Jump to content


Photo

Skirmish AI 2.3 Beta 8b - Post Comments In Thread!


19 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 28 April 2007 - 10:35 PM

All Testers please ensure AI is building on ALL its captured Strat points in a timely manner. We want to ensure its NOT "slacking off" ensuring its fortifying its conquered territory thus disallowing decapping by the enemy. :p Please look for this.

We are sitting on a Release Candidate here so lets arse-kick this one to the community VERY soon!

2.30b8b
- Fixed MAJOR LP building bug where AI on some maps fails to build LPs on certain stratpoints. Now AI will build on all its captured stratpoints as it conquers new territory.

2.30b8
- Fixed Tau build programs
> Drone squad replaced by Skyray
> Tau barracks in starting build order
> Carnivore strategy working now
- IG fixes
> Command squad reinforces always during harassing time
> Command squad builds 2 priest and only a psyker if stealth units/buildings are present
> Decreased basilisk rating
> Forced 2 Commissars in allround strat
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 ThetaOrion

ThetaOrion

    title available

  • Members
  • 676 posts

Posted 28 April 2007 - 11:08 PM

Sounds like a plan to me! :p

--

It will be interesting to see what further Larkin or Inquisitor come up with.

--

Missed seeing Zenoth around here. Hope all is well with him!

#3 LarkinVB

LarkinVB

    title available

  • Members
  • 1,488 posts

Posted 29 April 2007 - 09:39 AM

3 vs 3 at bloodshed ally.

Necron were building generators at front LP. Other LPs which were not that exposed were available as build positions but ignored.

I observed CS and it reinforced fine at harass but I'm not sure it was reinforcing later.

Edit: I rechecked and CS did reinforce after harass later but very slowly. Often CS was build and general did fight and died alone.

I never watched the tau commander jump at all for some weeks now. Even with the new emergency retreat it seems reluctant to use his jump pack.

I think this is all for now. From my point of view the AI is fine and should be released.

Edited by LarkinVB, 29 April 2007 - 11:17 AM.


#4 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 29 April 2007 - 12:04 PM

Necron were building generators at front LP. Other LPs which were not that exposed were available as build positions but ignored.

I would check that. In some RARE cases if the AI thinks its enemy is away from it furthest from a frontal LP then this can happen. I'm not too concerned as 9/10 times it will build Gens furthest away from the enemy. Regardless.. sounds like a map prob then anything.
Advanced Skirmish AI Team Lead for the coolest Warhammer40k PC RTS out there:

Dawn of War Advanced AI Headquarters

Latest DoW Advanced AI Download!

#5 LarkinVB

LarkinVB

    title available

  • Members
  • 1,488 posts

Posted 29 April 2007 - 12:44 PM

It is a bit suspicious that necron generators do appear twice here in this code. They are build SafePlace and not BaseBack as other generators. SafePlace seems to prefer LP close to enemy instead of those close to own base sometimes. Necron were at position 2 on bloodshed alley if you want to check.

	elseif (cpu_manager:IsGenerator(iBuildingID)) then
		return "Safeplace"
	elseif (cpu_manager:IsGenerator(iBuildingID) or
			iBuildingID == cpu_manager.stats:GetBuildingID("necron_summoning_core") or
			iBuildingID == cpu_manager.stats:GetBuildingID("necron_forbidden_archive") or
			iBuildingID == cpu_manager.stats:GetBuildingID("necron_greater_summoning_core") or
			iBuildingID == cpu_manager.stats:GetBuildingID("necron_energy_core")) then
		return "BaseBack"
	end

Over and out for v2.3. Good job, thanks to all involved and have a nice day.

Edited by LarkinVB, 29 April 2007 - 12:50 PM.


