21 replies to this topic

[ArkhanTheBlack]

This 'should' be the release version.

Changes.

- Some slight race optimizations

- CompassMove deactivated to avoid jump crashes

- Updated docs

[npr]

Been looking forward to your ai release more than the SS hotfix. Seriously, you people are my favorite people on the internet right now

[thudo]

Up for the testers!

[dreddnott]

Earlier today I goofed around with CompassMove and Larkin's suggestions but no combination of functions would prevent Necrons on Eden from crashing.

I just ran 16 games on Eden with this release candidate, no crashes and good performance from the AI in general. Sometimes the Necron AI would put a power generator in a spot where the builders would all get stuck. This basically froze everything for Necron AI except basic unit production - no teching and very little reinforcement occurred.

Interestingly this was the only circumstance that could allow the IG AI to beat Necron AI. IG suffers from reluctance to build turrets (they need them more than other races), Tactica Control not being built early enough (for grenades), and a constant failure to do Attack Moves (walking into or even past Necron Warrior squads is suicidal).

I also noticed that you took dancing out of the Necron Warrior infantry tactic, I'm not sure this is the best idea: Necron Warriors still do great damage while firing on the move, unlike basically any other unit, and as long as the dance distances are very short (long dancing distance is still a problem with the Soulstorm incarnation of the Skirmish AI mod) their slow speed shouldn't be a problem.

Of course Necrons in particular seem to be the most easily used by the AI - there are always plenty of infantry (Warriors and others) massed at every point of the game, something that the other races always have trouble with. Also since Necron Warriors do such great damage while firing on the move, Moving instead of AttackMoving isn't as much of a problem.

[LarkinVB]

Necron dancing is away from CC squad, therefore fire on move won't help them nor will they outdistance the enemy. This is all theory though.

[dreddnott]

That's true, the way the AI dances is very different from the way I do it (back and forth to keep enemy in range of other friendly squads).

Either way I don't think Necrons are going to suffer a serious disadvantage when they're up against other races.

[Zenoth]

LarkinVB, I'm curious, did you get any crashes with Beta 5?

[dreddnott]

"sisters_leader_melee_upgrade_1" is still in the darkeldarbuildbasestrategy.ai file. Doesn't really have any ill effects but clogs up the console with a scary-looking error.

[thudo]

@Dreadnott - I checked the SoulStorm code for the Sisters.. there is a buildable "sisters_leader_melee_upgrade_1" and "sisters_leader_melee_upgrade_2" so totally unsure why you could be getting console errors?

I also played a few games with Sisters.. no console errors so have no clue whats going on with your build? Anyone else? I can't reproduce this so I assume its his copy.

[dreddnott]

Now check the darkeldarbuildbasestrategy.ai file - when Dark Eldar try to research it, it doesn't have prerequisites and dumps errors in the console.

Also: Wyches are still freezing in place instead of charging into close combat, and it's still because of the attempt to use combat drugs in the darkeldarinfantrytactic.ai file. I guess there's no real fix for this other than removing the ability, is there?

Edited by dreddnott, 05 April 2008 - 11:52 PM.

[thudo]

Ahhhhhhh ok.. Bahhhhhhh! Not sure what Arkhan had put there by accident. It'll be looked at in the next build..

[ThetaOrion]

LarkinVB, I'm curious, did you get any crashes with Beta 5?

--

Well, I'm curious to know if this Beta 7 is Crashing on Broken Lands and Zenoth's machine, if Zenoth can get a moment in between his beers to give it a go.



Need another one of Zenoth's detailed reports to make this all complete.

[Zenoth]

LarkinVB, I'm curious, did you get any crashes with Beta 5?

--

Well, I'm curious to know if this Beta 7 is Crashing on Broken Lands and Zenoth's machine, if Zenoth can get a moment in between his beers to give it a go.



Need another one of Zenoth's detailed reports to make this all complete.

'Sup Theta!

Don't worry I'll be preparing a detailed report tonight, I hadn't much time to test since the past 48 hours or so, except for some Skirmishes in Broken Lands. The thing is I kept crashing on Broken Lands since Beta 3 up to Beta 5 for the reasons we all know, Necron Warrior tactic being the recurring lines from the A.I logs after a crash. Then Beta 6 arrived, and the crashes stopped. Believe me guys, they really did stopped, I must have played a total of 15+ Skirmishes (excluding in the Campaign) in Beta 6 on Broken Lands with 2 to 3 Necron teams, and it just refused to crash.

Which is why I thought it would be safe to "announce" that Beta 6 was finally stable because I assumed that since I was the tester with the most amount of crashes related to that bug then if I finally was able to play multiple games without problems it would almost automatically mean that the build was finally stable and that Arkhan indeed fixed whatever was needed to be fixed. But I was wrong... is that surprising?

