The PvT project

Help - Resources - AI Scripting - Grafting - Plugins - Modding Projects
User avatar
Myk
Posts: 257
Joined: Sat Jun 13, 2009 12:56 pm

Re: The PvT project

Postby Myk » Sun Jan 06, 2013 10:53 pm

Taranok wrote:I'm just wondering, do you know what the Template Terran is? I was looking at some old forums, and I see it referenced quite a bit.

TvP options are probably:
marine+tanks+medic (add ghosts or science vessels)
tank+goliath+vulture
one of those into mass air(?)

It's actually quite interesting that in small battles, zealots>vultures (with even army worth), but in larger battles (I tested 12 vultures vs 9 zealots), the vultures win quite easily, even with zealot speed. I think one of the main problems with mech is that support units quickly die, and then tanks can easily get wiped up by zealots or something.

The only way to afford throwing so many units away is to be ahead of your opponent in economy, which is not impossible, considering mech is built for being turtly. My anti-carrier AI went for double expand after detecting an expand build, however I doubt that's very viable once some counter build detects it.

I finally made my normal TaranokAI a formidable PvT opponent. :p


Template Terran was a generic Terran script with bio and mech builds by Archon_Wing. Kind of a cool idea, but I remember being disappointed with the performance of it and I ended up trying to rewrite it. If you're after an already written Terran to build on, I'd look at Accel T. It will take a little bit of effort to understand, but it's really the better way to write scripts.

Nice looking score % on TaranokAI. I'll have to check it out.
User avatar
Taranok
Posts: 256
Joined: Tue Sep 07, 2010 9:41 pm

Re: The PvT project

Postby Taranok » Tue Jan 08, 2013 12:50 am

I guess one of Myk's posts disappeared :O

I was able to Terran mech AI that was able to beat all of the Protoss AI's that didn't happen to 2 gate rush (didn't test against dt rush). This makes me feel like mech may be possible, but maybe I'm getting overeager because my AI is the most greedy AI ever written, because it goes expand, barracks, expand again.

Contrary to Racine style mech, the mech AI I isn't a bio mech build but purely vulture + tank. obviously in its current state it would still lose to a carrier rush, but it could probably easily transition into goliaths if detected. It doesn't get vulture speed either, in an attempt to keep the vultures from rushing too far ahead of the tanks.

it beat unleaded, JH24, racine, zelduck, peppe. A_Wing's is probably the most significant.

I'm kinda left wondering why mech is deemed not viable? Is it just because mech is just so cost inefficient that you have to be ridiculously ahead in economy (like my AI generally is) to win?
User avatar
Taranok
Posts: 256
Joined: Tue Sep 07, 2010 9:41 pm

Re: The PvT project

Postby Taranok » Wed Jan 09, 2013 5:57 am

Myk wrote:Played around with 1.71 tonight and couldn't get a mech build to beat it. The best I got was an AI that could take expansions and barely hold onto them. Even at it's best, I can't match that crazy protoss macro enough to launch any sort of viable attack.

Is it Battlecruiser time? :lol:


Took me 3 years to figure out how looped training should work....lolz. at least macro works now :)

The key to mech is just pumping attack constantly. Although I don't know how to afford that yet if the AI has to defend a 2 gate. However, with the 2 gate removed from a test version of TaranokAIv1.71, my mech v1.0 actually still managed to beat TaranokAI! Pretty awesome. I'll post the scripts later, when I'm not so busy studying for finals.

2 gate ruins just about every strategy ever...ha

TaranokAI currently doesn't have much of an answer to battlecruiser, but it may be enough for it to defend against a few, like Accel T throws in. Archons + feedback should be enough.
krasi0
Posts: 254
Joined: Thu Dec 18, 2008 12:42 pm

Re: The PvT project

Postby krasi0 » Wed Jan 09, 2013 8:27 am

For a constant mech push check Nova bot. It's a very aggressive terran bot that goes mecha in TvP & TvT
User avatar
bajadulce
Posts: 52
Joined: Fri Feb 01, 2013 7:02 pm

Re: The PvT project

Postby bajadulce » Wed Jan 09, 2013 7:12 pm

I've been so busy w/ other things that haven't had a chance to respond to your thread as have been meaning to do.

