DarkBlizz

Hahaha, I'm laughing WITH RAGE!

Please, "LAN.GAME" We want to play
^Nobody is pointing that out?!!? English pplz

Basic:
"LAN.GAME" 하고 싶어요.
I want to play "LAN.GAME".

Prefered:
"LAN.GAME"  할수있게 해주세요.
Please make it so that we are able to play "LAN.GAME"

Quote from: Rosenberg on March 06, 2010, 12:52:43 AM
Something is wrong, either I'm good or all the AI available are too easy for me. I was eager to try this version but turned out to be a little easy for me (won as a zerg twice against protoss and terran). I have to say, it's better than the previous versions though. Thanks for the hard work anyways guys, keep it up.

if you played any other RTS this is a picnic. Yeah but I fought worse players than this.

Forgot to add, too many zealots with Protoss. Zealots get owned too easily, plus they don't attack early enough with zealots.

I think it's not a huge improvement but improvement nevertheless. Protoss seem to be not doing very well and zerg likes to spam but doesn't do that well. Terran was the most fun but no counter to units means crush crush crush. Felt like playing against a noob-mediocre player and that's all I ask from an AI.

Purifier/Soul Hunter: they fired one missile and as they kill more biological units they could fire upto 3 missiles at once.

Quote from: sLayed on March 05, 2010, 10:14:41 AM
For those who say building a fully mixed army from the get-go is a failure tactic, please just watch some jaedong fantasy games. The amazing thing about zerg is every unit the race can make stems from the hatchery, allowing for a very diverse and ever adapting build especially considering the cost of the units.  This is of course beautifully countered by the fantasy build, where the individuals skill is truly brought out and great games of starcraft can be played.  Both races are tested to the limit, and the players ability to utilize each unit in counter to their opponent is tested miraculously.  The fantasy build goes up literally 80% of the techtree in 5 minutes or so.  Early vulture harass into wraith or dropships, followed by possible valkyries or goliaths or even another swap into medic marine tank army.  The "vulture => valkyrie => mm army complimented by tanks" can be seen in the 2009 OSL Finals, Jaedong vs Fantasy: '09 Batoo OSL Finals - Jaedong vs. Fantasy 1set 1/2 (Eng. Com.)  Hope you enjoy this great game of starcraft.  This whole series is really nice I find as it's a great and unique display of Zerg against Terran.  Perhaps you might also want to watch some Jaedong vs Bisu games as they both typically use a similar mixed unit style.  For example corsair reaver to dt drop, zealot drop, to high temps and archons.




Okay, something more important now.  I have created a folder in my Starcraft II Beta folder called TriggerLib and I have placed the .galaxy files into that folder.  Is that all I need to do to play with this new Protoss AI?  What should I select as the difficulty with the Launcher?  Thanks!

I wish you could learn to read. None of us said variety is a fail, we said the build AI does is fundamentally wrong and the rational is that you will not have enough money to sustain it. Let's look at FvJ; Fantasy was able to make variety of units because Fantasy knew he was in front (J had expanded + no troops). I didn't say diverse units are weak (harder to counter in SC2; read plz), it's just that you will be economically behind and your reward is having far less units but "variety". You probably noticed that Jaedong freaked out when he saw all kinds of units because he had not seen pro players do mass variety with one expansion, he wasn't prepared for it. You should ask yourself why so many pro players do not make so many variety with one expansion. Then ask yourself how many situations do you see coming out on top, having only one expansion heavily mixed units?

Quote from: sublime on March 04, 2010, 06:31:52 PM
really bad time to buy IMO. Basic economics man, supply and demand. when beta first came out, $300+ was the going rate for keys. Some even went as high as $400 or more. Once the second wave hit (blizzcon attendees), the # of auctions on ebay jumped threefold. I remember once seeing 180 results for "starcraft 2 beta". The price was going for barely over 200 and some people were making off with it for under 200. everyone knows another wave of invites will hit in a matter of days, the supply on ebay will be outrageously high again and thus the price will fall again. i was thinking about picking a key up off of ebay as well but it just isn't worth it at the current price. If 10,000 invites go out tomorrow you are going to be hurtin when you see the avg closing price on ebay. Just check the box for completed listings on the ebay search to get an idea of the avg closing price.

Plus if a multiplayer crack becomes available that will drop to like 2 bucks.

Quote from: sublime on March 04, 2010, 06:25:07 PM
didnt wc3 make a group of units selected together all move at the pace of the slowest unit? this means that regardless of units having different speeds they would all travel the same speed as a group.


i havent really paid attention to this being the case in SC2. i guess it's not. cant say i really care

yes, they removed in SC2 because they realised it was the wrong decision. Think about it, do you want to micro when you are not near enemy units or when you are getting slaughtered? This is a very logical choice. When you are retreating from battle select all and retreating will make every unit slow as mothership. This is LOL
When you are attacking an enemy base, you are not getting killed by enemy units, just micro them. It's a small price to pay for not having to micro a full scale retreat.

v0.1
Looked back at some replays as well and the AI has a horrible income:production ratio, why have millions of structures that you can't possibly produce from? The AI literally built almost every structure while on his natural expo, that's just not right. It's a classic noob mistake we have done at least once, but this AI does it over and over again.
Just need to fix that.