LarkinVB then experienced a crash, and then another... and he could reproduce it easily. Now I really don't understand what happened in Beta 6, why does the code felt like it didn't want to crash on me anymore but looked at Larkin like its enemy, so hey it said "why not give that guy some problems and leave Zenoth alone, after all he has good tastes when it comes to beers!". So far I haven't tested Beta 7, but I'm about to, tonight I will do so, and I'll once again try a couple of Broken Lands Skirmishes to make 100% certain... and even when I'll believe that I'm 100% certain that certainty will certainly translate into something like 1% because as lucky as I am someone or who knows perhaps something will come here saying "meh, I just crashed with this build" in turn completely destroying what's left of my reputation not as a useful tester but a good beer drinker...

[thudo]

Tested on 2p_Eden 1vs1 with 2 AI Hard Necrons.. 5 Games. All details max, AI Donations + Heroes + Annihilations = No Crash
Tested on 4p_Broken Lands 4player Hard AI Necrons FFA.. 4 Games. All details max, AI Donations + Heroes + Annihilations = No Crash

Tested on Beta7!

Oh and.. Email notifications are being sent again! Hooraarryyy! Makes getting responses in faster!

[Zenoth]

Players: SM (Me) + Orks, VS. DE + Necrons
Winner: Me + Orks
Map: Mountain Trail
Difficulty: Hard
Resource Rate: Standard
Resource Sharing: Disabled
Heroes: Enabled
Massive Battles: Enabled
Fortress: Disabled

Additional Comments: One of my best matches since I test this DoS version, because for once (although it does work quite regularly) the full potential of my "forward defenses" strategy was definitely at its best here, and thanks to the suicidal Dark Eldar rushes against my turrets they ultimately payed the full price (hey, who said that being a Dark Elf, an Elf nonetheless had to automatically make you intelligent?). Those foolish attempts resulted in the death of their Mandrakes, Warriors, and eventually their Archon too, and because they had to replace all of those poor souls and didn't have enough Requisition for that and researches they just basically destroyed themselves before I put the final blow. I didn't attack them all that much, which is the actual goal of my tactics with the strategy I employed, and it was a risk too, but it payed very well.

Overall the match was won by two things, being from what I've observed A) my forward defenses strategy and by B) the Dark Eldar's apparent inability to stay away from obvious critical danger, which as mentioned above resulted in a massive troops loss in early Tier 1, and yes three Mandrake squads and a Warrior squad along with an Archon, all of them perishing before Tier 2 is what I like to call a massive loss. Fortunately their ally, the Necrons, played satisfyingly well. And my own buddies the Orks also did quite a nice job at both capturing and holding territory, and at staying away from great threats. There wasn't too much gathering either, overall well balanced even for the DE, in this build it seems good, I believe that Arkhan/Larkin/Thudo did decrease the army strength requirement/requisite for gathering behaviors, so it certainly helped (although that might well be dating from Beta 6 and not this one, but this is where I mostly observe the changes so far).

What should be "blamed", I'm not sure. In fact I've seen the DE play much worse before this build. And I doubt that I have to seek something to blame each time I win a match, because this A.I isn't meant to be irresistible and immortal to human players. If it can be beaten it doesn't mean 'x' and 'y' has to be tweaked or "fixed" for a millionth time. The suicidal runs by their troops in this match however particularly struck my mind. The thing is I'm not referring to the actual chosen target here, remember guys (Arkhan, Larkin, Thudo) that now I do remember about the fact that targeting selection is hard-coded, that's fine (well it's no it's not, but you get my drift, as long as this A.I is concerned, it is, but as long as hard-coded stuff impede on the potential of this A.I and the game overall is the "not cool" part). The actual problem is that there's an obvious threat and they stay and fight, but because those turrets tend to shoot bullets and because bullets have quite a bad tendency to inflict mortal wounds fast enough the DE kinda didn't mind all that much... well hey I'm not the one saying "I thirst for violence!", the Archon says that.

1) Suicidal runs by DE, not recognizing threats levels?



Target selection is hard-coded, but is the threats-checking code also hard-coded? I believe not? If not then I highly suggest revising that specific part of the DE code to see if there is indeed anything that could be improved, or not. If not then I guess it had to do with a map-specific issue, or that along with a pathing issue, but I doubt it. My reasoning is the following: the Warriors, Mandrakes and even the Archon blatantly keep fighting while their numbers are being eaten by the turrets alone, the Archon does not retreat "for good" to regroup, he did retreat for a very short distance, immediately came back and stubbornly wanted that LP, builder or turret dead, however as mentioned above, the problem isn't the actual target he attacks, but the fact that his health is going down and doesn't retreat to safety... same thing goes for the troops, namely the Mandrakes and Warriors.

Briefly said, see if there's anything to potentially tweak for retreating behavior and/or threats-checking.

2) Building of upgrades research structures at LP's, what's the current status about that?





That happens quite often, if not always honestly. I mean on pretty much all of the maps I've played, the A.I will build such structures some times at close LP's/Relics, but also will do so at distant ones, depending on the map. And also depending on the actual map's layout it could force unnecessary exposure and vulnerability due to them being in close reach of the nearest enemy. In Mountain Trail it ain't that big of a deal, but on other maps like Testing Grounds that'd be another story. I just prefer seeing all of the important buildings as much as possible built inside the main base, just like it was in Dark Crusade.

And here's a little contradiction...

