Jump to content


Photo

Skirmish AI 2.5 Beta 4 - Post Comments In Thread!


34 replies to this topic

#1 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 31 July 2007 - 03:15 PM

Changes
======
- Mostly performance tweaks.

Go at it, testers! :)

2.50b4 Changes
==========
- Removed Defense print command (showed in console)
- Reduced IG Psyker rating
Advanced Skirmish AI Team Lead for the coolest Warhammer40k PC RTS out there:

Dawn of War Advanced AI Headquarters

Latest DoW Advanced AI Download!

#2 Inquisitor

Inquisitor
  • Members
  • 121 posts
  • Location:Ordo Malleus

Posted 31 July 2007 - 03:22 PM

Nice... I'll do some intensive testing tonight. I hope to do some online mp tests as well.
The Emperor knows, the Emperor is watching.
Want to zoom out further? Try this camera mod.

#3 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 31 July 2007 - 03:26 PM

Excellent! Some online testing would be a great benefit!
Advanced Skirmish AI Team Lead for the coolest Warhammer40k PC RTS out there:

Dawn of War Advanced AI Headquarters

Latest DoW Advanced AI Download!

#4 Inquisitor

Inquisitor
  • Members
  • 121 posts
  • Location:Ordo Malleus

Posted 31 July 2007 - 04:55 PM

@ArkhanTheBlack: Did your positioning bug fix make it into b3? If it did, I'm sad to report that it didn't work.

AI Version: 2.5b3
Matches: 1v1 six times
Factions: Necron AI vs IG AI on the same map, but different starting positions
Game Modes Used: Annihilate
Map: Valley of Khorne
AI Difficulty: Harder
Resources Rate: Normal
Resources Sharing: Off
Starting Location: Fixed
Winner: IG 3/3 on position 2 and Necrons 3/3 on position 2

Comments: I wanted to see if the position related bug was fixed in b3.

Observations of interest:
  • Most importantly: IG won all games on position 2 and Necrons won all games on position 2.
  • In two of the games with Necrons on pos 1, the NL appeared to be stuck, when he was initially built. He didn't move and only engaged the enemy on ranged until he died and was rebuilt and then there was no problem.
  • In one game with Necrons on pos 1 the Necron had two upgraded Monoliths. Why use 300 power on a useless upgrade when army strength is low?
  • In game 3 with Necrons in pos 1, IG built a Mars pattern right in the entrance to the Necron base! Not a problem because the Necron had almost no army, but a bit risky anyway.
  • Necrons still like to upgrade all LPs. This is totally unnecessary and a complete waste of power, and should have low priority unless near to enemy base. They are nice in the defense, but they yield no economic benefit for the Necron, so they are pure expense.
  • Very nice use of bunkering by IG in most of the games. It is soooo nice to see the AI do this! :)
Suggestions:
We have to fix the positioning bug before release, if at all possible. Testing the AI and not knowing wether it performs badly because of it's position (on mirror maps) is almost pointless. I suggest that the bug must reside in a single function that has to do with determining the positioning of friendly buildings and/or the enemy. ArkhanTheBlack mentioned the friendly base array contained wild pointers... Maybe we could somehow validate these positions at startup? Or maybe another function is writing to it (by mistake)? I'd be willing to help out if you could instruct me how to debug the code (just send me a PM with links or something).

Replays: Position wins the game (6 replays on Valley of Khorne).

Remember: I'm just trying to help out here. My role as a tester is to find bugs and demonstrate how to reproduce them.

Edited by Inquisitor, 31 July 2007 - 05:06 PM.

The Emperor knows, the Emperor is watching.
Want to zoom out further? Try this camera mod.

#5 ArkhanTheBlack

ArkhanTheBlack

    title available

  • Members
  • 814 posts

Posted 31 July 2007 - 07:40 PM

@ArkhanTheBlack: Did your positioning bug fix make it into b3? If it did, I'm sad to report that it didn't work.

Well, the fix doesn't equalize start positions, it fixes a problem with invalid move positions resulting in chaotic unit movements. That however doesn't mean that there isn't still a tactical advantage of a specific position which can lead to an always win or loose situation. The question is not IF they win or loose but HOW they win or loose. The moving problems don't occur on every map. I've seen them in Blood River, but not in Battle Marshes, Kasyr or Mountain Trail.

