9 replies to this topic

[LarkinVB]

I would like to get some suggestions on how to improve the AI any further. I would concentrate on *really* noticable improvements in fighting strength and fun, not stuff like "builders are wandering around" or "three turrets at bigger generators".

I think micro is ok now (as long as we can't target squads) but attack/defend strategies should be improved.

[Malkor]

A lot of the time warp spiders, instead of warping away, will warp right into an enemy army. I always found that kind of silly.

As for major improvements/additions, well... How about different build orders? Such as, SM could go mass dreadnaughts or mass predators or all out land raiders for their vehicles or the like. Some could try to tech early or something... just different styles of playing.

[oozish]

How about addressing the lone commander attacking base syndrome? Like only move out squads to take 'safe' LP's until commander can join (1-2 squads?). Too many times I've seen the ai pump out 1 squad like a factory that just goes running into battle without reinforcements and/or commander.

1 squad with FC is ok.

2 squads is ok together....would make attacks much more effective wouldn't it? Just thinking out loud.

Personally I would never attack with a simple lone squad, and if I attacked with my lone FC, it would be only yto harrass or deter the capture of a CP or LP.

Edited by oozish, 30 August 2005 - 04:16 PM.

[LarkinVB]

AI shouldn't attack alone except on small maps. Can you show a replay. Usually the commander does attack with one squad as we have min_units (to attack) set to 2.

[thudo]

As for major improvements/additions, well... How about different build orders? Such as, SM could go mass dreadnaughts or mass predators or all out land raiders for their vehicles or the like. Some could try to tech early or something... just different styles of playing.

AI actually do this: its based on a demand system and I've once seen 3-4 LandRaiders by the AI (nasty!) so each game you play its quite random.

As for improvements (priorities from our Objectives List):

1] Assassinate - Ensure Commander Unit is kept away OR minimizes his exposure to combat (AI still keeps his Commander too involved in combat and, on average, gets him/her killed) - Flenser

2] Control Area/Take And Hold - Ensure continued AI aggression under both these victory conditions - Flenser

3] Allow AI to recreate his HQ on another unthreaten Listening Post if his original HQ is destroyed - Flenser/Larkin

4] Creates a secondary HQ + unit production buildings w/ defensive turrets (optional) all built away from the enemy - Flenser/Larkin

#1 and 2 are really Flenser's territory but he's busy at present. #3 is far more important than #4 as having the AI recreate his entire base elsewhere is key. HQ, once original is taken down, should move somewhere else. A new base of operations. This used to happen in vanilla DoW but there are positives to the code as it stands that keep the HQ at startpoint.

Other objectives will happily wait til expansion (3rd-4th week of Sept) as Relic will HOPEFULLY have some good AI news for us! <prays>

[LarkinVB]

Well ,I was looking for something interesting and spectacular ... :-)

[oozish]

Regarding the lone Commander attack, I've seen this alot on valley of korne, especially SM SM...the commander isn't attacking with a squad, but a squad shortly follows. I've also seen this same thing with ork.

I don't t hink I have any replays at the moment but when I do I'll post it.

[Excedrin]

Simply keeping all of its units together would be a huge improvement. Having one squad lagging behind or way out in front is really bad.

Going past that, if it could do things like attacking both sides of two sided maps, that would be another big improvement.

Ideally it would have a big force countering the opponent's army, preferably setup in cover when possible, with the commander(s) slightly out front to soak up fire and melee or jump troops nearby to jump in and tie up as much as possible. At the same time, a second force is running around trying to take out LPs and generators and decap.

The AI vs AI game I posted shows other serious problems, like 900 unspent req and little power, the insistance on heavy bolters and pointlessly losing the commander.

Edited by Excedrin, 30 August 2005 - 07:02 PM.

[oozish]

RE: HBs....I think HB's for the first 2 squads (before tier 2) is fine...but save slots for when you get the heavy weapons upgrade and more importantly tier two (plasma) level.

Early on, I don't see anything wrong with keeping the squad to 4 or 5 and equiping the SM's with HB's; or even just 1 each squad...but something! It definately gives the edge needed in the earlly game, in my opinion. However, I can see the logic in just amassing a big squad and saving ALL the slots for plasma later....it's just not something I do.

Now, I'm a bit out of touch and writing this without addressing the HB's thread, so I'll have to go read that now.

Edited by oozish, 31 August 2005 - 02:49 AM.

[Markoso]

For the most part, I'll disagree on most of the points brought up thus far. In 1.3, there aren't too many attack plans you can go with, unless we built in the alternate SM BO's (ASM builds anyone ?), not a lot else that can be done beyond fine tuning the hell out of the micro.

Stuff like giving the AI the ability to harass (while it IS akin to a rush, it's a viable tactic especially for eldar and SMs, unless they're facing orks) would make a pretty big difference as long as the harass was something that only fired once, and preferrably just ran in looking for points to decap until it's squads attack rating fell...hmm, much lower then the enemy forces on their ass. Restrict the code to running ONLY for a scout squad in the case of SM, and only for eldar after FoF is researched, and you've got something you can play around with.

Edited by Markoso, 03 September 2005 - 08:55 AM.