87 replies to this topic

[ArkhanTheBlack]

Larkin - is there a way to specify a distance from the Startpoint or HQ assets so that the "Repair Radii" can be increased? It would be useful "to a degree" to have the closest LPs be available for repair and thats it. No outside that area. Also.. I've noticed engineers mid-game (where most damage to buildings occur) will sit idle most of the time so is there a way to increase the repair-frequency checks after, say, Tier3 onward? Tier1-2 the engineers are busy building their bases so no repairs will likely be needed.

Yes, I think we can expand the repair radius. The repair frequency is 10 sec at the moment, but if the IsIdle() really works (I'm still not 100% sure...) then we maybe could skip the timer entirely. I'll try this out...

groups of A.I being lost

Actually I think I know the reason why the AI is sometimes waiting at an allied base doing nothing. If I'm right, then the AI is gathering troops for an attack. I've seen that too and it mainly happened at the start of a game. At this time often more than half of the army is busy capturing listening points and is not available for attacks, at least not yet.
I maybe should correct this in the attack strategy, so that he doesn't consider the entire army strength for an attack, but only the available squads. This way the AI squads shouldn't wait at an allied base for 5 minutes eating popcorn an watching TV, but instead start the attack immediately or not at all.

[Malkor]

All seems well except for the SM using word of the emporer when threatened by 3 rocket launching orks (which were not even remotely threatening to their 20/20 army), and the dreadnaught rushing SM taking way too long to get their armory. Their armory should have a priority higher than infantry and heroes and they should get it ASAP along with tier2 and the machine cult. Same with the Eldar; they wasted too much time piddling around in tier1 and as a result most of the rushes were too slow to be effective. (On an FFA in penal colony which I don't have the replay for).

[ThetaOrion]

rushes were too slow to be effective

--

In the previous betas, I mentioned not liking the rush tactic, especially on the larger maps. I wondered if there were ways to turn off the rush tactic on the larger 6 player and 8 player maps, so that the AI enemies and AI allies don't waste tons of resources preparing for a rush that's never going to materialize or work out anyway.

The rush tactic is only effective (and annoying) on the smaller maps. On the larger maps, the rush tactic always hurts the AI performance, especially the more complex the map or the more distant the enemy bases. On the larger maps, the AI is better served by tiering up to the next level as quickly as possible and skipping the rush tactic completely.

[Malkor]

I can rush effectively in an 8player FFA with vehicles and get away with it. I don't see why the AI can't, as I am very lazy and sloppy and don't even follow a strict BO. The AI just wastes resources losing it's units and rebuilding them, though. At max I get two scout squads soley for capping and MAYBE a turret if I have spare resources.

[ThetaOrion]

The AI just wastes resources losing it's units and rebuilding them, though.

--

I finally isolated this problem for myself in the 1.61 Beta Two, where I complained about it.

In Beta2, the Imperial Guard were locked into a Tier One rush the whole game while I was finishing research and deep striking Final Tier stuff into their base.

At least this time around, in the Beta3, the IG TechPriest Enginseers shouldn't be locked into the Tier One/Tier Two rush, just simply rushing my turrets and listening posts to no effect. It was the strangest thing to see the IG producing only First Level Infantry, a Commander, and one TechPriest after another, and that line of TechPriests leading straight to my nearest turret where the IG Engineer just simply died.

The Rush Tactic really kills the effectiveness of the AI, in my opinion. Obviously, others see it differently.

[Malkor]

In starcraft, I made highly effective battlecruiser and carrier rushes. The AI didn't get ANYTHING but the required structures until it acquired it's first set of heavy air, then it trounced the enemy's unsuspecting forces. Should the enemy live, it then acquired a set of supporting units and more of the heavy air and finished the job.

This is how the DoW rushing AI should work. It shouldn't be in offensive AT ALL until the first vehicle pops out. It should avoid combat completely until it has it's vehicles. Is that hard to do, or something? It shouldn't even have a command squad.

