Jump to content


Photo

Beta 7


80 replies to this topic

#21 Malkor

Malkor

    Eternity

  • Members
  • 375 posts
  • Projects:Loladins of Legend.
  •  Inanely inane inanities!

Posted 08 February 2006 - 03:32 AM

Okay. Let's say I wanted to start writing build orders for you. I'm looking at guardstrategyinfo_hard.ai, which I assume is also being run for the Insane difficulty. If not, can I specific one to be run at the insane difficulty?

Keep in mind I have never TOUCHED scripting before.

SquadLimits = 
		{
			standard = 
			{
				guard_squad_enginseer = 0,
				guard_squad_guardsmen = math.random( 5,7 ),
				guard_squad_command_squad = 1,
				guard_squad_kasrkin = math.random( 3,5 ),
				guard_squad_assassin = 1,
				guard_squad_ogryns = math.random( 4,6 ),
				guard_squad_psyker = 3,
				guard_squad_priest = 4,
				guard_squad_commissar = 4,
				guard_squad_sentinel = math.random( 3,5 ),
				guard_squad_hellhound = math.random( 3,5 ),
				guard_squad_basilisk = math.random( 3,5 ),
				guard_squad_chimera = 1,
				guard_squad_lemanruss = math.random( 4,6 ),
				guard_squad_baneblade = 1,
			}

Are these permanent values that cannot be altered later on? Or, is that what the math.random is for? If I changed this to...

		SquadLimits = 
		{
			standard = 
			{
				guard_squad_enginseer = 0,
				guard_squad_guardsmen = 2,
				guard_squad_command_squad = 0,
				guard_squad_kasrkin = math.random( 3,5 ),
				guard_squad_assassin = 1,
				guard_squad_ogryns = math.random( 4,6 ),
				guard_squad_psyker = 1,
				guard_squad_priest = 3,
				guard_squad_commissar = 0,
				guard_squad_sentinel = math.random( 1,2 ),
				guard_squad_hellhound = math.random( 1,2 ),
				guard_squad_basilisk = math.random( 1,3 ),
				guard_squad_chimera = 1,
				guard_squad_lemanruss = math.random( 4,6 ),
				guard_squad_baneblade = 1,
			}

Would it explode?

Also, I want these build orders to work specificially for insane.

For win percentages, unless I get triple rushed, I will usually beat my SC AI, and almost always beat the DoW AI on insane. Starcraft's standard AI I can take on 1v6 and win quite easily.

edit:

Also, Carriers in starcraft are hardly uber all-around units. They are extremely vulnerable to anti-air. If I know my comp is getting carriers I can destroy him in seconds.

Edited by Malkor, 08 February 2006 - 03:36 AM.


#22 thudo

thudo

    Wacko AI Guy!

  • Division Leaders
  • 12,164 posts
  • Location:Lemonville North, Canada
  • Projects:DoW AI Scripting Project
  • Division:DoW
  • Job:Division Leader

Posted 08 February 2006 - 03:49 AM

Those aren't buildorders - those are AI limits to how many units of that type can be fielded on-screen at once. That section above is about limiting those unit types in a game for the AI. To make em be used for insane only would take some minor coding although why bother? Use em for HARD and up for now. Its all the same.

Well.. judging by your win percentage even in Starcraft.. I think I'd have to pay both Larkin and Arkhan $50-75k per annum each to write thousands of lines of AI code to simulate the best possible synaptic player possible. Otherwise, no pro player will ever truely be quenched of their thirst. As I said: 90+% of those who grab our Advanced AI are low-to-average players. However, you guys (the pros) help us try to exceed our own grasps! :p Thats a good thing!

As for Carriers.. well.. bring along vehicles to eliminate the anti-air opposition. Further, you DO NOT cluster Carriers together, you micro-them staggered in formation to eliminate ANYTHING before you. Force-firing them on small enemy squads = suicide. If micro'ed they are unstoppable. Of course, bringing ground units to the offensive mix is a nice bonus. I loved playing as Protos. Ahh those were the days circa '99 and a little in 2000.
Advanced Skirmish AI Team Lead for the coolest Warhammer40k PC RTS out there:

Dawn of War Advanced AI Headquarters

Latest DoW Advanced AI Download!

#23 Malkor

Malkor

    Eternity

  • Members
  • 375 posts
  • Projects:Loladins of Legend.
  •  Inanely inane inanities!

