62 replies to this topic

[thudo]

We need to chat in MSN x 50 ! I have too much to say regarding your new proposal which I love. You need to see what has transpired since last Summer. I've already accomplished what you've done late last summer BUT you probably have a far far more refined and optimized system. I even have the "link points" that cause "foreign mods" to enter into the AI mod and be added/removed with ease. However.. we must watch involving this endeavour and not diluting the project's original focus. Most mods are still a) not complete b) not balanced c) offer some wonky gameplay elements (although not a bad thing and VERY few do anything radical - which is good @ a coding perspective).

Its that last point that allows this to be an entirely feasible side-project that has been undertaken for many months. We really need to chat in MSN.

[Quietdeath]

my msn: quietdeath2000@[antisp4m]hotmail.com

[ThetaOrion]

THUDO:
1] I'll re-instate the Ubers units for STANDARD on the next build. No problem there.

2] STANDARD skill will force the AI to tech a little slower due to the "Techbreak = 2" flag being set to "2". Its only slightly slower. STANDARD also has 2 buildprograms available to it whereas HARD+ has 4.

Thank you. I definitely want the Ubers back in the next build for the STANDARD setting.

And, I won't mind if the STANDARD AI have the full 4 BuildPrograms, especially if that gives them more variety and a chance to mix up their build order. However, if FOUR BuildPrograms somehow makes them harder, meaner, nastier, more agressive, or more rushy, then the 2 BuildPrograms for STANDARD are fine. I don't know what the extra buildprograms do. What do BuildPrograms do? Anybody?

Teching a little slower due to Techbreak = 2 is fine with me, because I think that's what is actually giving them more of that STANDARD feel than what we were getting before.

Edited by ThetaOrion, 25 January 2006 - 12:13 AM.

[ThetaOrion]

ARKHAN:
BTW, the jump troops problematic is the same as the deepstriking problem. They are too only allowed to jump in range of the target position. If it's not in range...well, bad luck! Never though that such a little change can have such drastic consequences...

I thought about this last night. I don't know if you are talking about the Jump Troops of Flying Troops as well. But, I think there is something else going on besides range, unless the range thing is global and affects all the troops.

In beta 3, the Raptors are standing there right before the gates and then they fly over the gates inside the castle or they fly up onto the gate overlooks right next to where they are standing to attack the IG troops.

But, in Beta 4, the Chaos Raptors aren't flying. I don't know if it is because I'm getting IG grenades sooner and keeping the Raptors on their butts or forcing them to flee by killing off too many of them. But, in beta 4, the Raptors aren't flying. They just stand there, get beat up, and then flee.

Whereas before in Beta 3, the Raptors were flying onto the overlooks or over the gates into the castle soon to be followed by deep striking Horror Demons and/or Obliterators. There's no deep striking follow up in Beta 4 with Horror Demons and Obliterators, even though the Chaos troops before the gates and next to the overlooks should be able to see onto the overlooks and onto the other side of the gate.

Edited by ThetaOrion, 25 January 2006 - 12:14 AM.

[ThetaOrion]

Zenoth: I won [STANDARD], but I believe I did so quite easily. A bit too easily actually.

I usually only play the STANDARD setting with a layout or a scenario that I know I can't win at the HARD setting. After trying it a handful of times a the HARD setting and losing every time, I then try the same layout or scenario at the STANDARD setting, and then the STANDARD setting isn't quite so easy. Also, if I win it at the STANDARD setting, sometimes I can use some of the same tricks to then get a win at the HARD setting, and sometimes not.

BTW, your test was brilliant. We sometimes make some of out greatest discoveries by accident or by not trying to win. I had a feeling that the AI wasn't really rebuilding its base late-game after properly defending its base, and you kind of confirmed it or proved it. I never thought about purposefully backing away and giving the AI the chance to rebuild. Kind of a genius thing to do there, if you don't mind my saying. Sometimes the smartest tests are the simplest.

[Excedrin]

I tried v1.65b4 last night, and it seems not quite right. I'm not sure if the intent is to make it play well in team games exclusively, but in 1v1, it really doesn't try to fight for the map at all.

On Fallen City (IG vs IG), I had captured 6 points almost immediately (2 of the AI's points) with no resistance. In another game with the same settings, I attacked the first point it takes immediately, and it chose to abort its attack and come back to defend. The result was, I took away the first two points it capped, then started destroying its tech (armory, inf command, gens) because it was too far away to get back in time.