Edited by Malkor, 12 January 2006 - 03:20 AM.

[Zenoth]

I understand what you mean Malkor.

But I believe that early Infantry rushes are to actually affect the targeted enemy's ability to put out its first vehicle.

By doing so, you assume control over him. Keeping early pressure on an enemy that's only trying to get required buildings with a minimum of infantry to scout around while he is doing so will equal his death.

It does work in Dawn Of War.

It would also work the other way around. Everyone stays in their corner, building structures up to their first attack vehicle, then war ensues.

Perhaps in SatCraft most of the community plays like that, I don't know for sure. I do own it, but I haven't played it in a while.

Arkhan / Larkin / Thudo and perhaps others around the forum here surely agreed to at least give it a try (the early infantry rushes). It adds a new dimension of A.I capabilities to a genre that has seen so many cliché ways of opperation.

Think about Age Of Empires, or Empire Earth, for example. Or even Command & Conquer. In all those, the vanilla developer's own A.I builds required structures, train a minimum of infantry, make a few searches if possible, then once the first vehicle / cannon / whatever that's not a single infantry arrives, the "true" war starts, and is usually not very long before it ends. I'm not saying it's a bad thing. I'm just saying that we've experienced such A.I behavior in Real-Time-Strategy games so many times.

The DOS A.I modification allows for new things, new strategies. We'll try em' out. Those that makes sense, and especially those that are simply fun to play with. I believe that if the majority of comments from the people around here that enjoy the A.I Mod actually like the early infantry rushes, then there should be no reasons to "take that out of the Mod completely".

I'm sure the coders will find a way to balance / control / limit the whole thing.

And even though right now I believe it is somewhat un-balanced, the early rushes should stay, unless most of us, and especially the coders themselves think it'd be better to try something else instead, and eventually "code it out" of the Mod.

[Malkor]

Vehicle rushes in DoW are primarily for team and FFA games. In Dawn of War, I actually find it EASIER to tech uncompromised than in starcraft. In starcraft, the battle can start one minute into the game and get steadily larger in scale.

In Age of Empires and Command & Conquer, the developer's AI SUCKED. In starcraft, the developer's AI SUCKED. In Dawn of War, the developer's AI SUCKED. The only game I've seen have decent computer players in is in Conquest: Frontier Wars. That's where we come in.

If I was capable of coding I'd be trying to find a way to make vehicle rushes work. LUA is infinitely more capable then the vague crap I have to use for starcraft (I can't even dream of touching micro or abilities or economy related stuff).

And I'm talking about vehicle rushes exclusively. Infantry rushes I've never actually had difficulty defending against in any instance.

Edited by Malkor, 12 January 2006 - 05:32 AM.

[ThetaOrion]

Vehicle Rush

That's a different concept for DoW, though some might call all the IG Sentinels a vehicle rush, it's just that the IG vehicles don't rush to my base, only the infantry.

It's my claim that the infantry rush does not work on the larger maps and should be eliminated if possible on the larger maps.

But . . . on the smaller maps, the infantry rush or Commander rush can be quite effective and annoying and can indeed mix things up. My biggest problem with the infantry rush in the 1.61 betas was that I seemed to always get the same one, the same lame one where the IG were permanently locked into early tier rush mode for the whole game on a small map, and to be honest, that kind of sucked.

If they are going to do the rush, whether a vehicle rush or an infantry rush, they should try to figure out some way to make different types of rushes come into play beside just the same type of rush over and over again. If you could get a mixture of three different types of infantry rushes out of ten games, combined with three different types of tank rushes out of ten games, and four games out of ten games with no rush tactic employed -- that would mix thing up a bit and make it interesting.

But personally, I really do believe that the infantry rush tactic should be eliminated completely or only used one time out of ten on the largest maps and/or the 6 to 8 player maps.