Posted 08 February 2006 - 03:55 AM

My comp will get 200/200 carriers and I gaurantee you if I know he's getting them I will destroy them repeatedly. Same with a player. I play guys who are from PGL and stuff.

I've been working on Starcraft AI for five and a half years and I'll tell you you'd need to rewrite the game to ever have a hope of touching stuff related to micro and stuff. By the way, my Starcraft AI does in 100 lines what blizzard's do in 3000 lines. I've come a long way in those years but a good AI opponent in starcraft is still an unreachable dream. That's why I'm here. To help make my DoW dream come true.

Also, I do recognize the audience this AI is directed for, that's why I want to do this for the insane AI specificially. I want to create an opponent hardcore players like myself can fight and fear. There are still limitations, but I believe I can improve the Insane AI to a point where it is far more challenging than it currently is.

I guess I should make a new thread concerning my inquiries about modifying the build orders. I'll do that later.

#24 thudo

thudo

    Wacko AI Guy!

  • Division Leaders
  • 12,164 posts
  • Location:Lemonville North, Canada
  • Projects:DoW AI Scripting Project
  • Division:DoW
  • Job:Division Leader

Posted 08 February 2006 - 04:08 AM

Re: Carriers - I was referring to a pro (microing god) using them staggered and perfectly microed. No AI for 10s of years will do that as well as a cunning tactically-superior human.

Btw, what specifically do you want the DoW/WA AI on insane or anything above HARD to achieve? Please post your proposal(s) in point-form and we'll try to make it happen.
Advanced Skirmish AI Team Lead for the coolest Warhammer40k PC RTS out there:

Dawn of War Advanced AI Headquarters

Latest DoW Advanced AI Download!

#25 Zenoth

Zenoth

    title available

  • Members
  • 469 posts

Posted 08 February 2006 - 04:27 AM

Just played in Crossroads (3 Vs 3), at Hard.

Very impressed.

And guess what, no no ... it's not a crash, but much better, I caught the Orks just before the game ended (before me and my allies won) that just trained a Squiggoth ! :p

But too little too late ... poor Squiggoth died faster than he appeared. I do not blame the coding for that, or the build orders, or teching manner, but simply because the maps only had two Relics for six players, and sometimes the factions "exchanged" it. One lost it, than another took it back, etc. None of the factions really had a Relic for themselves for the whole game. I myself couldn't get the nearest one until near the end, and didn't had enough time to train an Avatar before we destroyed the Orks.

Here's a picture ...

Posted Image

And, technically speaking, I believe that we could consider this Beta build the last one. Unless someone eventually find something important enough to work on before a final build. But so far ... I noticed nothing "critical" to fix. The only thing I can think of right now is the structures getting destroyed, and worker units constantly trying to build them back in a "destroyed/build back" loop. But I have yet to see that again in Beta 7.

So far so good. Superb work guys.

Of course I can't comment on the final release, as in I simply don't know what Larkin/Thudo/Arkhan are considering right now. But personally, if I may advance my stance on the current Beta build, then as I said, I believe we should be ready for release.

Edited by Zenoth, 08 February 2006 - 04:29 AM.


#26 ThetaOrion

ThetaOrion

    title available

  • Members
  • 676 posts

Posted 08 February 2006 - 04:46 AM

The AI had 3 commisars but didn't so much as attach them. If he's going to get them, he may as well attach them.


I was going to mention that in my previous game, that the IG Commissars weren't attaching, but I wanted to verify it. I saw a Priest or two attached to Ogryn or Kasyrkin or Guardsmen, though. The Priest was attaching up.

When Arkhan changes the Commissars into Tier 3 units, I imagine that they are somehow skipping their attachment code. That would be my guess.

--

SECOND MESSAGE:

Getting a BaneBlade, like getting a Squigg or Avatar is hard. How do you expect the AI to quickly tech to it if your wearing him down.


In my last game, this argument applied very well to the Orks and the Squigg. All the fighting was done in my Ork AI's territory, and they really didn't have any chance to get a Squigg.

But, the logic falls down in one respect.

I got a BloodThirster at the exact same time that the SM showed up with a LandRaider. No excessive fighting done in SM, IG, or my Chaos territory, but the IG didn't get a BaneBlade at the same time that we other untouched factions got our Avatars or Titans.