#6 ArkhanTheBlack

ArkhanTheBlack

    title available

  • Members
  • 814 posts

Posted 01 August 2007 - 06:16 PM

I tried two games with IG vs Necons on Valley of Khorne myself and Necrons won on both positions. But I do think that position 2 has a slight advantage since it took a lot longer for Necrons to win on position 1.
More importantly, there were no bugged movements.

In one game with Necrons on pos 1 the Necron had two upgraded Monoliths. Why use 300 power on a useless upgrade when army strength is low?

This is to avoid that they get into a tech trap. Necrons are severely bugged, and they sometimes try to upgrade the wrong monolith it there is more than one available.

#7 LarkinVB

LarkinVB

    title available

  • Members
  • 1,488 posts

Posted 01 August 2007 - 09:46 PM

I watched some more battles. All in all it seems fine but I find that armies are moving too much through enemy fire. If the AI has some decent ranged army it is moving too much in combat. Instead of shooting it out is seems to get commands to move to another position. Perhaps it is a good idea to double check move commands in attack strategies which force squads to move instead of fighting and should be replaced by attack moves.

Tip for IG : Force priests before psykers to attach them to guard/ogryns.

Edited by LarkinVB, 01 August 2007 - 09:46 PM.


#8 ArkhanTheBlack

ArkhanTheBlack

    title available

  • Members
  • 814 posts

Posted 02 August 2007 - 09:35 PM

I watched some more battles. All in all it seems fine but I find that armies are moving too much through enemy fire. If the AI has some decent ranged army it is moving too much in combat. Instead of shooting it out is seems to get commands to move to another position. Perhaps it is a good idea to double check move commands in attack strategies which force squads to move instead of fighting and should be replaced by attack moves.

The move commands are fine, it's more a conceptual problem. I might have some ideas to improve this behaviour, but I can't afford the time to do that at the moment.

#9 LarkinVB

LarkinVB

    title available

  • Members
  • 1,488 posts

Posted 04 August 2007 - 11:41 AM

There is still stance code in TauInfantryTactic:Update()

function TauInfantryTactic:Update()

	 ......

	-- Set stand and ground stance for ranged troops
	if (self.squad_ai:IsRanged() and self.squad_ai:GetStance() ~= SquadAI.STANCE_StandGround) then
		self.squad_ai:DoSetStance(SquadAI.STANCE_StandGround)
	end
end

Isn't this interfering with Tactic:UpdateStance() as both are called now and isn't it better to put it into TauInfantryTactic:UpdateStance() instead to override ?

#10 ArkhanTheBlack

ArkhanTheBlack

    title available

  • Members
  • 814 posts

Posted 04 August 2007 - 12:35 PM

I don't know what AI version you have, but in my 2.5 AI, there's no TauInfantryTactic:Update() method anymore. And I made those changes already in beta 1.

#11 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 04 August 2007 - 03:20 PM

v2.50b4 is up. Please do an overall test and we'll get this baby out the door!
Advanced Skirmish AI Team Lead for the coolest Warhammer40k PC RTS out there:

Dawn of War Advanced AI Headquarters

Latest DoW Advanced AI Download!

#12 Inquisitor

Inquisitor
  • Members
  • 121 posts
  • Location:Ordo Malleus

Posted 04 August 2007 - 05:24 PM

Good work, guys. In my tests I have found nothing besides what I already mentioned, so it's all super-green imo.

Here's a hard won win by me on Valley of Khorne (pos 1) on harder:
Necron Valley

I have to say that I really had to sharpen up to win, even though this map would be considered a Necron auto-win vs. IG by most pros, due to the many near-base LPs and small map size. The IG AI on pos 2 on this map is just mean! The skirmish AI coders have done a really good job with this release. :)

I tried two games with IG vs Necons on Valley of Khorne myself and Necrons won on both positions.

Amazing! How on earth? Here's 10 out of 10 games won by IG on pos 2 with 2.5b4:
IG pwns Necrons
Nerf IG! :p
The Emperor knows, the Emperor is watching.
Want to zoom out further? Try this camera mod.

#13 LarkinVB

LarkinVB

    title available

  • Members
  • 1,488 posts

Posted 04 August 2007 - 07:47 PM

I don't know what AI version you have, but in my 2.5 AI, there's no TauInfantryTactic:Update() method anymore. And I made those changes already in beta 1.