And, it really gets annoying if the AI uses the rush tactic in every game, especially since the only rush tactic that ever worked for the AI was the 1.40 IG grenade launching rush on small maps, a rush tactic employed in every game that was hated and voted down by the user community because the IG were unbeatable on the small maps. With 1.41, the rush tactic no longer seems to work for the AI, so it is or would be kind of annoying to have it employed most of the time in the AI Skirmish Mod.

[ThetaOrion]

And even though right now I believe it is somewhat un-balanced, the early rushes should stay, unless most of us, and especially the coders themselves think it'd be better to try something else instead, and eventually "code it out" of the Mod.

--

I wouldn't necessarily vote to completely eliminate the AI rush, except on the largest maps.

But, I also don't want the AI rushing in every game.

If this were a democracy, my vote would be for them to mix it up. 3/10ths of the time do an infantry rush, but different types. 3/10ths of the time, if possible, do a tank rush on the small or medium maps. And about half the time do no rush at all on the small maps.

Mix it up -- that would be my vote.

But, they need to know that AI rushes often break the AI, so I would also agree with the often revisited claim that completely eliminating the AI rush would probably solve a lot of the flaws in the AI.

However, if they mix it up on the small and medium maps and/or eliminate it only on the larger maps, the AI being broken by rushes won't happen as often or at the times when the rush is the least effective. And, it will keep us all on our toes.

[thudo]

AI doesn't rush every game.. remember.. it randomly chooses the program and 1/4 of them is an All-Arounder. I can see the point:

All-Around or Infantry Tier1 Rushes programs = small maps
Tier2 Infantry or Tier2+ Vehicle Rushes = large maps

Makes sense as the last two programs are too close to the enemy and thus are not ready, militarily, to repress early rushes on tiny maps. However, on large ones the AI has time and breathing room.

[ArkhanTheBlack]

I think there's a bit of a missunderstanding with the rushes.

In principle, there are no rushes at all. The AI just checks if there's a target it can handle, and if this is the case then it will attack. The calculation is for tier 1 - 4 always the same, regardless of race or build program. The so called rush - items in the buildprograms force the AI to build only some specific types of units in the current tier, for example assault marines. Since IG have only one squad type in tier 1 (The command squad doesn't count!), IG tier 1 diversity is pretty much a joke.
The AI's need early troops in small and medium maps, because if they don't have them, the opponent can fast tech without any problems himself, rush the AI or get easy map control.
The current problem is to find a good compromise between early troops and fast teching. In beta 3 the AI's have 4 minutes for tier 1 fun and then stop all unit buidling/reinforcing and fast tech to tier 2. I'm not sure if this is fast enough. Maybe 3,30 min or even 3 min is better, not sure. One problem of the AI early game is to build/reinforce too much useless scout troops. I'm going to fix this for the next beta. I'll also tweak the attack strategy a bit to get rid of the lost AI squads problem.
Fast vehicle rushes are available in 6 player+ maps, but they are definetly not an all winning strategy. I've seen lots of AIs beaten by a combined tier 1 attack of two AI's. Fast teching AI's are dead against such attacks.

Edited by ArkhanTheBlack, 12 January 2006 - 04:55 PM.

[thudo]

I've seen lots of AIs beaten by a combined tier 1 attack of two AI's. Fast teching AI's are dead against such attacks.

Indeed. Whose to say that a pro-player decides one game to invoke the fast-tech-to-vehicles strategy but is them summarily-owned by another pro-player who rushed with infantry? So the vehicle-focused player took a risk and lost. Thats life.

[Malkor]

I think there's a bit of a missunderstanding with the rushes.

In principle, there are no rushes at all. The AI just checks if there's a target it can handle, and if this is the case then it will attack. The calculation is for tier 1 - 4 always the same, regardless of race or build program.