Why would I as Chaos and the Space Marines be able to get our Avatars so much faster than the IG? I wasn't wearing the IG down by the time the Space Marines got their LandRaider. Why do the Space Marines seem to tech up twice as fast as everyone else?

Edited by ThetaOrion, 08 February 2006 - 04:47 AM.


#27 thudo

thudo

    Wacko AI Guy!

  • Division Leaders
  • 12,164 posts
  • Location:Lemonville North, Canada
  • Projects:DoW AI Scripting Project
  • Division:DoW
  • Job:Division Leader

Posted 08 February 2006 - 04:55 AM

Why do the Space Marines seem to tech up twice as fast as everyone else?

BING-FRICK'N-OOO ! Now you figured it out! :p Surprised it took you this long to understand Dread-spam and why SM and sometimes Eldar/CSM can tech extremely fast to their Tier2 units and thus are incredibly hard to stop sometimes. Thats why in WA its easy to play as SM to get to Dread-spam.

Btw.. IG is VERY hard to get to their uber (lots of addons and buildings to get to) compared to the other factions. Even Ork take time...
Advanced Skirmish AI Team Lead for the coolest Warhammer40k PC RTS out there:

Dawn of War Advanced AI Headquarters

Latest DoW Advanced AI Download!

#28 ThetaOrion

ThetaOrion

    title available

  • Members
  • 676 posts

Posted 08 February 2006 - 05:01 AM

THIRD MESSAGE:

They cap stuff.


Actually, this is probably one of my biggest complaints about the Beta 7 Orks. They don't fight to cap if the strategic point is contested by the Space Marines, and the Orks don't recap. We were having problems with this and the Space Marines in previous Betas, and now the shoe is on the other foot, or so it appears. The Orks won't move to capture an empty or decapped strategic point if they have lost the previous attempt to take the resource point in their territory. And, the Orks also won't decap an enemy resource point in their own territory, even if it doesn't have a listening post and should be easy to decap.

Ork AI should be moving out and smashing down nearby SM listening posts, but in Beta 7, the Ork AI is afraid of the Space Marines and essentially refuses to do any destructive action at all in Space Marine territory.

On Mountain Trails, the Orks are walking right through Space Marine territory in order to get to the Imperial Guard, but the Orks don't stop to fight until they get to Imperial Guard territory. The Beta 7 (and Beta 6) Orks are afraid of the Space Marines.

That's my observation.

Edited by ThetaOrion, 08 February 2006 - 05:09 AM.


#29 Malkor

Malkor

    Eternity

  • Members
  • 375 posts
  • Projects:Loladins of Legend.
  •  Inanely inane inanities!

Posted 08 February 2006 - 05:05 AM

Thudo>

Basically what I wanted to do was try out the builds, namely for the IG, I've been talking about. That is, super fast tech to Hellhounds, then straight to leman russ tanks and the baneblade. No commisars, nothing like that, just straight up hard tech.

I was also going to take the opportunity to try out a number of other things, like early warp spiders + vypers, early mass land speeders... Basically I was going to see just how fast I could get the vehicle rushes happening, and have the AI rush as fast as possible to the heavy hitting tier4 units, especially Avatars and Squiggoths.

These builds would have been built SPECIFICIALLY for the 4v4 + FFA games I play, and would be limited to the "insane" difficulty, as by the time I'd have them working like I wanted, they would be VERY hard to beat. Brutallity factor at it's finest. I could give you a build order kinda thing if you want, or the unit combos I was going to try.

#30 ThetaOrion

ThetaOrion

    title available

  • Members
  • 676 posts

Posted 08 February 2006 - 05:16 AM

BING-FRICK'N-OOO ! Now you figured it out! Surprised it took you this long to understand Dread-spam and why SM and sometimes Eldar/CSM can tech extremely fast to their Tier2 units and thus are incredibly hard to stop sometimes. Thats why in WA its easy to play as SM to get to Dread-spam.

Btw.. IG is VERY hard to get to their uber (lots of addons and buildings to get to) compared to the other factions. Even Ork take time...


So, we are going to be blaming some of this on balance then?

The original game isn't balanced enough to expect each of the factions to be able to put out is Uber or Avatar at the same time, all other things being equal.