#6 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 29 April 2007 - 01:02 PM

No sir.. Safeplace I coded is the complete polar opposite of Front2 which biases all builds away from the enemy instead of towards. This was thoroughly tested. It works great as the last thing we want is the Necron AI piling all Gens at the Startpoint thus bottlenecking Destroyers , H.Destroyers, and TombSpyders. This drastically helps reduce this while keeping vulnerable Gens away from the enemy.

Actually, now that I see above we could remove the BaseBack for Gens but since Inquisitor has not said anything bad about the Necron AI I assume its ok the way it is now.

Try another map like 2p_FallenCity or 2p_BattleMarshes and watch the Necron Gens be placed around and back from the HQ. If it doesn't find any more room around the HQ, it will place em on LPs away from the enemy. Sometimes where the LPs are on a map will be a little screwy so this can make SafePlace a little more challenging.
Advanced Skirmish AI Team Lead for the coolest Warhammer40k PC RTS out there:

Dawn of War Advanced AI Headquarters

Latest DoW Advanced AI Download!

#7 Inquisitor

Inquisitor
  • Members
  • 121 posts
  • Location:Ordo Malleus

Posted 29 April 2007 - 01:05 PM

Well, I have been doing some testing all around and I really like what I see with most of the races. I have also tested a lot of mp (LAN most of it) and the AI performs really well as an ally as well as an adversary.

I think the improved Necron plays much better and I think that thudo's lab proves that as well. Thank you for incorporating a lot of my suggestions, ArkhanTheBlack! I think the improved building of gens gives the Necron the economy it needs to keep up momentum in the later game. I also think that the new NL research patterns are a step in the right direction. I esp like that there is default pattern, the Phylactery is now part of all of them and that Solar Pulse is part of two. One thing I would like to have more than 25% chance is the Lightning Field since it is T1, makes the NL much more resilient in CC and it is potentially powerful since it keeps charging indefinitely.

Suggestion: LF should be made part of the default pattern (i.e. pattern 4) instead of Veil of Darkness. Veil of Darkness is already part of pattern 2 and is T3 and quite costly. This is just a suggestion and just my opinion.

My impression is that Tau is the race that the AI plays the best. I can't put the finger on what it is, but I guess it is making the most of the race's strengths and the dancing code makes it offset it's weakness (i.e. it is a glass cannon army).

I have no other remarks on this build currently and I would just like to commend you on a job well done! :)

I hope to get some more mp testing done tonight and I will ofc report interesting finds.

...

On a longer term I would like the team to consider more specialized anti-race builds, for instance:
  • An anti-necron IG build/behavior might be an interesting first thing to discuss. Objective: The AI IG should be better at fighting Necrons.
  • An anti-Tau Necron build/behavior is also very interesting, since Necrons have no chance against Tau (and it is the other way 'round when humans play). Objective: Necrons should win most of the time in 1v1.
When I think specialized builds/behavior I think of a tailored build that has about 100% probability when playing a 1v1 vs. the race. It might require more than just tweaking of the builds. The team could maybe think up new strategies and write new code to implement the behavior.

This exercise might even help the team improve the AI overall as some of the new "inventions" might be useful strategies that can be used in general as well. :xcahik_:
The Emperor knows, the Emperor is watching.
Want to zoom out further? Try this camera mod.

#8 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 29 April 2007 - 01:08 PM

@Inquisitor - well I'm open to some IG builds to combat Necrons but I highly doubt we'll find em. IG just plain blow donkey balls against ONLY Necrons and perhaps Tau. Everyone else they are decently balanced. Anything that can help would be great but I've covered everything IG vs Necron-related on other boards recently outside this board. Tried all suggestions.. Nope.. no go.
Advanced Skirmish AI Team Lead for the coolest Warhammer40k PC RTS out there:

Dawn of War Advanced AI Headquarters

Latest DoW Advanced AI Download!

#9 Inquisitor

Inquisitor
  • Members
  • 121 posts
  • Location:Ordo Malleus

Posted 29 April 2007 - 02:01 PM

Actually, now that I see above we could remove the BaseBack for Gens but since Inquisitor has not said anything bad about the Necron AI I assume its ok the way it is now.