Quite frankly this is a rreeaaalllly sloppy way to do things. Can't you make a custom "attack group" kind of thing? No wonder the AI has so many problems regarding attacking and targeting. There's so many inherit problems with this, obviously the major ones regarding late-game armies getting constantly wasted by ones fifty times their size (I see this a LOT), armies split up all over the map, and the vehicle rushes being completely screwed.

With this known now, I honestly believe your next biggest step is to establish determinable attack grouping with it's own threat detection. This attack group thing would work a bit like this does in starcraft,

train 60 hydralisk
train 6 lurker
train 3 ultralisk
attack_add 6 lurker
attack_add 50 hydralisk
attack_add 3 ultralisk
attack_prepare
wait 2500
attack_do
attack_clear

Obviously with LUA you can do a lot more but basically you get the AI to build what you want, add it to the attack group, group up the units close to the enemy base, attack, then clear the group data if it fails (if it wins you win the game. If it's FFA or team the group moves on to the next major target). This can also let you have "cleanup squads", such as a scout squad that will focus on capping while the main army keeps moving. Having this kind of setup will also prevent workers from getting added to the group, early units being used for "rushing" in a completely retarded situation, and other inanities we tend to witness a lot.

Edited by Malkor, 12 January 2006 - 05:31 PM.

[ArkhanTheBlack]

1.) The AI gathers its troops before it attacks!

2.) Attack targets are selected by a map analysis of a weak spot of the opponent. I wouldn't call that sloppy! But I admit that there are problems because of issues in pathfinding and terrain analysis.

3.) I had those attack groups in 1.6 where I waited for specific unit combinations and then unleashed an attack. It was a mess! A guy named Akaoz showed me in about ten replays that those attack groups are beaten quiet easily by fast techs. I was thinking too in starcraft terms, and that's a horrible mistake, because DOW is completly different.

4.) The worker bug is a 'hardcoded' bug from Relic.

5.) Split armies happen because of the attack moves which stop when an enemy squad is in range. I've improved this behaviour for defense but not for gathering attack troops. I might give it a try for attacks as well. We'll see if it works...

[Malkor]

Well in FFA I see troops all over the place getting shot to pieces by defense, and early on heroes will run off while their squad is attacking an LP and get raped by ork towers or the enemy's more concentrated force...

As for gathering... I only see that happening if their main force isn't completely obliterated and retreats then gets reinforcements and returns. However, though, they get obliterated to the point where if they retreat back to their base they lose (because the enemy just chases them over then blows up their buildings).

Perhaps for the unit groups, use it soley for the vehicle rushes, to force the AI to wait for it's first vehicle before it attempts offensive action?

[ArkhanTheBlack]

Actually, the hint with the gathering problem was very good. I checked it out and not all squads outside the gathering radius got the gathering move command, but instead all squads already inside the gathering radius. Ohh...I'm so ashamed !
I also replaced the attackmove for gathering by a normal move, so the units don't get stuck in combats all over the world.

The vehicle techs normally start their first attack with vehicles, so that's not a problem.

I've further restricted the scout reinforcing for chaos, eldar and SM, and build the armoury earlier. Had a very positive effect on teching speed in my first test games.

Repairing now works in 35 range around the start position and is much more reliable. They aren't disturbed by helping finishing a building anymore, have a higher building check frequency, and repairing works more reliable. They restore all HPs now if they've found a damaged building.

[ThetaOrion]

All-Around or Infantry Tier1 Rushes programs = small maps
Tier2 Infantry or Tier2+ Vehicle Rushes = large maps

Yes Thudo, you get it, except I would want a few more All-Around or non-rush on the small maps.

The tier one infantry rush is acceptable on the 2 player or 4 player maps or the small maps, but with the Beta1 and Beta2, I seemed to be getting it every time I played, which was rather annoying.

But, tiering up or Malkor's vehicle rush sounds like the better policy for the 6 player and 8 player maps and/or the large sized maps.