And, since you are in a sense classifying it as a balance teching issue, then all other things aren't equal. :p

Just as long as we get our stories straight before the interrogation starts. ;)

--

Is Relic ever going to balance this thing?

#31 thudo

thudo

    Wacko AI Guy!

  • Division Leaders
  • 12,164 posts
  • Location:Lemonville North, Canada
  • Projects:DoW AI Scripting Project
  • Division:DoW
  • Job:Division Leader

Posted 08 February 2006 - 05:30 AM

I blame it totally on WA = console game. SM totally and without a doubt atypifies the FAST ACTION.. MINIMAL TECH to GOOD SPAM faction. SM have it easy - some few steps to get to good Tier2 units. Other factions have to regrettably mimic its buildorder structure all due to f**king Dreads. If the Dreads weren't such b*atches in #s or even one that early can own you bad, then it wouldn't be so bad for other factions teching up. Ork and IG have it bad, period compared to Dreads, Defilers, and Wraithlords. Only thing helping IG are teching to Sentinels FAST as SH*T!

Anyway.. sorry for what appears to be general grumpiness of late. I just keep seeing so much feedback its a challenge to please everything and people's left testicle. <oh my>.

I need to better create a "priority list" based on POINT FORM observations. I cannot write paragraphs. Things in development are about short, concise, easy-to-read blurbs which can be prioritized.

Btw, Malkor.. Yer joking right?

Do you think for one gazilla-second we couldn't have done that? Its a joke of humanity to just let the AI "spam to its heart's content" while just teching with the most basic items. All we would do would just call various squads quantities, build addons or prerequisites required to tech to em, and then research no luxuries. You want that? If so.. how utterly boring. Thats just a massing spam-orgy of units always attacking. Of course we could do that and show you how. Not that its a bad idea.. What you ask is for the ELITE of the ELITE players: those who like their wips and leather and hot irons always up their arses. :p A totally custom dungeon torture! hehe

Edited by thudo, 08 February 2006 - 05:36 AM.

Advanced Skirmish AI Team Lead for the coolest Warhammer40k PC RTS out there:

Dawn of War Advanced AI Headquarters

Latest DoW Advanced AI Download!

#32 ThetaOrion

ThetaOrion

    title available

  • Members
  • 676 posts

Posted 08 February 2006 - 05:30 AM

But so far ... I noticed nothing "critical" to fix.


Well, I'm not ready to call it quits or label it finished.

I haven't yet had time to test Beta 5, Beta 6, and Beta 7 on the Castle Assault Map to see if the Raptors and the Assault Space Marines are flying again.

I remember back in Beta 4 the last time I had a chance to make the test, the Raptors and ASM weren't flying, unlike what they were doing in all the previous Betas. Gotta know if that has been fixed before I start feeling comfortable with discussion about release dates.

It would be nice to make sure that the IG and Ork AI allies aren't retreating and turtling on the Desert Gauntlet, as well. Who knows if all the annoying issues are gone and staying gone?

Whatever ends up being the final beta, we all need to make sure to give it a long shakedown on many different maps to make sure it isn't glitching someplace, in my opinion.

Of course, I guess there are hotfixes, but that's kind of messy and confusing for the end user.

#33 ThetaOrion

ThetaOrion

    title available

  • Members
  • 676 posts

Posted 08 February 2006 - 05:40 AM

I need to better create a "priority list" based on POINT FORM observations.


1) Figure out why Beta 7 IG Commissars are not attaching and make them attach.
2) Make Beta 7 Orks Decap and Recap Strategic Points, as well as contest resource points.
3) Do whatever else will put a smile on Thudo's face.

#34 thudo

thudo

    Wacko AI Guy!

  • Division Leaders
  • 12,164 posts
  • Location:Lemonville North, Canada
  • Projects:DoW AI Scripting Project
  • Division:DoW
  • Job:Division Leader

Posted 08 February 2006 - 05:51 AM

Believe I found the commissar problem.. the squad was incorrectly named..

local num_leader = self:CountSquads("guard_squad_commisar") + 
			   self:CountSquads("guard_squad_psyker")
			if num_leader >= num_attach then
			   squad_mod = self.squad_mod[i][2] --set modifier to 0
			end
Should be:
local num_leader = self:CountSquads("guard_squad_commissar") + 
			   self:CountSquads("guard_squad_psyker")
			if num_leader >= num_attach then
			   squad_mod = self.squad_mod[i][2] --set modifier to 0
			end
Odd that was ever changed? hehe.. Hey.. scripter make mistakes!