I would say that placing gens away from the base is high priority because as you said it will reduce the risk of stuck units in the base (which is a big problem for Necrons in many games!). So if the choice is between the base and at some forward LP I would prefer the forward LP. But if there exist code that could pick a point furthest from the enemy, but not in the base it should be utilized.

@Inquisitor - well I'm open to some IG builds to combat Necrons but I highly doubt we'll find em. IG just plain blow donkey balls against ONLY Necrons and perhaps Tau. Everyone else they are decently balanced. Anything that can help would be great but I've covered everything IG vs Necron-related on other boards recently outside this board. Tried all suggestions.. Nope.. no go.

That's why I suggest that we consider doing something new. We need to analyze why the AI fare so bad against the Necrons and doing so would involve looking at how good human IG players win over Necrons and looking at what goes wrong for the AI. If we can pinpoint some things that we want the AI IG to do or not do we can perhaps make it perform much better with little effort.
I will not say "no, it can't be done", when I know that most IG players said exactly that when 1.2 was 2 weeks old, but now there is a lot of evidence that it can be done and the mood is completely different. IG might be the underdog in the early game in DC1.2, but it is certainly one of the strongest races in the late(ish) game. So, a concrete idea might be: "how can we make IG survive long enough for it to get strong enough to win vs. Necrons". But again, I think we need to be more analytical about this and base our efforts on evidence and perhaps discuss it in more detail in a separate topic - if the team thinks it is worth a shot. In this way the opinion of ThetaOrion and others will give invaluable input to how the AI can be improved.
The Emperor knows, the Emperor is watching.
Want to zoom out further? Try this camera mod.

#10 LarkinVB

LarkinVB

    title available

  • Members
  • 1,488 posts

Posted 29 April 2007 - 08:17 PM

Whatever you like. I think this release is golden.

#11 ThetaOrion

ThetaOrion

    title available

  • Members
  • 676 posts

Posted 29 April 2007 - 10:55 PM

IT'S TIME:

It sounds like me that it's pretty much time to release it.

That will give you about few thousand people more to tell you what is wrong with it, and right with it.

--

WHAT A THOUSAND TESTERS WILL DO:

I suppose as Aralez and others play 2.3 over time, he will have some more AI ideas that he would like to see implemented. Aralez put in a number of AI requests the last time he did the UC Mod. You start seeing things. I even eventually started seeing things that were wrong with the 2.1 AI over time, like the Choas builders not building secondary bases. Arkhan went right in and fixed it.

I suppose the same thing will happen as thousands of people get a chance to look at 2.3 and then figure out for themselves what more the AI needs. But, you do need time to play it and see for yourself what more is bugging you, if anything.

--

A REPEAT OF HISTORY:

The past couple years, LarkinVB has had the best eye for what the AI Mod needs most, imho. If he says it's done or good enough, and if he says to release it, then it is indeed time to release it.

And for the UC Mod, Aralez will also look at it and maybe decide that some of it is balance issues and probably buff the IG and nerf the Tau in the UC Mod and call it done. I don't suspect that he's going to have to do nearly as much with 2.3 to get a great balance as Aralez had to do to get a great balance out of 2.1 and 2.2.

Arkhan told Riker's DoWXP people that 2.3 would be out this weekend, and then Arkhan said he would help port it over to DoWXP this week. I imagine that the process will work the same as last time, with Arkhan finding further things that he wants adjusted in the AI Mod and just doing so while he prepares the AI for DoWXP, thus getting a start on and paving the way for 2.4. That was a good system last time around, and it's great to see that Arkhan is willing to do it again. I'm actually eager to try DoWXP 3.8 with 2.3 AI mod for 1.2 Relic Patch. There have been some 'change's in DoWXP since last I played!