--

The current problem is to find a good compromise between early troops and fast teching. In beta 3 the AI's have 4 minutes for tier 1 fun and then stop all unit buidling/reinforcing and fast tech to tier 2. I'm not sure if this is fast enough. Maybe 3,30 min or even 3 min is better, not sure. One problem of the AI early game is to build/reinforce too much useless scout troops. I'm going to fix this for the next beta. I'll also tweak the attack strategy a bit to get rid of the lost AI squads problem.
Fast vehicle rushes are available in 6 player+ maps, but they are definetly not an all winning strategy. I've seen lots of AIs beaten by a combined tier 1 attack of two AI's. Fast teching AI's are dead against such attacks.

Yes, finding the right balance for it all or the 'good' compromise is what's needed. It does indeed sound like Arkhan is aware of the 'problems' and is making plans to address each of them. It would also indeed be nice, especially for the Space Marines, for the 'useless scout troops' to be toned down or not reenforced so much with resources.

My biggest complaint is that the AI seems to rush too often. Thudo suggests that only 1/4 of the games are All-Arounders or Non-Rush, and I would prefer up to half the games or 3/4 of the games on the small maps to be Non-Rush.

Get that good compromise on the tier 2 and tier 2 build, and combine that with the AI repairing its listening posts near its startpoint, and I think the next beta will be just about perfect and/or cover everything that everyone is requesting. Of course, somebody will find something wrong with the next beta, such is life. But for now, it sounds like Arkhan has taken to heart what is bugging us most.

--

Malkor sounds like to me that he's wanting a rush that actually works. But, more importantly, he wants a rush that is abandoned if at first it fails.

In the Beta2, the AI did indeed gather its troops for that combined rush into my base -- the problem is that it didn't abandon the rush tactic after the initial rush failed, and then there was a steady line of one squad after another or one TechPriest after another one at a time 'rushing' my turrets and firing listening posts. They didn't group after the initial rush failed, they instead just came one at a time in a lemming suicide march. It was a line instead of a group.

--

I don't know if Arkhan and Thudo will ever be able to find the right combination or the perfect balance that will satisfy everyone, and indeed part of the problem is that each different map provides a different set of problems. It sounds like Arkhan has tried a whole bunch of different strategies or goals or ways of doing it.

Maybe the way to go is to provide a Mix of everything -- sometimes use the current strategy, sometimes rush to tanks, sometimes use the strategy that Akaoz rejected, and sometimes turn off the rush. In other words, keep the human player totally confused and unable to show one replay after another of a locked repeatable strategy. Mix it up.

--
--

Thudo and Arkhan are going to have to think about it. And, this issue is so complex both for coding and in theory, that I imagine that they will be working on it and tweaking it right up to the final version of the AI Skirmish Mod.

Edited by ThetaOrion, 12 January 2006 - 10:10 PM.

[ThetaOrion]

Repairing now works in 35 range around the start position and is much more reliable.

--

When you get time, see if you can get them to repair their own listening posts within a reasonable radius around their start position. Maybe the repair listening posts would have to be a larger radius than the repair-building radius?

[ThetaOrion]

So far, with the Beta 3 games that I have played, I have had only good stuff to report. The person who watches my replays has also had positive stuff to say. The AI didn't turtle up at the point that it normally did, but then it was the Eldar AI that I had instead of the IG. But still, my AI ally kept moving and its moves seemed logical on the Winter Gauntlet 8 Player Map. And, the AI built nices bases and used the ThermoGenerators.

But . . . with the beta 3, I haven't yet had time to get to the small maps where all of these rush issues, and tiering issues, and worker bugs seem to appear and seem to mess things up for the AI.

--

During Christmas, Arkhan took a vacation from DoW, and I took a vacation to DoW. But, I'm back at work now, so now I only have time for one big game per day, or a couple of short ones, if that.

--

So far, the Beta 3 is looking really good for the large maps, the maps that I have been betatesting for a friend.