Also might have found out why the Orks are NOT throwing themselves more into combat. Arkhan can back me up on this but I changed there TurretPower to 250 (was 150 - the lowest of all factions):

if (sType == "TurretPower") then
		return 250
	elseif (sType == "SupportAllyFactor") then
		return 9 / 4

I also ensured ALL factions have a minimum of 8 / 4 support factor so they would more readily assist their allies. Eldar, CSM, and SM have the highest supportative values so figured why not boost the IG/Ork a little. Will test.. Arkhan will also need to discuss this..
Advanced Skirmish AI Team Lead for the coolest Warhammer40k PC RTS out there:

Dawn of War Advanced AI Headquarters

Latest DoW Advanced AI Download!

#35 Malkor

Malkor

    Eternity

  • Members
  • 375 posts
  • Projects:Loladins of Legend.
  •  Inanely inane inanities!

Posted 08 February 2006 - 06:38 AM

Thudo>

To clearify, I'd be having the AI do exactly what I do to win FFA and 4v4 situations.

Perhaps the easiest way to explain this kind of a build is to show you my carrier rush script, straight from my AI editor.

define_max 40 probe
define_max 1 zealot
define_max 1 dragoon
define_max 1 high_templar
define_max 1 dark_templar
define_max 1 dark_archon
define_max 1 archon
define_max 1 scout
define_max 1 corsair
define_max 20 carrier
define_max 6 observer
define_max 2 shuttle
define_max 1 reaver
define_max 2 arbiter
multirun money
build 1 nexus 80
build 1 assimilator 80
build 3 pylon 150
wait 800
build 1 gateway 150
defensebuild_gg 1 zealot
defenseuse_gg 1 zealot
build 2 cybernetics_core 150
build 1 forge 150
upgrade 1 p_plasma_shield 30
upgrade 1 p_plating 30
upgrade 1 p_air_weapon 30
build 3 stargate 150
build 1 photon_cannon 80
build 1 fleet_beacon 150
train 3 carrier
upgrade 1 carrier_capacity 30
attack_add 3 carrier
attack_prepare
wait 1600
attack_do
attack_clear
expand 1 exp
multirun probemelikeananimal
build 1 robotics_facility 80
upgrade 2 p_plasma_shield 30
train 6 carrier
build 1 citadel_of_adun 80
train 2 shuttle
defensebuild_ag 1 carrier
defenseuse_ag 1 carrier
defensebuild_aa 1 carrier
defenseuse_aa 1 carrier
defenseuse_gg 1 carrier
defensebuild_aa 1 corsair
defenseuse_aa 1 corsair
defensebuild_aa 1 scout
defenseuse_aa 1 scout
defenseuse_ga 1 scout
defenseuse_gg 1 scout
defensebuild_ag 1 scout
defenseuse_ag 1 scout
build 1 robotics_support_bay 80
build 4 photon_cannon 80
build 1 observatory 80
train 4 observer
expand 2 exp
build 1 templar_archives 80
upgrade 2 p_plating 30
build 8 photon_cannon 50
upgrade 1 observer_speed 30
upgrade 2 p_air_weapon 30
tech disruption_web 30
build 8 stargate 80
upgrade 1 scout_speed 30
build 1 arbiter_tribunal 80
tech recall 30
upgrade 3 p_plating 30
upgrade 3 p_air_weapon 30
upgrade 3 p_plasma_shield 30
:carrierfallback
define_max 20 carrier
define_max 1 scout
define_max 1 corsair

:carrierloop
expand 99 exp
train 20 carrier
train 2 arbiter
attack_add 2 arbiter
attack_add 20 carrier
attack_prepare
wait 2000
attack_do
attack_clear
wait 750
;   random_jump 25 skootfallback
goto carrierloop


Now. The numbers usually following things is a priority. This script will do everything from start from scratch to eventually encompass the entire map with defended expansions, manage workers, shuttles, and attack with huge amounts of fully upgraded carriers. This is all I have to work with when creating a Starcraft AI script.

Think of, for example, my proposed Land Speeder rush in the same way I did this.

