Bot tournament on ICCUP (Finished)

https://github.com/bwapi/bwapi
Heinermann
Posts: 362
Joined: Thu Apr 26, 2007 12:42 am

Postby Heinermann » Mon Jul 04, 2011 2:15 pm

Quick rundown:

krasi0 (as krasi1) (T) vs Aiur (as skynet_bt) (P)
Aiur
  • Did not spend enough. Often had over 1k minerals.
  • Could have taken out an engineering bay with no loss using dragoons (which would have made krasi0 lose 125 minerals and push back turret production).
  • Continuously suicided observers into turrets, which accumulated over time and became a huge loss of resources.
  • Did not get zealot speed upgrade (probably the most noticable, caused losses in battle because they were outmicroed or took too long to reach a destination fast enough, and by the time that happened reinforcements had already arrived).
  • Did not get dragoon range upgrade (another noticable cause of loss).
  • Suicided corsairs into turrets and goliaths when attempting to attack a valkyrie. (Maybe should have gotten disruption web?)
  • Didn't get any High Templars or Reavers. Didn't move to late game (Arbiter).
  • Defended a drop nicely.
  • Had 3 zealots in the enemy's natural line but decided to attack a bunker it had no hope of killing, instead of attacking SCVs which would have been a gain.

krasi0
  • Let a Science Vessel die.
  • Constructed a base far away for the 3:00 base. (Did the observer cause an issue?)


EISBot (as krasi1) (P) vs Skynet (as skynet_bt) (P)
EISBot
  • Let a probe nearly kill a pylon, just left it while it owned an idle zealot.
  • Did not rally its units when they came out of Gateways (You can use rally point).
  • Attempted to micro units for a lost cause (many attack) instead of retreating.
  • No mid-game units.

Skynet
  • No shuttle play. Reaver drops would have been devastating to EISBot and would have completely crushed it earlier on.
  • Constructed too many cannons later on.
  • Did not scout to finish off EISBot sooner.
Laccolith
Posts: 8
Joined: Mon Nov 30, 2009 3:12 pm

Postby Laccolith » Mon Jul 04, 2011 8:35 pm

We just ran the final best of 3. First replay corrupted, it causes Starcraft to crash when viewed, though it may be possible to fix, in which case I'll update this post if that happens. The other 2 games are fine to view however.

http://dl.dropbox.com/u/40551/Replays/K ... Skynet.zip
Kae
Posts: 35
Joined: Tue Jun 28, 2011 1:26 pm

Postby Kae » Tue Jul 05, 2011 5:31 am

Thanks Heinermann for your feedback. As Laccolith said, our reply with Krasi0 is corrupted and I can't watch it correctly. But I saw you certainely take an old version of AIUR for this contest, since I spot that some new features are not applied. But nevermind, it's not a big deal. ^^

Could you please give me a sum-up of our game with Krasi0?
User avatar
Myk
Posts: 257
Joined: Sat Jun 13, 2009 12:56 pm

Postby Myk » Tue Jul 05, 2011 5:55 am

Notice: Game 2 of Krasi0 v Skynet seems to desynch about halfway through for me.


I'm happy to see that these scripts are improving and evolving each time that I see them. However, there were some things that didn't seem to make a lot of sense to me. I've spent a lot of time with Starcraft, writing BWAI scripts for play against both humans and other BWAI scripts, theorizing and analyzing replays, as well as watching lots of Day[9]. The following are some of my thoughts and perspectives on Starcraft and the games I got to watch in this tournament.


Have an economic or aggressive focus in the early game, but not both at once. 3 Factory or 4 Gateway builds are great for rallying units to an opponent in an effort to break their defense, but if you're not utilizing all of those units, you are making yourself lose, by way of choosing a less economical route, for no reason other than having a slightly larger army at a time when you might not have needed it.


Mid-game should be where the major economic plays come in. A focus should be kept on limiting an opponents bases while expanding your own. Typically, Protoss and Zerg try to stay slightly ahead of Terran in bases, and if Protoss and Zerg are too far ahead, Terran is either doing something wrong or the other race is playing Greedy and may be punished for it. Drops and harassment takes place in this phase: Reavers, Vultures, High Templars with Psi Storm, and Mutalisks, they can all help slow down your opponents econ which translates directly do their army size.


As far as specific things that I saw in the tournament that seemed of concern:
Krasi0
-Vultures not using Spider Mines. These can be placed tons of places and be effective. Other expansions, peppered around the center of the map, or in the main base of an opponent if you get a lucky drop.
-Vultures are great at killing other workers.

Skynet
-Protoss vs Terran almost always revolves around getting relatively fast Dragoon Range. This can either be used offensively or defensively. In game 3 against Krasi0 it built 4 gateways and then sometime after it expanded upgraded range; try to push this a little faster, if possible.
-Reavers really benefit from Shuttles.

EISBot & AIUR
-Early game defense against opposing workers. It really sucks to lose a worker or a pylon really early.


I hope some of this helps. To anyone writing bots or playing, I do suggest checking out the articles on Liquipedia. It was a great resource for my own work and just might help you out, too. :)
Heinermann
Posts: 362
Joined: Thu Apr 26, 2007 12:42 am

Postby Heinermann » Tue Jul 05, 2011 4:45 pm

If others are wondering, krasi0's vultures do use some spider mines, but they don't use ALL their spider mines, and vultures die while having 1-2 mines left for placement.
krasi0
Posts: 254
Joined: Thu Dec 18, 2008 12:42 pm

Postby krasi0 » Tue Jul 05, 2011 7:27 pm

[quote name='Heinermann' date='05 July 2011 - 06:45 PM' timestamp='1309884313' post='10281']
If others are wondering, krasi0's vultures do use some spider mines, but they don't use ALL their spider mines, and vultures die while having 1-2 mines left for placement.
[/quote]
yes, they do. but the bot attempts to spread them on a big area at some distance from each other and sometimes the vultures engage the enemy before they spotted a good location to put a mine on. that's why they die before mining. It on purpose doesn't mine near sieged tanks as this is sometimes suicidal.

Return to “BWAPI (wrappers in other languages / questions and announcements related to BWAPI itself)”

Who is online

Users browsing this forum: No registered users and 1 guest