I think that at some point, the AI needs to be more geared towards building up small advantages and keeping them. Then turn many small advantages into a win. Otherwise, it seems either too aggressive or too defensive and basically doesn't react properly to many situations.

Other people have mentioned sending the entire army home to deal with a single scout squad, and I've observed similar behavior.

I recognize that improving those areas is difficult, so please don't take this as me bashing your efforts, the skirmish AI is constantly improving and it's come a very long way.

One more thing, Thudo, you should come back to #dowmods! It's fun to idle and occasionally read some AI related backlog.

Edited by Excedrin, 25 January 2006 - 01:18 AM.

[Zenoth]

BTW, your test was brilliant. We sometimes make some of out greatest discoveries by accident or by not trying to win. I had a feeling that the AI wasn't really rebuilding its base late-game after properly defending its base, and you kind of confirmed it or proved it. I never thought about purposefully backing away and giving the AI the chance to rebuild. Kind of a genius thing to do there, if you don't mind my saying. Sometimes the smartest tests are the simplest.

Thank you ThetaOrion.

Well, as I mentioned to Thudo, before being accepted within the Beta testers team for Dawn Of Skirmish, I was an "un-official" tester for a guy nicknamed Ballbarian_SW from the Star Wars Galactic Battlegrounds Heaven's community (and web site of course). I tested the very last of his modified A.I builds for Clone Campaigns (Galactic Battlegrounds' expansion pack). The A.I in question is named after his own nickname, Ballbarian_SW A.I.

In those "tests", one thing I remember I usually did was exactly what I did in that replay.

Stand back, surround the enemy, give it time, so the A.I can potentially go through all the scripts/orders and whatnots possible.

I thought it was a reflex to do it for Dawn Of Skirmish as well, even though the complexity of the A.I in this case is on a much higher level. I wasn't sure of the results. But it seems it does provide a good opportunity to test the A.I behaviors that way. Well at least to test the early builds behaviors in that very case in the replay.

[thudo]

One more thing, Thudo, you should come back to #dowmods! It's fun to idle and occasionally read some AI related backlog.

I'm on MSN now during the day speaking with various mod teams here and there. Hopefully, Arkhan and I will be on there regularily chatting about something PRETTY MAJOR he just did to the AI and thus extending the life of the project considerably. Have lots to discuss. Leave me a PM if you want to chat. I'm in Eastern Canada so ~ -5hours GMT. Arkhan is -8hours to me so my best bet chatting is in the early/mid afternoon for me which shouldn't be a prob. I can't use IRC @ my client's site.

And Excedrin is about a pro-player as you get and he's also a coder too.. lucky combo! heheh.

We're listening people..

Btw, I still ask: why do you guys play STANDARD if the HARD setting is 1:1? STANDARD retards the AI even before my changes as it has a slower resource rate thus you can outproduce it. However, you still find it challenging too which is interesting. I would figure HARD to be the optimum setting as the AI does not cheat. HARDER+ and the AI gets resource bonuses thus can outproduce you.

Anyway.. would love to chat with all on MSN. We get alot done on there.. ALOT.

[Zenoth]

Hopefully, Arkhan and I will be on there regularily chatting about something PRETTY MAJOR he just did to the AI and thus extending the life of the project considerably.

Can't wait to hear more about that !

And, here I present my third test, with a replay.

I didn't notice any "unknown" issues / malfunctions. I rather liked that game, simply because I was very well backed by my Ally. Something I haven't seen since a while.

It was a 2 Vs 2, in Refinery. Me (Space Marines) and my Ally, the Imperial Guards, against the Eldar and the Orks. The difficulty setting was Hard.

I'd simply like to point out the good performance (I personally believe they played well) of the Imperial Guards.

Basically, what I did on my side, was to make sure I survived, and kept my starting location clean of enemy attacks, so I could regularly build structures and make my researches.

I wanted to see if the Imperial Guards could make it on their own against both enemies.

I helped the IG faction on just a few occasions, but I didn't help them repel enemies from their base, as I would have expected, but simply gave them some support by attacking the Orks with them.

They managed to build their base, secure LP's and Critical Locations, all that while regularly pushing in Eldar's territory, as well as taking care of the roaming Ork groups here and there, which in turn helped me maintain an effective defensive position.

I simply waited for them to whipe out the Eldar before attacking en force as well against the Orks at the end.

Watching the replay proved me that Beta 4 indeed helped the IG, perhaps more than actually intended ? All I know is that the IG would have never managed to do that in the previous Beta builds, much less on the Hard difficulty setting.

There is however something I believe was mentioned by ThetaOrion before. Related to the infantry using jet-packs. In that case, the Orks. I did not see them use the jump capability at all, unless I missed it during the replay.