You also have KaiserSoze porting AI Mod to DoWPro or DCPro, with LarkinVB watching over his shoulders. They will see things and have requests over time.

You guys will also find other things you want changed or improved as you start porting 2.3 AI Mod into Tau Exiled Union, Steel Legion for DC, the Tyranid Mod, and Shin's Eldar Mod.

--

REGARDING INQUISITOR'S DESIRES:

I also suppose that once you release it, some of us might find ways to win as IG vs Necrons using only the 2.3 AI Skirmish Mod, if it is indeed possible to do. And, that too might help to feed into Inquisitor's desires and goals over time.

When I was last programming, I found out that if I could do it on paper, then I could program it. That was a long while ago that I last programmed, but the idea still pretty much applies.

If Inquisitor can get to where he is always winning with the plain vanilla 2.3 AI as the IG vs Necrons on a variety of 1vs1 maps, then he should be able to supply you guys with replays showing you how he did it, and then Arkhan should be able to use those replays to glean ideas and adjust the AI so that the AI can do it too, in 2.4

--

SUMMARY:

There are plans and steps for going forward. It worked well last time, and I suppose it will work well this time around. Just need to release it now, and let them at it.

Then 2.4 will take shape on many fronts, once everybody has their hands on 2.3. I have faith that it will be so!

Edited by ThetaOrion, 29 April 2007 - 10:58 PM.


#12 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 30 April 2007 - 01:55 AM

There has to be a compromise between Larkin and Inquisitor for smarter Necron Gen Placement. Right now, its balanced between both but it depends on the map and its startpoints. Right now, Necron AI will first build the first 4-6 Gens @ the Startpoint then the rest around the furthest LPs from the enemy. If the Necron AI looses those Gens around the HQ or on expansion LPs then it re-distributes them to less threatened areas.

As for going for Release Candidate! I do concur.. :lol:
Advanced Skirmish AI Team Lead for the coolest Warhammer40k PC RTS out there:

Dawn of War Advanced AI Headquarters

Latest DoW Advanced AI Download!

#13 ArkhanTheBlack

ArkhanTheBlack

    title available

  • Members
  • 814 posts

Posted 30 April 2007 - 10:41 AM

Okay, it looks like we're ready for release. Thanks to all testers and coders for your help. Thudo will handle the release now like always. :lol:

#14 Inquisitor

Inquisitor
  • Members
  • 121 posts
  • Location:Ordo Malleus

Posted 30 April 2007 - 11:38 AM

Great news! :unsure:

Suggestion: LF should be made part of the default pattern (i.e. pattern 4) instead of Veil of Darkness. Veil of Darkness is already part of pattern 2 and is T3 and quite costly. This is just a suggestion and just my opinion.

I would really like a comment to my suggestion... :lol:
The Emperor knows, the Emperor is watching.
Want to zoom out further? Try this camera mod.

#15 ArkhanTheBlack

ArkhanTheBlack

    title available

  • Members
  • 814 posts

Posted 30 April 2007 - 01:21 PM

I would really like a comment to my suggestion...

To be honest I'm not really a fan of this suggestion. Veil of Darness might be expensive but it's also a very powerfull ability. In DowXP they considered the 'stealth monoliths' even so OP that they excluded the monoliths from the Veil of Darkness effect (at least that's what I've heard). Detectors might be common in later tiers, but in larger battles detectors also tend to die very fast, especially since they have to be close to the front line to be of some use. Veil of Darkness can hide an entire army, while lightning field is only usefull for the NL and doesn't have any effect on the surrounding army. In later tiers, I don't consider it a match for Veil of Darkness. The time Veil of Darkness is researched is also later in the game when the economy is stronger. I therefore consider 300 power not that much for this time and definitely worth for such a powerfull effect like Veil of Darkness.

#16 Inquisitor

Inquisitor
  • Members
  • 121 posts
  • Location:Ordo Malleus

Posted 30 April 2007 - 01:34 PM

I would really like a comment to my suggestion...