The SM would rush ASAP to dual cults with FC and two scout marine squads with snipers. Then, it would pump to 10/10 speeders and attack the enemy WITH ONLY THE SPEEDERS. Then, it would move on to predators and land RAIDERS, and terminators, encompassing techs and upgrades for those units as it gets them.

And, no, I don't think you couldn't of done this. And I know why you didn't, which is why I want to now do it. These builds aren't designed for 1v1. They can used in 1v1, but that's risky as hell, even for an Insane comp. These are FFA/4v4 builds. They are NOT general combat builds. They're for VARIETY. One game the comp could go early speeders, next dreads, next straight up infantry. It's the unexpected factor, that mystery factor, that is what will set this up well in a large FFA. You won't know who you're up against, or what strategy they will go, until you face them.

Edited by Malkor, 08 February 2006 - 06:40 AM.


#36 Zenoth

Zenoth

    title available

  • Members
  • 469 posts

Posted 08 February 2006 - 07:06 AM

I got something to suggest though ...

To Arkhan/Larkin/Thudo.

I'm not aware right now (and you guys will tell me if that's the case) if, within the scripts/coding, there's any sort of priority, for attacking structures.

If there is indeed such a thing, then we might consider revising it.

I'll give an example.

It happens quite often to see some infantry or vehicles attacking "less important" (still not useless though) structures, once a base is under attack. I'm making reference to the main base, which is usually located at the starting point.

And that, I saw it from the very first Beta I tested up until now.

Let's pretend that a group of infantry ... say a Space Marines Chapter, enters a "main base", obviously to attack it, then they will attack some random Plasma Generator. Then, sometimes, they will suddenly change their mind and go for a Turret, if any. Then, no, they prefer to go for the Stronghold.

Or another example I can think of right now, happened in my last Mountain Trails test (in Beta 6), I was under attack by a single group of Chaos Raptors. I saw them enter my base. All my infantry was busy somewhere else, and my base was empty of military units. I only had three workers. I had one Turret guarding the Stronghold. I clearly saw them going for one of my worker units. They destroyed the poor robot, and then ignored the second and third one. They proceeded to attack the Stronghold for just a few seconds. All that while the Turret did its best against them (they never tried to go out of the Turret's range). That group lost two soldiers by then, and they kept getting reinforcements (adding units to the same group). My own infantry groups were coming back to my base to get rid of them. Eventually, the Raptors jumped for the Turret, they hit it a few times (and they could have destroyed it so easily), then they decided to kill my remaining workers. So what I did was to take both of them, and ordered them to a random location far outside my main base.

Conclusion ... the Raptors followed them both. My infantry units arrived. I got new workers. I repaired everything that was damaged. But I never lost my Turret, which should have clearly been a priority. And then those workers running for their lives were destroyed, and I never heard news from those Raptors again. They never came back after killing my remaining workers.

Perhaps other testers also saw such behavior. But in my case I can say it's not rare. Not everytime though, but it does happen. Such a behavior can make the difference between loosing and wining a game especially during the first moments in Tirer 1.

Edited by Zenoth, 08 February 2006 - 07:10 AM.


#37 ThetaOrion

ThetaOrion

    title available

  • Members
  • 676 posts

Posted 08 February 2006 - 07:33 AM

Zenoth:

I have seen the enemy AI get off task when I put out a new worker or when I get troops back to my base. I'm kind of glad that I can draw them off task and prevent them from hammering my Stronghold into oblivion.

I don't know how this could be improved or if it should be.

I have taken an inferior set of SM scouts and danced them around one of my firing listening posts until the pursuing Chaos Raptors are dead. It's nice from time to time to be able to lure the AI into a trap or to knock them off task. I think this is one of those situations where the AI team is not going to be able to please everyone in every situation.

Yes, it would be logical for the enemy AI to finish off a turret once they start destroying it, but the AI is programmed to put Builders (and probably Strongholds and Barracks and Power Generators) on a higher priority. They also tend to go for the nearest thing at hand after chasing and killing a worker.

I guess my expectations here regarding AI priorities aren't as high, or I'm glad that they can be drawn off task from time to time.

And, it's all made even more muddy by the fact that in the last couple of games the Chaos AI came straight in and wiped out my turrets completely right from the start, and then started on my Stronghold until it was gone. Sometimes the AI works too efficiently. <grin>

--

Malkor:

I know that a lot of people on the Relic Forums have mentioned that they don't like the IG, and I know that for myself that it usually takes me twice as long to win while playing as the IG. I can see why Malkor and others would want the IG to be better.