[ThetaOrion]

THUDO ASKED:

Btw, I still ask: why do you guys play STANDARD if the HARD setting is 1:1? STANDARD retards the AI even before my changes as it has a slower resource rate thus you can outproduce it. However, you still find it challenging too which is interesting. I would figure HARD to be the optimum setting as the AI does not cheat. HARDER+ and the AI gets resource bonuses thus can outproduce you.

Thudo, you need to learn to actually read the responses to the messages.

I have answered this question three times for you, and you don't seem to read the answer or accept my answer. My answer not acceptable?

You ask this question in every thread and I answer it in every thread. I answered it up above in this thread too, but here goes again.

I only play the STANDARD setting with a layout or a scenario that I know I can't win at the HARD setting. After trying a scenario a handful of times at the HARD setting and losing every time, I then try the same layout or scenario at the STANDARD setting, and then the STANDARD setting isn't quite so easy. Also, if I win it at the STANDARD setting, sometimes I can use some of the same tricks to then get a win at the HARD setting, and sometimes not.

I have got to save this link someplace so that next week when you ask why we use the STANDARD setting, I can refer you to it. You asked the question and it was answered last week and the week before and the week before that.

Did you read the whole message? Thudo, now did you really read the whole message? Common, Thudo, I know you didn't read it all, now go back and take the time and really read it, and then read it again.

When we encounter a HARD setting we can't beat, then we use the STANDARD setting to hone our skills, test ideas, and strengthen our muscles -- or we use a win at the STANDARD setting as a consolation prize. It has nothing to do with HARD being too hard, which is something that you keep suggesting in each of your threads where you ask why we use STANDARD. HARD is just right, unless you encounter an unwinnable scenario, like 3 of you vs 5 of them. In fact, the HARD AI are rather soft in Beta 4. No flying Chaos Raptors in Beta 4. The Raptors are all grounded.

Now, I need to go find the other three times you asked the question and didn't bother reading the answer and try to build up a library. STANDARD has its use, so we want it to be good, we want it to be STANDARD, not hobbled or brain dead or chunks of the AI cut out.

Now Thudo, have you really read this answer? Read it again please.

--
||
--

Thudo, here's the link to last weeks response to your question:

http://forums.revora...ndpost&p=236307

There, I tried to explain why we want a STANDARD setting and want it be good. Asked and answered, though I do try to take the answer from a different approach each time.

One of these days I will actually get your attention, and you will actually read the answer. It has nothing to do with HARD being 1:1 or HARD being too challenging for us. YOU ARE LOCKED into your answer and never bother to read ours.

Edited by ThetaOrion, 25 January 2006 - 05:30 AM.

[ThetaOrion]

Arkhan:

I watched a replay where Aralez won as 3IG vs 5Chaos on the Castle Assault Map at the HARD setting with 1.60 of the AI Mod.

After taking lessons from the master and the maker of the map, I then went and duplicated the win with Beta4 of the AI Mod.

In Beta 4, the Chaos Raptors didn't fly. And, about 1/4th of the way through the game, one of my IG AI allies retreated into the Castle and turtled up for about half the game. They didn't move even when their side of the Castle was finally being overrun. I had to lead the charge on their side to get them to come out of hiding, and, then they played normally as I would have expected them to.

I never saw any Chaos BloodThirsters either, though they should have had time to make some.

I think the Allied AI in Beta4 retreats too easily. I don't know what scared the IG ally into a full retreat, but they seemed to be doing just fine until they retreated and turtled up.

Another link to where I discussed my win:
http://forums.relicn...3&postcount=162

Edited by ThetaOrion, 25 January 2006 - 05:44 AM.

[Zenoth]

I must report a crash (yes sadly).

I was playing in Fort Atlantis, at the Hard difficulty setting. The Resources rate set to High. Game mode Annihilate. It was a 4 Vs 4, me (Eldar), with two Space Marines factions and one IG faction. The enemies were two factions of Chaos and two factions of Orks.

I played for a good 15 minutes without any noticeable problems or slow downs, nothing wrong.

And all of a sudden, the game froze for about 10 seconds, and then crashed to the Desktop.

I must say that is was the first time, ever, that the game crashed on me. It never happened since I bought it actually. Not even happened with the vanilla version before I bought the WA expansion. It's really and honestly the very first time it happens to me.

And since the game crashed, I have no saves sadly. It wasn't an "A.I error" that makes your game freeze with the error message on-screen. But a simple ol' style crash to Desktop, nothing else, nothing more.