To be honest I'm not really a fan of this suggestion. Veil of Darness might be expensive but it's also a very powerfull ability. In DowXP they considered the 'stealth monoliths' even so OP that they excluded the monoliths from the Veil of Darkness effect (at least that's what I've heard). Detectors might be common in later tiers, but in larger battles detectors also tend to die very fast, especially since they have to be close to the front line to be of some use. Veil of Darkness can hide an entire army, while lightning field is only usefull for the NL and doesn't have any effect on the surrounding army. In later tiers, I don't consider it a match for Veil of Darkness. The time Veil of Darkness is researched is also later in the game when the economy is stronger. I therefore consider 300 power not that much for this time and definitely worth for such a powerfull effect like Veil of Darkness.

Alright. Thank you for taking the time to comment.
The Emperor knows, the Emperor is watching.
Want to zoom out further? Try this camera mod.

#17 kreative

kreative
  • Members
  • 27 posts

Posted 30 April 2007 - 03:50 PM

After the discussion about IG vs Necrons I've really taken an interest to them.
Spent quite a while creating the banner and badges, and I really love playing them now!
I've also had some interesting fights vs Necron, and I beat them more and more.
I turtle, and spend early resources on upgrading LP's, turrets and Infantry Command close to LP's. Not sure how the AI handles garrison of units, this is essential for my success against necrons. I still need to play more maps to validate my strategy.
I'll get some replays up later.

#18 Weirwood

Weirwood
  • New Members
  • 1 posts

Posted 30 April 2007 - 05:37 PM

Just wanted to chime in and say how much your work is appreciated.

On the IG vs Nec front, one pretty counter-intuitive thing to do, that I've had some fun with, is to melee NWs with priest-attached GMs, while shooting them with HWTs and/or chimeras, as well as your commie squad.

Until you get weapon specialisation, a priest'd squad will deal more damage in melee than it would ranged with 3 plasma guns. With the priest's hp bonus, (and preferably, bot not necessarily, Battle Armor and sergeants) GM make very cost-effective meat-shields. The priest also easily takes care of wraiths. flayed ones, however, are a problem, that's when you fall back to your chims.

But more generally speaking, when and how to use GM in melee can make or break IG early game. I don't know to what extent you've implemented that or experimented with it, but priest melee squads can be an alternative to heavy weapon GMs in T2, if your opponent hasn't invested much in CC himself. And if he does to counter you, then you can still swith to priest plasma on the fly.

#19 mdcertainty

mdcertainty

    Sage-Otaku-Monk

  • Members
  • 329 posts

Posted 30 April 2007 - 06:41 PM

I must say you guys put a lot of hard work into this Project apparently for a long time now and i´m really impressed. The AI improvements of 2.2 were awesome and i´m looking foward to 2.3. I´m very grateful for your work.

A big THANK YOU to all AI developers and testers here.
You guys rock, and make a match with or against the AI a real challenge :) :)

keep it up :D

ec-sig-early.gif


#20 ThetaOrion

ThetaOrion

    title available

  • Members
  • 676 posts

Posted 30 April 2007 - 09:57 PM

With the IG vs Necrons, I'm sure the people will find a way around it, once they have their hands on the final product.

Gee, long ago, I noticed yet another trick. Most people, when they encounter the Necrons, post up and fight until they are dead. But, the Necrons are slow, and you are fast.

What I often did on the larger maps is to run up to the Necrons, fire at them to get their attention, and then run into their base. You can get to their base faster than they can get to yours. And, I noticed that most of the time, when I'm attacking their base, the Necron death march will actually turn around and return to their base rather than continuing forward into your base.

Of course, it probably won't work on smallest 1vs1 maps, and some other maps with unique designs, but it can work at times.

Lots of things to try, just need the AI mod to try them with, and I'm sure somebody will come up with lots of things to try.



Reply to this topic



  


0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users