66 replies to this topic

[ArkhanTheBlack]

Will I just have to change the AIUnitStartingRank struct on the host or should it be changed on all the clients?

I guess every host runs their own SCAR script, therefore you'd have to change it for all hosts.

Anyway, I have just tried disabling Heroes and then Harder works just fine in an mp game with two players and two AI opponents. smile.gif Have you tried running a mp game on Harder on your LAN setup?

Not really surprising. SCAR code is the trouble maker number 1 if it comes to sync errors.

[thudo]

Set them all to "1" and see if the synch error occurs? If not.. Hmmmm.. I guess we'll have to keep starting units on HARDER skills and up @ the same XP levels as HARD and below. Still.. AI gets enough of a boost with resouces on HARDER+. However, let me test on my end in my LAN to see what happens. I'm curious..

[Inquisitor]

I guess every host runs their own SCAR script, therefore you'd have to change it for all hosts.

Ahem, there is only ever one host in a mp game hence the name "host". In other words: "there can be only one!"

I fear that what you're saying is that I have to change the AIUnitStartingRank struct on all my friends computers...

Set them all to "1" and see if the synch error occurs? If not.. Hmmmm.. I guess we'll have to keep starting units on HARDER skills and up @ the same XP levels as HARD and below. Still.. AI gets enough of a boost with resouces on HARDER+. However, let me test on my end in my LAN to see what happens. I'm curious..

Yes, that seems to be the conclusion on this. It's a bit discerning, since I guess it was the main reason for including the Heroes mod, namely to make the AI opponent better on more levels than just resources. I think it's a bit supsicious that Heroes run perfectly on several severely lagged online games, but not if the AI starts with Veteran troops... Maybe ArkhanTheBlack can work his magic and find a bug?

[thudo]

Well the HARDER+ w/ Heroes is still bloody hard regardless. Further, 9/10 people who play our AI mod play on HARD and lower as most are average players and not pros. HARD is more than likely too hard for most.

[LarkinVB]

Quatra, 2 eldar vs 2 ork (all Ai)

Third ork squad should be shoota instead of slugga as AI is still bad with CC and sluggas are basically useless if not handled well (which the AI can't achieve). I did the change and orks could fight back much better. First ork vehicle is still a killa kan which is a bad choice vs eldar. It is too slow to have great impact. No early trucks, no trakks. At least not during the time slot in which they are effective.

[thudo]

Thanks Larkin.. that can be quickly tweaked.

[kreative]

Are there any tactics unique to certain maps that are written for the AI?

[thudo]

@Kreative - not that I'm aware of. We don't specifically code for maps of a type. We do have special code to avoid initial build problems on certain startpoints.

@Larkin - ok.. the 3rd initial squad will be Shoota. I also managed to get the Wartrakk out more often as well.

[kreative]

Being a new member of the beta tester squad, I find myself coming up with new features rather than looking for wrong build orders or bugs. I'm sure this will change in time though, just takes practice like everything else.
However, given the lifetime of this mod, and the state it's in, it's more adjustments, and not fundamental changes in the structure of the mod.
Not yet...
With the hopeful release of DoW 2, I'm pretty sure the AI will be just as challenging as before, and the skirmish mod will live on.
So I will be jotting down some thoughts on improving this mod further if that's ok. By the time DoW 2 eventually hits the shelves I will also be able to contribute with programming.

[ArkhanTheBlack]

I also agree with the third shoota squad. I didn't know that you can get them fast enough to not interrupt the early capturing process. However, I don't want the Kan to be changed. Maybe he wasn't as effective it this special confrontation vers Eldar, but in general, the early walkers rule. I tested this last year in lots of scenarios.

[thudo]

KillKans still come out alot.. now followed by Trakks.. and now occassionally Trukks.

I have a question: I constantly watch the Ork AI and noticed all the time it never tries to maximize its sqaud/support cap by trying to build the maximum # of infantry/vehicles at certain times of the game. ie. Just before the Ork HQ_Addon_1 is built, the Ork AI still has half of the support cap open so it can build more vehicles but never normally does UNTIL sometime after the HQ_Addon_1 is built. Is this MinArmyStrength related? Not too big a deal.. just wondering as sometimes I'd love to see the AI give priority to building troops/vehicles.