[ThetaOrion]

I have experienced quite a few crashes to desktop with Winter Assault and even with the 1.65 Betas.

All my most recent crashes to desktop with 1.65 Beta take place when I'm teleporting something through the Eldar Webway Gates. Last time, it was a FarSeer Commander attached to a Seer Council. When I pressed the button to release them from the gate at their destination, the FarSeer Commander said, "We have arrived," the game stuttered or stalled, and then dropped to the desktop where it gives that bug report thing that doesn't work because the other side of the internet connection is dead.

And, it happened another time before that. It seems like I was putting the FarSeer Commander through that time as well.

So, I think the Eldar have a problem or a bug. Were you transporting or teleporting something through a gateway when that CTD happened to you? Was the FarSeer Commander there or being accessed or doing any speaking when you had the CTD?

Recently, though, playing IG vs Chaos, I haven't had a single CTD. Recently, it seems like I have to have the Eldar there or be playing the Eldar in order to get a CTD.

Edited by ThetaOrion, 25 January 2006 - 08:50 AM.

[ThetaOrion]

Thudo Wrote: STANDARD also has 2 buildprograms available to it whereas HARD+ has 4.

Arkhan, what did he take out? What two buildprograms did he take out? What is a BuildProgram? Is it something we are going to miss? Is it something essential? Is it something we should fight for? Or is it something we should let go?

Thudo Wrote: I'll re-instate the Ubers units for STANDARD on the next build. No problem there.

Now that I know what an uber is, I'm glad that it will be reinstated in the next build, because I know that I want them to be there in the STANDARD setting as well. In fact, Squiggoth Ubers not being built in Beta 3 at HARD setting and BloodThirsters not being built in Beta 4 at HARD setting is quite noticeable to me. I want the ubers to be there at all difficulty settings, except maybe EASY.

--

Does anybody know whether the Beta4 Orks are building squiggoths? I have seen IG AI BaneBlades in Beta4, so I know that's working the way I like it.

[ThetaOrion]

Thudo Wrote:
STANDARD skill will force the AI to tech a little slower due to the "Techbreak = 2" flag being set to "2". Its only slightly slower.

Thudo Wrote January 16th: I think a slightly slower teching STANDARD (using Arkhan's "TechBreak = 2") should not break the AI so its not as competitive. Thats the ONLY change I'm gonna recommend. We will test and if the AI stinks then we'll regress back.

--

This I agree we do want, because I think it is the thing that is giving the STANDARD setting the feel and the pace of a STANDARD setting.

[Zenoth]

ThetaOrion, the Seer Council wasn't trained, and the Farseer was being trained when it happened. That, I remember clearly. I had only a few basic troops, but no "special" units at all. Not even a single vehicle. And I must also mention that I did not use the Webway transportation ability.

And I was used to play as the Eldar, in Fort Atlantis (it's my personal favorite when it comes to 4 Vs 4's). And as I mentioned, I never had a problem in that map. All races included, all difficulty settings considered.

I'm still trying to figure out what went wrong ... but I can't. It's just ... you know, it came out of the blue.

[Zenoth]

Ok ...

Arkhan, Larkin, Thudo ... I believe there is something wrong with this build.

The A.I behavior I have reported about that game I played in Black Fortress, you know, the "back and forth" movement between short distances, and the fact that the A.I did nothing else but increase their army size and variety up to their limits, and still remained in a specific spot, and refused to attack each others for hours.

Well, it happened again ... but that time it was actually way too obvious that it shouldn't have been like that.

I just played a game in Fort Atlantis (I had a crash issue earlier in that map, but I tried not to play as, nor include the Eldar, and it didn't crash, so I could resume testing).

And even though the game never ended after a good hour of play, I took screenshots, and I edited them to help me explain you guys what's wrong, because it definitely is.

First screenshot ...



That screenshot was taken some time after I, and my SM ally had both been defeated, so what you see are the remaining two "alliance" factions, north-west on the map, and all the rest being the enemies.

In Fort Atlantis, everyone starts on the edges of the map, and then need to take control of most of the center, with its Critical Location, to then have easier access to enemy bases from their main "front" entrance, and from both sides.

In the center though, there isn't a lot of space for 6+ players and full armies battling. So, it's quite easy to eventually be within range of one or more groups of infantry or vehicles' . So ... "ignoring" the presence of enemies in that map is virtually impossible.

What you see in more details in the screenshot above, are the "spot" that each armies decided to stay at. Within a specific, very small radius, in, and around their starting location.

