thinks it is advantageous (it’s still wrong a lot :P).
I definitely agree with this, since I only added the cannon after realizing the way my algo makes decisions made it super weak to maze algos, which a while ago were extremely popular so its rankings dropped significantly because of it. I found that the cannon was (usually) a good way to counter that.
yeah the weakness of maze algos is they are not good at targeting attacks, I see dumb dumb stacking up a lot of encryptors right in the center which my bot ignores. 1-2 EMPs can’t win against 8 shielded Scramblers or Pings. I just got a randomized Bot up and running, will try transferring my maze logic into that in piece by piece, so can be a bit more dyanmic.