86 replies to this topic

[thudo]

Yeaa Tier1 IG is not bad at all really. The Guard can usually be okay but once they get Vehicles they are brutal. I so love the Chaplain and Mega Armoured Nobz although why the f**k are they both created at the HQ? Amoured Nobz should be in the Boyz Hut with their Nob breathern. Chaplain should be in the Artifact with the Librarian for pete's sake! Also.. Chaplain should be attachable as the damn n00b goes running all over. Berserkers and Firedragons - their okay. MegaNobz rock though - love seeing the AI use their power surge ability. What a slaughter fest! Muhahahah!

But yep.. note how much has changed. So many balance changes and the tech tree is all different, to a degree. Really do not like the new Farseer voice - sounds like some has-been flirty valley girl on uppers. :(

Still.. AI for it has been coded and Larkin and I are tweaking it. FunFun..

[thudo]

Thanks Shin.. we're looking into all the changes. Almost all the factions have gotten buffs so now its not "roll Defilers across the land to win" anymore.

Still.. Eldar AI should first get DRs out then MAYBE we'll consider Banshees. If they are much better than before then coo... Should check to see if Rangers also are improved.

[oozish]

There is none but I heard an Image Crack is available. 

<{POST_SNAPBACK}>

oozish here...ya, i saw that image post(s)...looks like too much work for me.

[oozish]

Loaded up WA last night, it's good stuff.  Vanilla AI definately got an improvement as well, that, or either SMs, or I, suck...badly.  Also, tier 1 IG...man...they really don't need to defend all that hard, those flashlights of theirs tend to kill scouts just fine.  In fact, they fight quite well offensively at tier 1 against my pathetic SMs.

<{POST_SNAPBACK}>

Hey man, I'm feeling ya...I played SM vs IG on that new, first map in the list...I got clobbered each time. Seems the IG are a force to be reckoned with; I'm still not sure, but it really seems like the SM need some help defeating their very powerful gl's. I am enjoying playing the IG, the only other race I've really dabbled with. I think I'd like to learn more about them for sure.

AI I've read elsewhere has tanked badly with the release of WA...I'd have to disagree, it's definately better/harder for me.

[oozish]

btw, thud/larkin...did you see that post about the next patch I read somewhere? Someone on the relic team said AI would be a major focus on the next patch.

It would be great if they'd touch bases with you both. Cheers!

[thudo]

Someone on the relic team said AI would be a major focus on the next patch.

Where did you see this?

[Excedrin]

That's a Tranj quote from a thread in the DoW general forum on relicnews.

Oozish, I agree that the AI seems to play better. I think that part of the "mixed" opinion about the AI on the relicnews forums is due to people playing with different settings.

For example, if I play a 3v3 game on Kasyr with AI allies, they never build generators if shared res is on. With shared res off, they do much better. I believe that is due to the "build a gen when we have less than min_power" logic. With the human player making generators, the AI never runs that low on power, so it keeps leeching power. That makes the human make more and more gens, which never achieves the result of a decent power income. I suspect that the usual player who wants to play team games with AI allies doesn't understand the impact of shared resources, so they don't turn it off before playing skirmish games. Then they have the perception that the AI is much worse than before (since the older AI didn't tech much or use much power).

On the other hand, when playing a 1v1 against the default WA AI, my only serious complaint is that it doesn't fight for map control in the early game. To someone who doesn't do that themselves, it's a non issue. Therefore, I suspect that most people who try 1v1 skirmishes have the perception that it's improved from the earlier AI.

[thudo]

they never build generators if shared res is on. With shared res off, they do much better. I believe that is due to the "build a gen when we have less than min_power" logic. With the human player making generators, the AI never runs that low on power, so it keeps leeching power.

Thats interesting.. that couldn't be the reason why the AI is sometimes poor when a human ally is on the team? Just turn off Shared Resources and the AI works indepedantly. But wait.. there is NO CODE for the AI to leech resources off anyone AI or Human so how can that be a factor. AI doesn't even know how to delete units or buildings. The code has to be there for the AI to aware of it, no? Still.. worth seeing if Shared Resources makes a difference.

[Excedrin]

There is code for it. The AI only builds a generator if it has less than X power righ? If the human player builds 3 gens because he/she needs power NOW, then the AI doesn't drop below X power. I think the issue is that on Kasyr or another big team map, humans will typically build gens much faster than the AI.

In other news, I just played my first game against 1.5b1 in WA. I see that none of my "keep squads alive" code was integrated, so it's still very easy to run over the AI in the early game, exactly like 1.4. It was SM vs SM on Fallen City. I'll try IG vs IG on the same map next.

[thudo]

Thats news to me: if shared resources is abled then Power Plant output is shared among allies? Damn.. so if Shared Resources is disabled then the Allied AI on a human team will actually play MUCH BETTER than with that option disabled?

As for the "keep squads alive" code - we need time to first do the basic stuff then we'll get the "good stuff" in. However, Exedrin.. since we haven't tested your code in WA conditions, can you integrate into v1.5b1 to see if it works cool and send it my way. I still like your idea for the early game - we're just busy getting WA working at the rudimentary level.. expansion has only been out.. ohh.. a week.. many places in the world don't even have it yet.. :(

[Malkor]

With luck Malkor will have it saturday. With luck Malkor will beat all campaigns in that same day. With luck Malkor will have opportunity to play 4 IG vs 4 Orks for intense chaos. With luck Malkor will have super AI to test this with. With luck WA has much bigger/better 8 player maps.