Just wanted to say that have managed to find some time to watch a few replays of your script w/ the recent BWAILauncher that was updated and was very impressed. :wub:

Great job on coming up w/ a winning build order/timing. Looks great. :thumbs:

@mech build:
Mech builds are "human pro" builds meaning they need micro to be effective vs. Protoss. Something the internal AI engine has clearly shown it does only a fair job at. Racine attempted to build such a defense in his first non$ and it clearly failed. Best would be a mixed army of units to support one another such as infantry (fbats & marines), medics, and tanks for the internal AI. Not as "sexy" but way more effective at least in the early stages, which will be very important vs any protoss as they are without doubt the most powerful of the internal AI.

Something like the Nova bot Krasi0 mentions however would be different (and must have been quite challenging to build I would imagine) as one would assume the micro management of this mech force is vastly improved over the internal AI engine.
User avatar
Taranok
Posts: 256
Joined: Tue Sep 07, 2010 9:41 pm

Re: The PvT project

Postby Taranok » Wed Jan 09, 2013 10:07 pm

Regardless of issues with the AI's hardcoded micro issues, I created an AI that held the 2 gate and went on to beat TaranokAI, and Baja P and Ashara P. YAYAYAY! :D (sorry I'm tired)

This defensive script combines with the quick 3 base script to make one TvP script. Although I modified some of the loops that are also used in the 3base scripts so I don't know if it works as well or at all. It also has a detection loop for stargate, as it'll stop building vultures and build goliaths instead once it's detected. Might be a bit of an overreaction, I'll probably change it to fleet beacon.

It seems like just Mass vultures with maybe a few tanks here or there seems to do the trick. :| I'll post the script as soon as I'm not too busy with studying and not so tired either, and also after I test the 3base script one more time.

I find it how one of the older scripts - Baja's P - actually gives me the most trouble. It just zealot rushes early and repeatedly, and so often it kills the natural and does lots of economic damage. The only problem with the scripts is that it macros horribly (floating 2000 minerals at maybe 15 minutes into the game). And it goes Zealot Archon, which seems very good against mech (but almost everything is).
User avatar
Taranok
Posts: 256
Joined: Tue Sep 07, 2010 9:41 pm

Re: The PvT project

Postby Taranok » Fri Jan 11, 2013 7:29 am

changes first post to a general list of AI's made in this thread. zbwaip really isn't the focus anymore.

Time for the bigger news:
Mech v1.0 released! I think it beats most of the Protoss AI's, except for Taranok AI and Blizzard AI (weird combination, eh?) I only tried it on one replay, and that's Destination (it was custom edited to remove the back door and enlarge the main slightly, so not one you guys would have). A few game I looked at were on Tau Cross. I picked those two because they had the largest mains. With 9 factories, and small mains, like in Icarus and Python, units get stuck all the time.

Right now it works in 2 ways:
builds supply depot, and then checks for gateway:
true: builds bunker, then expands, builds factories, and then runs attack and expand loops
false: expands twice, then build factories, and then runs attack and expand loops

the rush command did not work at all. rush(5, bunker2) was true even when there were only 2 zealots and none in training. rush(4, bunker3) was true even when there were only about 5 zealots.

It does deal with carriers too. Although I'm not sure about ZBwaiP or Myk's P, it could beat Ashara's and Baja's carriers that came around 15 minutes. Just throws out vultures and builds goliaths instead.

Might have to get some editable maps from BWMN to get maps with good main size.
I think map choice will become more important for mech. Mech is pretty awful on maps with large, open 3rds (eg python).
User avatar
Myk
Posts: 257
Joined: Sat Jun 13, 2009 12:56 pm

Re: The PvT project

Postby Myk » Fri Jan 11, 2013 5:48 pm

Looks good. I'll use it to guide my TvP into something that makes more sense. I like a few more marines with my bunker, about 2 extra, just in case the bunker gets busted. Against BajaP's second zealot attack 4 doesn't seem like enough to hold it off. Oh well, use whatever works!