Why building a Summoning Core at the main base (as expected and certainly desired)...


If then you build a Greater Summoning Core at some random LP...


Can the A.I building location decision be controlled? If it can I would humbly suggest to revise that part of the code, and for all races I fear, they just all do that without exceptions. Some buildings are welcome at LP's, or perhaps not "welcome" but certainly tolerable and of not much relevance, but an expansive building of Tier 1 requisition should not in my opinion be built far from the main base due to obvious reasons. If at all possible and feasible, I suggest to take a look at this Arkhan/Larkin/Thudo. But if I am just wrong about all this and you guys are all right then please let me know (this isn't meant to be offensive towards any of you, I'm a tester, not a balance expert, I just want to legitimately know if either it can't be adjusted or if for instance you guys just have not enough tangible proof of any "lack of efficiency" for the A.I team when it occurs during matches to be tweaked at all). I'm bringing this point because I did mention this exact issue a few builds ago, I believe during Beta 3 or 4, I'll have to check about that.

Closing Comments: Well this is what I've observed from that match so far obviously, I certainly have more to look at and perhaps am just not aware of it right now, I need to watch the replay more, or just start a new Skirmish soon. But overall, for that one, the only "poorly performing" faction was the Dark Eldar. Sincerely, I saw nothing I'd call "wrong" about the Orks nor the Necrons, they both fought themselves with regularity, and the Necrons helped the Dark Eldar when I started to assault their main base. The Orks also helped me a little earlier in Tier 1, which shows that that behavior works properly and with a fast response time. I saw no capturing issues, and the building units didn't idle for no reasons (I.E idling only when idling is expected to happen, that's ok then).

Finally I can already see, and sense that this build is indeed worthy of its own Release Candidate title.

We're definitely close guys

Attached Files

•  4P_MOUNTAIN_TRAIL.2008_04_06.00_35_45.rar   171.28KB   3 downloads

Edited by Zenoth, 06 April 2008 - 07:18 AM.

[LarkinVB]

LarkinVB, I'm curious, did you get any crashes with Beta 5?

I had no crashes up to B6 but I didn't test much and not on 'critical' maps. I found Rokclaw foothills to be the best map for fast crashes with B6. I could get it to crash just after gamestart with necrons by forcing the buggy part of code.

In my honest oppinion I think the cause of crash is in since the use of CompassMove(). It is just that it will only crash in very rare circumstances on certain maps. I remember that I had similar crashes with DCpro last year while testing some new generator placement code for necrons.

My theory is this: CompassMove() is taking a jump squads position and will calculate 8 positions round this at jump distance. These positions will then be sorted, nearest to hq first etc.
If one of these positions is out of map bounds and tested for validity by CanJumpToPosition() it will crash. This is very rare as usually the position sorted by distance to hq are on the map and will be used before checking the buggy one. Therefore a crash will only happen while :

a) squad is forced to jump
b) one of the 8 positions is of map
c) the positions before this crash position are not valid.

Edited by LarkinVB, 06 April 2008 - 07:35 AM.

[LarkinVB]

@Zenoth: Got a sync error while trying to watch your replay. Guess you have some extra mods active.

I also watched some instances of T1 squads trying to take down an LP2. I think we should raise the defense value of turrets and LP2/3. Afaik the defense strength value is currently 300. 400 to 500 is possibly a more appropriate value.

[Zenoth]

Thanks for your reply concerning the Beta 6 crashes, it's indeed quite interesting. Basically as I can understand it, if this A.I could recognize the map's bounds then it could freely use any jump code without crashing, right? I can understand why it crashed on Broken Lands, because there's also lots of impassable terrain, high hills and curved places, that map basically is just not flat at all, or almost (only at the center it is). At least you've come up with those three conditions under which a crash can occur. But like Arkhan said, we never had a problem about that in Dark Crusade, it's a shame, Iron Lore had to touch that part of the code and maybe they just broke it under its hard-coded part.

About the sync error that's odd... I mean I'm using Beta 7, it's the same than yours I'd guess? The only mod I'm using is Dawn of Skirmish, nothing else, but I'm using Massive Battles and Heroes (Fortress is Disabled), maybe try to activate those in your options before trying to load the replay? I'd not see why it wouldn't work really

[LarkinVB]

About the sync error that's odd... I mean I'm using Beta 7, it's the same than yours I'd guess? The only mod I'm using is Dawn of Skirmish, nothing else, but I'm using Massive Battles and Heroes (Fortress is Disabled), maybe try to activate those in your options before trying to load the replay? I'd not see why it wouldn't work really

I do not use massive battles nor heroes. I don't have them installed as I don't like them.

[Zenoth]

About the sync error that's odd... I mean I'm using Beta 7, it's the same than yours I'd guess? The only mod I'm using is Dawn of Skirmish, nothing else, but I'm using Massive Battles and Heroes (Fortress is Disabled), maybe try to activate those in your options before trying to load the replay? I'd not see why it wouldn't work really

I do not use massive battles nor heroes. I don't have them installed as I don't like them.

That's ok but is that even the cause of the sync error in the first place? I was just suggesting that it might be but I'm not sure.