Quote from: Zalamander on March 04, 2010, 04:07:31 AM
AMAI2 v0.01 Released.
http://www.filefront.com/15733535/AMAI2_0.01.zip/

First simply prototype version, only contains a simple build for each race right now.

from readme:
Designer: Zalamander
Coder: AiAndy

v 0.01 (none cheating)

Experimental first release:
- Simple early mid game builds for all races.
- Enhanced harassment groups.


Next versions plans:
- Better harassment.
- Multiple builds based to counter enemy tactic.
- Better resource/expansion handling.

AMAI2 will have the same goal as our original AMAI for WarCraft 3.
To make an as challenging and unpredictable opponent as possible while remaining as fun as possible to play with and against.

This first version use the state system but enhanced with custom state switches no timer. Future version is going to have many different states to counter different strategies and do state switching based on enemy intell.


Reference:
Original AMAI for War3: http://www.hornes.pwp.blueyonder.co.uk/AMAI/

But... AMAI wasn't that good in wc3 T__T

step in the right direction though.

until blizzard says something it's not true.

Plus for blizzard games' release date the formula is
expected release date + 10 years

Yeah, played it. Good news and bad news
The Build
Protoss AI has variety in his/her arsenal but it seems to be Jack of All Trades and master of none. During the Battle I saw 3 phoenixes, 2 collosi and 2 immortals and good mix of stalkers and zealots. You would think that's a decent army but no. I was played Terran 3 times(teching (bio/mech) and rushing), I came to this conclusion. AI is just too poor economy wise. Even you don't harass, it will have a poor economy because it's building too much variety. Don't get me wrong, variety is good in SC2, harder to counter. However years of SC taught me this, go with one genre (e.g. bio, mech or air) after you expand and have spare resources then go variety.The fundamental build of this AI is wrong. Why would the AI go so many different kind of units with only one expansion? It does not have enough resource to sustain it. Yes, if you tech and never attack, the Protoss army becomes pretty large. Even then once you destroy their army AI doesn't have the resource to recover.

AI Micro
Slightly better than v5.5 and goose's AI. I didn't see anything terribly stupid from the AI.

Fun or Not?
I guess it's always going to come to this. The answer is sorta. The AI is like a noob SC player. This is a vast improvement compared to other AIs which feels like you are playing against a box. The AIs will not start moving into your base and back and base and back. Nor the AI will suicide its units. When reapers try to harass they will respond fairly quickly, some probes even moved towards me (didn't attack me though). Still the same problem as they move away as soon as reaper hides.Computer picks a battle then goes all in. I liked that part.

6.5/10 below expectation

Quote from: 1337 on March 02, 2010, 08:06:10 PM

After downloading and investigating some of the various AIs on this forum, I have determined that most of them focus on the "macro" aspect of gameplay, with build orders and such. I have been trying to unravel the "micro" AI to modify individual unit AIs. My first project is to code the High Templars to use Psionic Storm. Currently, the AI will build Templars if told to in the build order but they just run toward you dumbly without doing anything.


Here's my first hack, which compiles and runs but does nothing...


Replace the entire AIThinkHighTemplar method with the following in TactProtAI.galaxy:

Code Select Expand

order PsiStorm (int player, unit aiUnit, unitgroup scanGroup) {
    order ord;
//    aifilter filter;
    int psiStormRadius = 2;
//    unitgroup enemyGroup;
//    int enemyCount;
//    int i = 1;
//    region r;
//    unitfilter f;
//    unit enemy;
//    unit target;
//    unitgroup enemyAreaUnits;
//    region enemyArea;
//    int enemyAreaCount = 0;
//    int maxCount = 0;
   point castPoint;
   unitgroup stormGroup;


    ord = AITacticalOrder(player, aiUnit, c_AB_PsiStorm);
    if (ord == null) {
        return null;
    }


//   f = UnitFilterStr("-;Missile,Dead,Stasis");
//   r = RegionCircle(UnitGetPosition(aiUnit), AIRangeHighTemplar(player, aiUnit));
//   enemyGroup = UnitGroupAlliance(player, c_unitAllianceEnemy, r, null, c_noMaxCount);
//   enemyGroup = UnitGroupFilter(null, c_playerAny, enemyGroup, f, 0);
//   
//   enemyCount = UnitGroupCount(enemyGroup, c_unitCountAll);
//   while(i <= enemyCount)
//   {
//      enemy = UnitGroupUnit(enemyGroup,i);
//      enemyArea = RegionCircle(UnitGetPosition(enemy), psiStormRadius);
//      enemyAreaUnits = UnitGroupAlliance(player, c_unitAllianceEnemy, enemyArea, null, c_noMaxCount);
//      enemyAreaCount = UnitGroupCount(UnitGroupFilter(null, c_playerAny, enemyAreaUnits, f, 0),c_unitCountAll);
//      if (maxCount < enemyAreaCount)
//      {
//         target = enemy;
//         maxCount = enemyAreaCount;
//      }
//      i = i + 1;
//   }


