Wednesday 4th of February 2015, 23:00 CET 15:00 CST @ SSCAIT Stream
Sound only on 480p and 360p idk why
Getting there... Both final contestants, XIMP and LetaBot, breezed through the Ro42 group phase with equal ease, both going 37-4. In the Ro8 they 1-0'd Skynet and krasi0bot respectively. The semi finals were equally one-sided, with LetaBot dominating the 2012 mixed champion ICEBot 2-0, while XIMP set an equal score in its match vs UAlbertabot. So running into this final best of five, we have got 2 contenders with both a score of no less than 40-4, or a win-ration of 90%. For both, there is only one thing left to prove.
The interbellum Both programmers got, and used, the brief spell between semi finals and finals to make some final modifications to their AIs. Those of you who have been watching the stream may have seen some of the new strategies already, for now let their new-gained skills remain in the dark, for the final versions of each bot which will play the finals are still unknown...
War will rear its ugly head for the last time, as XIMP and LetaBot finally meet each other. Will the carrier prince once more resort to carriers, or has Vajda finally taught his pony to do another trick? Although the first one, teching to carriers behind cannons, was very effective, it was highly predictable. LetaBot, which used to go marine only might have a hard time killing the static defences. So what will it have come up with to counter the cannons? And what should XIMP do to counter the counter?
The same goes for LetaBot: its trimmed-to-the-bone barracks rushes (proxy and bunker possible too) are very predictable. What should it learn in the short week between semi finals and finals that will make it either withstand the carrier attack, or take out its enemy before carriers are out? Might not making alterations to delicate programs at such short notice introduce the danger of either AI stalling? It is all up in the air, and the only thing that can be certain, is that whoever wins this best of five will be the deserved champion of the Student Starcraft AI Tournament 2014.
On February 05 2015 13:21 Crunchums wrote: pretty silly to see LetaBot completely switch strategies in order to hard counter what the opposing bot was doing : P
Programmers shouldn't be exempt from the mind games.
On February 05 2015 13:21 Crunchums wrote: pretty silly to see LetaBot completely switch strategies in order to hard counter what the opposing bot was doing : P
If the bots were not allowed to change, then my bot would 3-0 Ximp due to the old version of Ximp not being able to cope with a bunker rush. The new version of Ximp hard counters a bunker rush, but didn't take the other strategies from LetaBot into account (the strategy that you see in the video was already in LetaBot, but not activated until now).
On February 05 2015 13:21 Crunchums wrote: pretty silly to see LetaBot completely switch strategies in order to hard counter what the opposing bot was doing : P
If the bots were not allowed to change, then my bot would 3-0 Ximp due to the old version of Ximp not being able to cope with a bunker rush. The new version of Ximp hard counters a bunker rush, but didn't take the other strategies from LetaBot into account (the strategy that you see in the video was already in LetaBot, but not activated until now).
sure; i'm just saying that i think it's weird to allow the bots to change
On February 05 2015 13:21 Crunchums wrote: pretty silly to see LetaBot completely switch strategies in order to hard counter what the opposing bot was doing : P
If the bots were not allowed to change, then my bot would 3-0 Ximp due to the old version of Ximp not being able to cope with a bunker rush. The new version of Ximp hard counters a bunker rush, but didn't take the other strategies from LetaBot into account (the strategy that you see in the video was already in LetaBot, but not activated until now).
sure; i'm just saying that i think it's weird to allow the bots to change
We are saying that it's weird to say that it's weird
The replays from the Ro42 can be found here. If you want, you can freely cast these replays.
why did you make firebats?
At this moment, I am making the wraith micro to be something like what the Berkeley Overmind did with its mutalisks. But wraiths are not a staple unit in TvP, so I made a BO specifically designed to be used against Protoss.
The BO that LetaBot uses against Ximp was based on a build order used to counter other Protoss bots that go heavy on the zealot count (a firebat takes down zealot shields faster than a Marine). The barracks lifting issue had priority over changing the BO to lessen the amount of firebats. As you can see in the finals video, I didn't finish the improvements for that issue, so I never got around to changing the BO.
On February 11 2015 07:51 BisuDagger wrote: I'm casting over the existing youtube videos at 9EST tonight. Everyone come enjoy the brilliance of these guys!
@Letabot, if you have the sc reps can you PM them to me. I want to see if I can get a overlay to works and cast from the game client itself.
On February 11 2015 12:57 Assault_1 wrote: [img]http://imgs.xkcd.com/comics/game_ais.png[img]
not helped by the fact that people still create mostly cheesebots when the AI has SO MUCH potential when it comes to midgame/lategame mechanics and macro. the amount of N-pool bots in this tourney many years after the creation of berkeley overmind is simply saddening.
EPT![[image loading]](http://i.imgur.com/7VDYgBy.png)
![[image loading]](http://www.teamliquid.net/images/forum/wiki_icon.png){kind=icon}
![[image loading]](http://imgs.xkcd.com/comics/game_ais.png)