[ThetaOrion]

I'm really glad to see Inquisitor doing multiplayer tests with this mod and pinpointing any further potential multiplayer bugs with the AI Mod. I always kind of felt that this team needed a betatester covering the multiplayer angle, and now you guys seem to have someone!

Gee, you even have someone who is looking into the save and reload issue!

Good going guys!

A lot of people excited to see this one!

Of course take you time to get it right, as it's going to become the core of everything else that is yet to come.

[ArkhanTheBlack]

I have a question: I constantly watch the Ork AI and noticed all the time it never tries to maximize its sqaud/support cap by trying to build the maximum # of infantry/vehicles at certain times of the game. ie. Just before the Ork HQ_Addon_1 is built, the Ork AI still has half of the support cap open so it can build more vehicles but never normally does UNTIL sometime after the HQ_Addon_1 is built. Is this MinArmyStrength related? Not too big a deal.. just wondering as sometimes I'd love to see the AI give priority to building troops/vehicles.

Hmm, I wasn't aware that maximizing squad and support caps is the new winning strategy?! I don't think players would be impressed if we send huge waves of cultist squads against their bases.
The troop building limit is the min army strength and the force tech times. At the moment the AI tech very reliable to tier 4 in about 20 - 25 minutes. I don't think that's too far away from human players.
Huge armies are only usefull if they consist of powerfull (=high tier) troops. If they get killed to easily, the player gets a tech advantage and tank rushes the AI. The economic control is the only discipline the AI is often better than a human, therefore this advantage should be used.
Sure, you can max out all caps, but since the AI will then be constantly attacking, loosing troops and try to replace them, I'd really like to know how the AI should ever reach tier 4 with this strategy.

[thudo]

No no.. leave it the way it is since we've had no complaints specifically on this. Was just wondering. I was more wondering why the AI waits for a major addon to be finished before building units? It works the way it is so all is good. Otherwise, our betatests would have had some observations directed to it. Thanks Arkhan!

[Inquisitor]

I'm really glad to see Inquisitor doing multiplayer tests with this mod and pinpointing any further potential multiplayer bugs with the AI Mod. I always kind of felt that this team needed a betatester covering the multiplayer angle, and now you guys seem to have someone!

Thanks, ThetaOrion! I'm just glad that I'm allowed to contribute to this awesome mod!

Based on my findings I suggest a change in the help text for heroes (in the Heroes_local.lua).
The text as it is now:

"Squads get battle experience and gain in power. Don't use if you experience heavy lag or sync errors in multiplayer games."

I suggest something like (my english is not the best, so feel free to correct or rephrase it!):

"Squads get battle experience and gain in power. Don't use in multiplayer games on Harder or Insane level as it will cause sync error."

As you know we have played a considerable amount of multiplayer games on Hard with up to 4 AI (8P) and with heavy lagging, so I'm rather confident that the mod is working perfectly in mp - even if it's lagging heavily. We have not had a single problem, so I think the sync problems with lagging games might be a thing of the past, so it would be better if the help text were more precise as to what to avoid.

Please let me know if there are specific areas that you would like to be tested and I'll see what I can do...

[LarkinVB]

eldar vs IG meeting of minds

1. Four eldar guardian squads at gamestart ? I think 3 is enough, than aspects.
2. IG had 12/12 infantry but *never* increased this through to the game up to T4. They did build lots mines, turrets but no additional infanty command (or whatever they do need to increase infantry cap).
3. Two fire dragon squads. I still think one is more than AI can handle. I suggest limiting FD to 1, guardians to 3.
4. CS with general, psyker and priest. When the general was killed the psyker and priest stayed at range stance.
5. FS is attaching while hurt but AI can't handle this as it is constantly toggling FOF on/off. Therefore I suggest to not attach FS but keep her at ranged stance while severely hurt.
6. Just play this setup at this map to observe lots of stuff which can be improved.

[thudo]

1. Four eldar guardian squads at gamestart ? I think 3 is enough, than aspects.

4 initial Guardians is due to the map. If the map is determined to be LARGE then it builds 4 squads. If Small or Basic, then three. This goes for the Orks.. they build 2 Sluggs and then 2 Shootas.

2. IG had 12/12 infantry but *never* increased this through to the game up to T4. They did build lots mines, turrets but no additional infanty command (or whatever they do need to increase infantry cap).