   
//    OrderSetTargetPoint(ord,UnitGetPosition(target));


   stormGroup = AIEffectGroup(player, c_EF_PsiStormArea, scanGroup);
   castPoint = AIBestTargetPoint(
        stormGroup,
        3, // min hits
        40, // damage base
        3.0, // score
        psiStormRadius,
        UnitGetPosition(aiUnit),
        AIRangeHighTemplar(player, aiUnit),
        c_unitAttributeNone
    );


    if (castPoint == null) {
        return null;
    }
   
    OrderSetTargetPoint(ord, castPoint);
    if (!UnitOrderIsValid(aiUnit, ord)) {
        return null;
    }
   return ord;
}


void AIThinkHighTemplar (int player, unit aiUnit, unitgroup scanGroup) {
    marker mark;
    order ord;
    unitfilter f;
    region r;
    unitgroup enemyGroup;
    int enemyCount;


    if (AIEvalTacticalData(aiUnit, null)) {
        return;
    }
   
   if (UnitGetPropertyInt(aiUnit, c_unitPropEnergy, c_unitPropCurrent) < 75)
   {
      return;
   }


    //  If we already have a psi storm order, ignore any new orders since psi storm is more important.
    //  If we already have an PhaseShift order, ignore any new PhaseShift orders.
    if (UnitOrderHasAbil(aiUnit, c_AB_PhaseShift) || UnitOrderHasAbil(aiUnit, c_AB_PsiStorm)) {
        return;
    }
        f = UnitFilterStr("-;Missile,Dead,Stasis");
        r = RegionCircle(UnitGetPosition(aiUnit), AIRangeHighTemplar(player, aiUnit));
   enemyGroup = UnitGroupAlliance(player, c_unitAllianceEnemy, r, null, c_noMaxCount);
   enemyCount = UnitGroupCount(enemyGroup, c_unitCountAlive);
        if ((enemyCount < 2) ||
            AIAllyEnemyRatio(player, UnitGetPosition(aiUnit), f, AIRangeHighTemplar(player, aiUnit), c_MinThreshold) > c_EnemyMultiplierHighTemplar) {
            return;
        }


    ord = PsiStorm(player, aiUnit, scanGroup);
    if (ord != null) {
        AICast(aiUnit, ord, c_noMarker, c_castRetreat);
        return;
    }
}

I based my code off of AIThinkInfestor() and its associated InfestedTerrans() method, which work and *ARE USED IN GAME BY THE AI*. This is very important since Infested Terrans is one of the few abilities that the AI will actually use in game.
The code is also similar to AIThinkMothership() and the Vortex() method.
However, my code appears to have no effect. The commented lines in the PsiStorm() method represent my first attempt, and the uncommented lines are a different method of targetfinding, both of which do nothing.


In my build order I force the AI to build High Templars and research Psi Storm but he just never uses the ability.




My hypothesis as to why this code doesn't work is that it never runs. I found this gem of information in the TactTerrAI.galaxy (reaper AI code):

Code Select Expand


void AIThinkReaper (int player, unit aiUnit, unitgroup scanGroup) {
    // **Reaper AI reproduced in XML**
    //  Melee AI does not call this AIThink routine.


    //  Reaper AI is left in TactTerrAI to serve as an example of two different ways to write the same
    //  AI.  To see the xml equivalent of the Reaper AI, look in TacticalData.xml and TargetFindData.xml.
    //  In general, AI implemented in galaxy scripts will be slower than AI implemented in data.
    //  If it is convenient to express AI in data, one should do so.  In this case, converting reaper AI
    //  to xml resulted in a 2x performance gain.



It appears that there are two different sources for the unit AI, and which source is chosen is completely arbitrary. I have scoured the MPQ files for the locations of the AIThink calls and I can't find any. This means that the decision whether to call the AIThink routine or use the XML version is made in the main SCII executable.


Here's what really confuses me though - Psionic Storm is listed in the XML with all the proper fields in the AbilData, EffectData, TacticalData, TargetFindData, but it still is never used.
The aforementioned Reaper AI is also listed in the XML files, but Reapers don't have active abilities so...




I thought I would post what I have done so far up here and see if anyone else has had luck in getting the AI to use activated abilities. Also, if anyone knows where the AIThink routines are called from, that would be most helpful.[/code]

O_O did you check if the AIs research storm? Remember most AIs don't even research blink

Quote from: MiCrOMaN1 on March 03, 2010, 12:09:05 AM
Kinda way too easy for me...

that's prob because you played a lot of sc1 like me
People who aren't used to RTS will have problems. Unless you are saying you beat sc1 AI first time you played? XP

Quote from: KabouterNL on March 02, 2010, 08:48:25 AM
Just played the game, but you can just press a worker and then select the damaged building and he/she will start repairing it.

Second: I haven't found a function where you can see all your damaged building, just click in them.

東風春風: Unable to feel a thing even if the eastern wind were to blow
In other words, one is oblivious to what another had said.

I'm pretty sure there aren't any functions for it.
You can try to find it here http://galaxywiki.com/wiki/Main_Page
Full list of AI functions on that wiki.