 |
| |
|
|
Mongoose
|
|
Harrisburg,PA USA
Apr 1999
time: 00:13
|
|
I have only had that type of thing occur during IP starts. In those instances, it was because there were players in the starting screen before the host selected 'new game'. No splash sequence at start, and most factions had units available to build for which they hadn't the tech. I THINK they are designs held over from the previous game. No idea why this might have occured in SP mode, unless the time warp option invokes some part of the MP code.
In any event, there's no other alternative to a restart, at least in IP, because the game is unstable and crashes very quickly anyway.
|
|
| |
|
|
|
Googlie
|
 |
Commanding Officer, CRYPTEIA
Apr 1999
time: 21:13
|
|
Ogie:
yes, I have relished the last couple of games with accelerated start.
Just finished one using KSG (Kinjiru's Static Gambit). Start a normal game, but just keep clicking enter for 30 or 50 turns to give the AI a head start, then get going.
I was Domai, gave the AO a fifty year headstart, then opened my base for business. Within ten turns Yang came calling, and we fought a naval battle every four or five turns, with the odd transport gettting thru and dumping a couple of impact rovers on my soil.
I micromanaged like you wouldn't believe to squeeze every last credit out, then in 2350 or thereabouts got the 20 year warning that Morgan was cornering the energy market, which he did, in 2403.
Did I ever get my butt kicked.
(double blind, transcend, 128 x 128 map, mostly islands)
great game tho, even though I lost to the AI
Googlie
[This message has been edited by Googlie (edited August 11, 2000).]
|
|
| |
|
|
|
Misotu
|
 |
Leamington Spa, England
Dec 1999
time: 05:13
|
|
Ogie/Googlie
I've tried accelerated start and it works well for beefing up the AI, but I'm still trying to find a way of getting a better game without it (particularly since the game crashes on my PC when I've set up two or three accelerated starts. Such a pain.)
I realise that a created scenario spoils a lot of the discovery of the early game (for the creator), but I don't want to edit alpha.txt or get too artificial about helping the AI (ie giving them all my techs all the time or whatever).
So I've been messing with the scenario editor, intending to create a 2-human vs 5-AI game that would be seriously tough to win. Since this is my first attempt with the editor, I completely overdid the help for the AI but when I played through the game, the results were fascinating.
Map of Planet - humans are PK/Gaians in the centre island. Surrounded by UoP, Hive, Believers, Drones, Pirates in an unbreakable five-way pact. The Hive are in the Monsoon jungle, and all the AI factions have three good bases and three formers, including at least one sea former. Pirates started off with one land base, to overcome the usual problems they get themselves into. I chucked a bit of forest around the AI bases to give them a hint about terraforming, too 
I played through quickly to test the set up with an AI pactmate, not putting in too much effort really - and was absolutely slaughtered. The AI did things I've never seen before including:
o It's conventional wisdom that the AI doesn't use choppers. Well ... not in this game. EVERY AI faction, with the exception of the Believers, built choppers. Not only that, the Hive actually built choppers in preference to needlejets and overcame the range problem by housing them in a captured Gaian base. Prior to having a base on my island, the Hive did not return its choppers to a base at the end of each turn and thus achieved the required range. It was quite an experience watching the Hive plink all my crawlers in about 3 turns
Even worse was watching their chop and drop attack that took my first base effortlessly. Three choppers, followed by two armoured shard paratroopers. Which would have been quite fun if my best weapon had been better than missile
o The Pirates built foil probes. Yup. That thing the AI never does. And they built them in serious numbers, sending them in pairs or threes against my bases.
o The AI is usually reluctant to mount a serious attack on human bases using air power. Hah! They all went straight for it. The Pirates started hitting me with squadrons of missile/chaos jets even though I had AAA silksteel/aerospace complex. And they kept going for it until they had taken the base.
Within 100 years, the Hive tried for the Supreme Leader victory, and would have won effortlessly if two of their pactmates hadn't abstained They tried this twice in the game.
All of the AI factions terraformed like crazy and built sensible improvements in their bases, including (gasp) tree farms built by Hive, Drones, UoP. All right, the terraforming wasn't brilliant but the forests I'd seeded expanded well and were not converted into farms by the AI. AI research rates were *stonking*, they took the PKs and the Gaians apart in no time.