But, I have a question for Malkor. Is he and has he been completely satisfied with the performance of the Ork AI since 1.41 of WA came out? Maybe the Orks do better in FFA and/or on larger open maps?

--

ThetaOrion:

I need to see or determine if the Eldar AI is still working fine and doing fine. They seemed to be really aggressive but rather weak AI allies on the Beta 6 and Beta 7 Mountain Trails tests. I assume that they are just Eldar being Eldar -- not too hot and not too cold. I imagine that everything is 'normal' with the Eldar, but I don't yet have enough evidence to back up my assumption.

I also want to test and see if the STANDARD setting is working right. So many things to try, and so little time.

--

And, thank you Thudo for looking into all of this Beta 7 stuff that we are bringing up.

#38 ThetaOrion

ThetaOrion

    title available

  • Members
  • 676 posts

Posted 08 February 2006 - 08:15 AM

I don't care what people say, the Eldar were made for killing Chaos. The Orks tend to give the Eldar more grief than the Chaos do.

Mountain Trails:

2Eldar vs 2Chaos at HARD setting Beta 7. We Eldar beat the brains out of the Chaos, 22 minutes of clock time.

The Eldar are working just fine in Beta 7, from what I can see.

Obviously, the game was too short for anyone to get their Uber, including me.

I like playing as the Eldar.

Yep, everything back to normal.

Thudo done went and made dem Eldar good.

Edited by ThetaOrion, 08 February 2006 - 08:18 AM.


#39 Zenoth

Zenoth

    title available

  • Members
  • 469 posts

Posted 08 February 2006 - 09:04 AM

I see what you mean Theta, and I do agree.

But technically speaking, such behavior simply makes no sense. There's nothing logic about attacking a Generator, while a Turret makes members of the attacking group history. There should be some sort of priority if we seek to provide some sense in the A.I behavior.

I do not mind to lure the A.I around a firing Listening Post indeed ... but you'd never see a human fall into such a humiliating trap (well not me anyway). And of course the A.I can't get a human brain to behave the same way so ... but as far as scripting goes, it's possible to "fix" such a situation, simply by coding in some priority list.

A Raptor squad approaches a LP with a worker idle by its side.

Priority says (let's pretend): #1 -Turret- #2 -LP- #3 -worker unit(s)

Dynamically it could go like: (start search priority = x (x being a vairant for Turret) x = 1, destroy Turret, search for LP, LP = 1, destroy LP, search for worker, worker = 1, destroy worker) ... of course I know pretty much nothing about coding ... but you get my point. Technically I know it can be done, since I had to test an "A.I priority behavior" back when I was a beta tester for Ballbarian_SW's A.I for Galactic Battlegrounds.

However there is a downside to priority lists.

If in the situation above a group of infantry arrives (enemies to the Raptors) in the area, then the Raptors will ignore them until the "priority queue" ends, then, they will proceed for the enemy infantry, even if "infantry units" was of higher priority than the worker they were attacking when the enemy group arrived.

Well ... I'm sure Arkhan will have something to say about this.

But the bottom of the whole subject I bring here is simply that when a single group attacks a base with multiple struture types then they seem to attack randomly. And as I said, it doesn't happen everytime. When it doesn't they do attack things in a logical manner. So since it's not a constant issue, where it'd happen everytime with obvious lack of sense in their tactics then i'd say it'd be something we would have to look at.

But that's not the case. It seems to be something that happens only a few times. It's not "critical", but still quite obvious when you see it yourself.

Edited by Zenoth, 08 February 2006 - 09:16 AM.


#40 LarkinVB

LarkinVB

    title available

  • Members
  • 1,488 posts

Posted 08 February 2006 - 10:24 AM

Just two notes.

1. Thudo, the commissar code you corrected is not responsible for the attaches. This is handled in the commissar tactic file. It is responsibe to check that the number of commissars fits the number of regular squads they can attach to. Therefore it might be possible the AI did build more commissars than it had squads.

2. I would love to see better AI code to fight for initial points. This is one main weakness as a player just has to contest them a bit and the AI will give up soon, falling behind with their economy. Instead of gathering its attack force (commanders are usually ideling around) it should protect their initial cappers.



Reply to this topic



  


0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users