BTW, there's some decent vulture rushes that can bring a lot of pain to some of the scripts that skip too many dragoons. Raynor P and I think Ashara were the victims. Doesn't work out too well with P has cannons or DTs, but everything else seems to be fair game.

It's pretty surprising how effective a group of vultures are. I thought it was all about siege tanks, but a bunch of vultures clean up some of the rushes pretty well. They need a not_so_many_mines_in_the_main(True) function. :P

You might try putting the factories in the expansion script instead of putting 9 in the main. I made a decent attempt at 3 at each expansion and that worked out alright.

About Myk P from Entropy: I've got no idea how good those are. It's been a while since I looked. :)
User avatar
Taranok
Posts: 256
Joined: Tue Sep 07, 2010 9:41 pm

Re: The PvT project

Postby Taranok » Fri Jan 11, 2013 10:44 pm

Yeah zealot rushes are quite a problem especially since the rush command doesn't work (or at least not consistently anyways). Versus Blizzard AI I even tried to put 3 bunkers up, but 3 bunkers didn't even kill any of the 9 zealots. I think maybe some of this is due to how the bunker targets enemies (?). 2 bunkers is pretty safe against Baja AI.

I'm thinking about looking into some of the semi-early pressure that can be done too. Although I think Unleaded and TaranokAI shouldn't have many problems dealing with that, against other AI's it could be effective, especially if the AI could snipe some probes.

I don't think I've quite figured out how to put factories in the main and the natural at the same time. I just dumped a player_need (because its universal(right?)) and hoped it would build some at the expansions.

I'll probably try to edit the AI to try and fix these sort of problems. Will look at your entropy P too.

Hey do you know how to fix this thing in my Mech AI where it builds 2 goliaths? I don't know why this happens, it happens after the 2 machine shops are built...
User avatar
Myk
Posts: 257
Joined: Sat Jun 13, 2009 12:56 pm

Re: The PvT project

Postby Myk » Sat Jan 12, 2013 4:19 am

Taranok wrote:Yeah zealot rushes are quite a problem especially since the rush command doesn't work (or at least not consistently anyways). Versus Blizzard AI I even tried to put 3 bunkers up, but 3 bunkers didn't even kill any of the 9 zealots. I think maybe some of this is due to how the bunker targets enemies (?). 2 bunkers is pretty safe against Baja AI.

I'm thinking about looking into some of the semi-early pressure that can be done too. Although I think Unleaded and TaranokAI shouldn't have many problems dealing with that, against other AI's it could be effective, especially if the AI could snipe some probes.

I don't think I've quite figured out how to put factories in the main and the natural at the same time. I just dumped a player_need (because its universal(right?)) and hoped it would build some at the expansions.

I'll probably try to edit the AI to try and fix these sort of problems. Will look at your entropy P too.

Hey do you know how to fix this thing in my Mech AI where it builds 2 goliaths? I don't know why this happens, it happens after the 2 machine shops are built...


Get some more marines. Maybe one or two more. It's not going to break your AI to get a few more marines. I can't keep the bunker vs Ashara P's rush, but I do clean up the zealots with just marines.

As far as the factories go: As long as they make it out of the main with 9 in there, might as well continue to do that. I had a lot better luck building my 6 inside the main instead of split since the enemy units get extra stuck finding their way into the main from the natural. I still build one factory with an addon at each base to keep the tank count okay.

I still can't grab a win off of Taranok or Archon_Wing P. Just launching an attack before the Protoss goes into insane attack spam mode hasn't happened. Maybe earlier upgrades might help? Hrm.

Return to “StarCraft Modding”

Who is online

Users browsing this forum: No registered users and 1 guest