AIIDE 2017 results announced
AIIDE 2017 full results are out. I’ll discuss tomorrow. ZZZKBot came out on top with more cheese builds by learning which cheeses to play.
AIIDE 2017 full results are out. I’ll discuss tomorrow. ZZZKBot came out on top with more cheese builds by learning which cheeses to play.
Comments
krasi0 on :
krasi0 on :
Congratulations to PurpleWave and Iron. Those are the real winners in my eyes!
Also not sure how PurpleWave would have crashed (25 times!) the Broodwar process when it's running as a client. That needs to be investigated IMO.
LetaBot on :
Dave Churchill on :
Crashes are up this yer, but crash rates for the re-submitted 2016 bots are almost identical, so I don't think it's a tournament issue.
PurpleWaveJadien on :
krasi0 on :
In general, I am a huge open source user and proponent myself (main OS Linux years for 20 years now; I have been using professionally only open source IDEs, text editors, etc. since forever; I have contributed to multiple open source projects on GitHub and so on...).
That said, BWAI development as a hobby of mine (since '09) is kinda of an ego thing for me (wanting to become the next king of the hill all on my own) which I guess doesn't speak well of my character but everyone's got flaws, right? :D
Of course, that doesn't mean that I don't consult other bot developers from time to time online for opinions, known strategies and advice. For example, it was yesterday, I think, when I asked tscmoo and jaj22 on Twitch about the logic behind the valkyrie and vulture patrol micro.
But for now, I have forbidden myself to read / use other bots' source code. If one day, I hit a wall and reach my limits, I may do what you did, but for now I will go on fighting the uphill battle and see where this gets me :D
That's the awesome thing about the SSCAIT ladder - so many entries submitted, so many different styles and approaches to the toughest AI challenge in the most popular contemporary RTS game - BroodWar. And in the end, all we've got is the bloody arena with the starting 4 workers and 1 town hall for each player where we measure our smarts and endurance :)
Antiga / Iruian on :
#begging, at very least swear on the dead body of Blizzard North that if you ever find SSCAIT / making the bot not interesting you release the bot to the public before you disappear into the internet ether time machine so that all that knowledge isn't lost. That's the truly scary part, having knowledge not build-able for future AI projects or lost.
krasi0 on :
But there is also the curse of some (successful) open source projects which see some popularity and the end result is effectively burnout for the authors / maintainers. It's a huge responsibility, burden and sometimes a big headache. Just imagine how you often need to answer support requests (support work is tedious and time consuming on its own), everyone expects you to fix bugs quickly and for free. Additionally, the reality is that no one else is likely to step up as the next sole maintainer.
Then there are the shameless copycats who will just strip out your name as the author and publish the bot as their own without giving you any credit.
There are even more reasons that come to mind, but the above ones should be sufficient for now :)
Jay Scott on :
Jay Scott on :