In the center, most of them were duking it out when they happened to be in range of any possible enemy attack(s). But they always refused to actually fully engage each others. Even though they saw each others clearly and were even being attacked regularly from some distance away.

Basically what happened was that each armies' numbers decreased very slowly, and the dying units and destroyed vehicles were immediately replaced, and went back in the "radius", and were randomly fighting here and there when they saw an enemy.

It was quite a show to look at.

And if that's technically "forces gathering" within the A.I code, then it lasts way too long, or is simply in need of a revision.

Now, the second screenshot ...



Just take a look at the resources, and make it the case for all the A.I factions present.

That's all they do. They build necessary units to fill up their army if they lost one or two units here and there randomly, and nothing else. So the amount of resources dramatically increased over time.

To get some "extra" requisition or power at the end of a game, yes I have no problems with that, and it's even normal. But not that much, I'm sure about that.

So ... while some games with Beta 4 may seem normal in terms of A.I behavior, there is, so far I believe, a problem, and I am trying to explain it in this actual post. We'll need to investigate further, because either I'm getting senile, or there's a little something to look at in the scripts.

And of course I'll need comments on this from everyone involved in the development and testing process.

It'd be much appreciated.

Oh, and ... as I side note, I think it would be important to say that I never experienced such behavior from the A.I in any of the previous Beta or public builds. It's unique to Beta 4.

Edited by Zenoth, 25 January 2006 - 09:04 PM.

[LarkinVB]

I'm not involved in the b4 build, therefore I can't comment but I'm sure your excellent report will help Arkhan to nail the problem pretty fast.

[ArkhanTheBlack]

Well, I experienced the 'let's have a gathering' behaviour myself yesterday with the restructured beta 4. There are definitely several things broken in this build. I don't consider that as a big problem since I know most of the 'guilty' code parts. Beta 4 contains more code changes and additional code than the last three betas together. I also abused it for some experiments. Some of them were successfull, others not. The sad thing is that the not so successfull experiments tend to ruin the entire AI, whilst the successfull experiments only slightly increase the behaviour .
It's a bit chaotic at the moment, since the beta 4 contains a lot of new content with lots of teething problems, and at the same time there's another big change in the code structure which will have major consequences for all other mods using our skirmish AI. But since the restructuration of the AI code is almost successfull completed, I (hopefully) can slowly get back to the gameplay problems of beta 4.

The crash is more problematic. If you're sure that it's related to Eldar, then I'm 90 % sure that it's a problem of the jumps. Some inbuilt AI functions crash, when they get an invalid coordinate. This happens mostly on maps with hight differences. It would help if you could make an AI trace, but this really slows down the game speed. Anyway, if I know the map, I'm pretty sure that I can reconstruct this crash. BTW, it would be interesting if the other testers get crashes, too.

[Zenoth]

The crash is more problematic. If you're sure that it's related to Eldar, then I'm 90 % sure that it's a problem of the jumps.

Well, as I mentioned, it was the very first time ever that the game crashed on me. And I did use the Eldar in Fort Atlantis before, but never experienced such an issue with any of the previous DOS builds nor even the vanilla A.I. But that does not actually mean that others didn't. I believe that, for example, Quietdeath experienced a crash during one of his tests in Beta 3.

ThetaOrion also seem to be the victim of a crash every now and then, and he reported that it usually happens when the Eldar are present, either playing as them, or that one or more of the A.I factions present is/are also Eldar.

Today I will try again to "replicate" the crash, with the very same settings, and I'll try to take note of everything. But if it crashes then of course I'll have no replay.

BTW, it would be interesting if the other testers get crashes, too.

As I said, I believe they did experience one or more crash in previous builds. But any testers correct me if I'm wrong.

And, Arkhan, I'd like to point out something that I just noticed ...

Related to the gathering behavior.

Now that I think about it ... you know what ? That behavior happened after one or more factions was/were defeated. I'm not sure if there is a link about that and the behavior itself. But it'd be important to take note that at the beginning, up to some point in the game, everything looked fine, and as soon as the Eldar died in the Black Fortress game, then the behavior started. And, in that Fort Atlantis game, is started after I and my SM ally were both defeated.

They must have not gathered any specific forces so they could actually come to my base and fully engage my own forces. Or if they did, then they did it properly, and it didn't last long at all.

I believe there's some sort of pattern there.

All the gathering code works fine, until perhaps (not sure, just an idea) one or more factions are defeated. Then ... something wrong ensues.

Edited by Zenoth, 25 January 2006 - 09:26 PM.