[Excedrin]

I played against IG (watched a movie since my last post), it was ... a bloodbath. I'll upload the replay in the hope that it sheds some light on one way to play IG.

http://dow.lerp.com/...0929.223756.rec

Will the AI ever be able to put troops in buildings?

[thudo]

Will the AI ever be able to put troops in buildings?

Depends if Relic was able to put that code in. That and allowing the AI to repair buildings, focus fire, allow AI to use Infiltrate (Scout) or Rampage (Squiggoth), Inter-AI communications...

[LarkinVB]

AI will have no chance against early infantry massing + pressure, with or without 'keep squads alive'. It can't use terrain, can't focus fire and has some squads capping elsewhere. So its inferior in numbers and skill, what do you expect ? AI can only shine in bigger battles, not 1v1 against good players doing high pressure.

Btw the AI currently tries to tech to vehicles fast so IG starts with a 2 gen build and is even more vulnerable. But a squad more wouldn't make a difference against your attack.

Edited by LarkinVB, 30 September 2005 - 06:08 AM.

[Excedrin]

Micro isn't all about focus fire. Infact I rarely focus fire (my micro is only decent). Now that the targetting priorities are better (no more accidentally shooting buildings), it's almost completely unnecessary. The exception to that is commanders, but it seems like moving towards a squad and then setting melee stance works well (I think that's what the current DoW Skirmish code does, but I could be wrong).

With my changes, it almost survives SM vs SM scout rush. Without them, it's pretty easy (remember that "how fast can you beat the AI?" thread?). I'm not saying that InfantryTactic:AvoidDeath() solves all micro problems, just that it's a good starting point to build better micro.

As far as being inferior in numbers and having some sort of rigid build order, there's no reason that it has to have that. It should probably base the number of squads/gens it has on the number of squads/gens its opponent has.

Also, it can use terrain to some extent. I don't expect it to jump warp spiders behind a wall to avoid CC troops like a human will, but it can move thru cover on the way to some goal.

The primary reason it fails so badly against an aggressive attack is not because it can't focus fire, it's because A) it has squads all over the map, instead of all in one location and B) it lets squads fight it out until they're dead.

[LarkinVB]

I will add your code to next beta Show me a replay doing the same strat. If the AI survives I will give you all the credit. Not fighting to death will lead to retreating. You follow up, rinse and repeat until there is no more point to flee to. But it won't hurt so the idea is fine. To stop your attacks AI must field superior numbers at the point YOU choose as you are the attacker. Of course it has squads all over the map. They try to capture. If they won't and you decide to tech they will loose too. Its simple as that : Against a good player AI will loose always. My fun in coding AI is for big team battles, not to stop a 1v1 aggressiveness. If you can add code which will allow both I will be happy.

I would also like to see your code were the AI can use cover to good effect in comba but I think it's not possible.

EDIT: To use AvoidDeath I need a new customstats.ai with IG and new units. Don't know the values.
EDIT2: Your code is incomplete as it won't put the squad into a substate. Therefore it will try to disengage only to get another order from attack/defendplan a second later.

Edited by LarkinVB, 30 September 2005 - 08:38 AM.

[Excedrin]

Without being able to use buildings, it'll be difficult to get IG to survive early attacks, but it should still be somewhat better.

The number is basically max squad size. I think the following numbers are correct.

guard_squad_enginseer =
{
  unit_max = 1,
  melee = false
},
guard_squad_guardsmen =
{
  unit_max = 10,
  melee = false
},
guard_squad_command_squad =
{
  unit_max = 5,
  melee = true
},
guard_squad_kasrkin =
{
  unit_max = 10,
  melee = false
},
guard_squad_assassin =
{
  unit_max = 1,
  melee = false
},
guard_squad_ogryns =
{
  unit_max = 6,
  melee = true
},
guard_squad_psyker =
{ 
  unit_max = 1,
  melee = true
},
guard_squad_priest =
{ 
  unit_max = 1,
  melee = true
},
guard_squad_commissar =
{ 
  unit_max = 1,
  melee = false
},
guard_squad_sentinel =
{
  unit_max = 1,
  melee = false
},
guard_squad_hellhound =
{
  unit_max = 1,
  melee = false
},
guard_squad_basilisk =
{
  unit_max = 1,
  melee = false
},
guard_squad_chimera =
{
  unit_max = 1,
  melee = false
},
guard_squad_lemanruss =
{
  unit_max = 1,
  melee = false
},
guard_squad_baneblade =
{ 
  unit_max = 1,
  melee = false
}

I'll integrate my stuff into 1.5b1 in the next couple days unless someone else does it. Also note, I don't really consider my code finished. There's issues with melee staying set to ranged forever and it still spreads stuff all over. I guess you could consider it an attempt to improve "B" above.

Regarding the lack of using a substate, the unintentional effect is that squads get pulled slightly out of firing range until they form a big wall of ranged fire. Which brings up another point, I haven't tried to tune it for melee units.

Edited by Excedrin, 30 September 2005 - 08:58 AM.

[LarkinVB]

I'll integrate my stuff into 1.5b1 in the next couple days unless someone else does it.

I'm currently adding it to my 1.5b2 , makes it 1.5b3 . Needs serious testing as I did some code cleanup which might easily mess the fragile substate system.

EDIT: Did submit to Thudo. Disabled AvoidDeath() for CC as they went back/forth often.

Edited by LarkinVB, 30 September 2005 - 07:04 PM.