Hmm.. I've noticed this as well. To build addditional guard_infantry to increase squad cap AI needs 300req/1pwr. Perhaps the game you played you forced the AI to draw too many resouces towards the fight and not enough for larger armies. If you increase the squad or support caps but the AI has no money to build extra units anyway since your exhausting him then those cap increases won't do anything anyway BUT if the AI is NOT being attacked and has the time to recover but doesn't build/research those squad/support caps.. then.. well..

3. Two fire dragon squads. I still think one is more than AI can handle. I suggest limiting FD to 1, guardians to 3.

Currently its set as a random between 1 or 2 squads as I've seen 2 FDs squads controlled by the AI really own. Yes its a random chance but I would have no probs going for only 1 FD squad forever.

4. CS with general, psyker and priest. When the general was killed the psyker and priest stayed at range stance.
5. FS is attaching while hurt but AI can't handle this as it is constantly toggling FOF on/off. Therefore I suggest to not attach FS but keep her at ranged stance while severely hurt.

Good observations Larkin! Thank you!!

[LarkinVB]

Hmm.. I've noticed this as well. To build addditional guard_infantry to increase squad cap AI needs 300req/1pwr. Perhaps the game you played you forced the AI to draw too many resouces towards the fight and not enough for larger armies. If you increase the squad or support caps but the AI has no money to build extra units anyway since your exhausting him then those cap increases won't do anything anyway BUT if the AI is NOT being attacked and has the time to recover but doesn't build/research those squad/support caps.. then.. well..

It had enough resources to build another mechanized command (though this cap was not maxed out) as well es multiple turrets and minefields. For me it looked like a real problem. The game went T4 with baneblade and stuff but the AI refused to max out infantry.

It had also queued a HW team which could not go out as max cap was reached.

Edited by LarkinVB, 18 April 2007 - 10:14 AM.

[Inquisitor]

More MP tests
Version: Skirmish AI 2.3b4
Map: Tartarus Center
Match: 2 vs 2 multiplayer via Hamachi
Game Modes Used: Annihilate
AI Difficulty: Harder
Resources Rate: Normal
Resources Sharing: Off
Factions: Necron and SM vs Chaos AI and IG AI
Winner: Chaos AI and IG AI
Replay: Not available
Comments: Three multiplayer games on Harder vs. Chaos and IG. We got our asses kicked in each game, so in pure frustration I did not save the replays . The AI performs great and IG is a pain to meet because they field GL early and my NW has no chance. Chaos was not surprisingly a real challenge with e.g. several harassing Raptors. Great games where the AI attacked together and defended together all the time. Really hard to match.
Observations of interest: In one game I managed to assault the IG base and had him on his heels (I think) only to find one or more CSM squads with HB standing and waiting and then just mowing down my NW down.

Version: Skirmish AI 2.3b4
Map: Tartarus Center
Match: 2 vs 2 multiplayer via Hamachi
Game Modes Used: Annihilate
AI Difficulty: Hard
Resources Rate: Normal
Resources Sharing: Off
Factions: Necron and SM vs Chaos AI and IG AI
Winner: Necron and SM
Replay: Available
Comments: After three times defeat, we needed to reaffirm ourselves and took a game on Hard (no Heroes). I chose a two monolith build and massed NW. We won without too much trouble.
Observations of interest: IG massed a lot of GM squads and plays the CS and psykers really well. IG built two Basilisk but it was too late as I was already in his base. Chaos assisted with Defiler, preds and CL attached to possessed, which took a lot of hits from my ally as they were en route to the IG base because he was assaulting the Chaos base!
Suggestions for improvement: I still think CC units behave a bit weird as they sometimes seem to regret attacking for no obvious reason. For instance I saw a full squad of Raptors assault a full SM squad, fight five seconds taking practically no hits and then jump out of combat back towards the chaos base. I have described the behavior a couple of times before.

I will probably test more tonight as some of my friends have already asked me if I want to. We will test more games on Hard with Heroes on and I will probably also try AI donations, to see if it works. We will be three or more human players.

Question: How about Take and Hold games? Are there any AI logic related to the timers? Would it be interesting to test it in mp?

[LarkinVB]

Question: How about Take and Hold games? Are there any AI logic related to the timers? Would it be interesting to test it in mp?

I think so. But it makes AIs life more difficult so I suggest playing it on harder for a challenge.

Edited by LarkinVB, 18 April 2007 - 03:21 PM.