It was the dcpro build based on 2.5b3 made my Thudo. Seems he missed these changes.

Arkhan, please note that IsMoving() is using
return (bMoveState and distance_sqr(self.m_vLastPosition, self.m_vCurrentPosition) > 0)

Slight low level AI squad repositioning will be reported as move though the squad is not really moving at all (resulting in toggling on FoF for example). I suggest checking for > 4 instead of > 0. I used and debugged same code for scarab builders and found that they 'move' up to 4 distance while building a generator. Since this check is made every 2 seconds 4 should be easily exceeded if the squad is 'really' moving for 2 seconds.

Edited by LarkinVB, 04 August 2007 - 07:49 PM.


#14 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 04 August 2007 - 09:42 PM

Ouch yes sorry.. not sure how I missed that being thorough. Larkin.. trust you will correct that for the next DowPro build. :)
Advanced Skirmish AI Team Lead for the coolest Warhammer40k PC RTS out there:

Dawn of War Advanced AI Headquarters

Latest DoW Advanced AI Download!

#15 ArkhanTheBlack

ArkhanTheBlack

    title available

  • Members
  • 814 posts

Posted 05 August 2007 - 09:55 AM

Slight low level AI squad repositioning will be reported as move though the squad is not really moving at all (resulting in toggling on FoF for example). I suggest checking for > 4 instead of > 0. I used and debugged same code for scarab builders and found that they 'move' up to 4 distance while building a generator. Since this check is made every 2 seconds 4 should be easily exceeded if the squad is 'really' moving for 2 seconds.

Done!


Amazing! How on earth? Here's 10 out of 10 games won by IG on pos 2 with 2.5b4:

I always play with heroes. Maybe that's the difference...

#16 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 05 August 2007 - 01:48 PM

Same.. always play with HEROES and AI donations.. at least for single play. For multi.. we need more testing online with those game modes enabled.
Advanced Skirmish AI Team Lead for the coolest Warhammer40k PC RTS out there:

Dawn of War Advanced AI Headquarters

Latest DoW Advanced AI Download!

#17 Inquisitor

Inquisitor
  • Members
  • 121 posts
  • Location:Ordo Malleus

Posted 05 August 2007 - 05:57 PM

Same.. always play with HEROES and AI donations.. at least for single play. For multi.. we need more testing online with those game modes enabled.

I can verify that Heroes works flawlessly online. We have played many games with it, both on Hamachi and via the online lobby. I haven't tested AI donations, though.
The Emperor knows, the Emperor is watching.
Want to zoom out further? Try this camera mod.

#18 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 05 August 2007 - 07:46 PM

Thanks Inquisitor.. AI Donations is also critical to know: we need to ensure that works also online. Thanks a mega ton!
Advanced Skirmish AI Team Lead for the coolest Warhammer40k PC RTS out there:

Dawn of War Advanced AI Headquarters

Latest DoW Advanced AI Download!

#19 Inquisitor

Inquisitor
  • Members
  • 121 posts
  • Location:Ordo Malleus

Posted 06 August 2007 - 04:21 PM

Thanks Inquisitor.. AI Donations is also critical to know: we need to ensure that works also online. Thanks a mega ton!

I'd like to test it but I'm changing my ISP and they failed delivering today, so I'm without internet (I'm currently connected via my phone, which costs $1.5/MB). I don't know when I'll be back online, but at least it will be on 20Mb down, 2Mb up... :popcorn:
The Emperor knows, the Emperor is watching.
Want to zoom out further? Try this camera mod.

#20 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 06 August 2007 - 06:33 PM

Thanks Inquisitor!

As for the current build..

1) Yeaa I see alot of enemy troops running by enemies trying to get to a distant LP of theirs without first fighting to get there. I gather this is VERY hard to code as too many scenarios are possible here.

2) I have noticed something usual where the AI will "sometimes" build a building set in the building placement as "Backbase" on the Front (ie. Front1 designation). This never ever happened before. So in other words, I saw the Necron Ai put the Energy Core (which is clearly a BaseBack building) right on the front. Strange eh?
Advanced Skirmish AI Team Lead for the coolest Warhammer40k PC RTS out there:

Dawn of War Advanced AI Headquarters

Latest DoW Advanced AI Download!



Reply to this topic



  


0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users