So I'm going to have to fine-tune I think But it's definitely showing promise if I can get the balance a little better. What interests me is how come the AI can use choppers well, and knows about foil probes, but doesn't usually build them? Do they have to be in a massively superior position before they will build these units?
|
|
| |
|
|
|
Beör
|
|
Copenhagen, Denmark
Aug 2000
time: 05:13
|
|
Hear, hear!
|
|
| |
|
|
|
Helium Pond
|
|
Los Anheles, California, Good Ole U S of A
Nov 1999
time: 05:13
|
|
Misotu:
Wow, yeah, great findings. What I really wonder is what the critical adjustment is for the AI to start behaving in this way. If you can isolate it down to one or two small changes, it might be possible to alter the faction files or the alpha.txt so that the AI behaves this way in every game!
Here's a thought: is there a way to adjust how might is calculated? If the more intelligent AI behavior is based on the AI's perception of its own might, perhaps one could rig it so that the AI always thinks it is way ahead of you?
(And if it turns out that the unbreakable pact thing is a significant factor, I'm going to laugh at everybody who razzed me for suggesting a SUBMISSIVE parameter be added to the faction files...)
|
|
| |
|
|
|
Misotu
|
|
Leamington Spa, England
Dec 1999
time: 05:13
|
|
Red Fred, Beor, HP: I'd be delighted to provide you with all the details. I'll zip and send you the scenario too, if you'd like it. I'm no expert with the scenario editor though, so it may take me a while to get the balance better, it's *way* out right now.
The other difficulty is that I intended it to be an MP game, rather than SP. My experience with an AI pactmate was not good in the early stages - Deirdre wouldn't exchange techs until the WP was built, and did a lot of dumb things when we were under fire. (Like transferring practically her entire airforce to a single, tiny base of mine, which didn't have the minerals to support all those units). So the balance for an SP game would have to be quite different. Still, I'll give it a go, can't be that hard.
I'm sure that the unbreakable pact thing is significant - I put that in because I remembered Zsozso's account of his fast transcendence game. He reported that, once all the AI factions are submissive and there is no-one to attack, they turn into builders and stop military development. Now that doesn't happen here, of course, as they are in vendetta with the PK/Gaian axis. But, perhaps because their situation is more secure than normal, they do seem to focus on building more than is typical.
My instinct says that the AI behaves differently when it perceives that it is very strong. And see this quote from Stuntman19, posted yesterday, coincidentally, in the Tech Share Challenge thread on AC-Strategy:
"Also it seems the AI for some reason becomes smarter when it is way stronger than you. It built choppers, probe foils, crawlers, bore holes, orbital facilities , and university even built skybases. Most of these I had never seen the AI build and especially in large quantities. I wonder if the AI has to perceive itself as way superior to build some of this stuff? Anybody more knowledgeable have some insights?"
So it seems that my experience with this scenario is not an isolated one.
In Zsozso's Ultimate Builder scenario the human player played Cha Dawn against 6 AI's in an unbreakable pact and the results were quite different. The AI factions did nothing at all unusual when I played it - but then, their lead over me was never as pronounced as in this game. So it could indeed be that "might" is the key.
It would be somewhat ironic if the only way to get a decent game against the AI were to let it win
In which case HP, that's an interesting idea - rigging the AI's perception of its own might in order to produce a more interesting game. I still find it confusing that the AI should play differently because it's winning, but if it makes for a better game, that's fine by me
I'll tweak the scenario and play it through again. We'll see if the AI suddenly goes back to skulking in a corner if it isn't so far ahead ... if anyone wants my file to play with (on the understanding that it's not really playable right now unless you're a masochist!) just drop me an email at misotu@hotmail.com or post here.
[This message has been edited by Misotu (edited August 14, 2000).]
|
|
| |
|
|
|
Helium Pond
|
|
Los Anheles, California, Good Ole U S of A
Nov 1999
time: 05:13
|
|
Here's an idea, Misotu, to test out the effects of might. Edit the alpha.txt so that one of the user-defined techs has values of 100, 100, 100, 100, and make it unresearchable, but give it to all the AIs. I've tested this a little, and it makes their might graph go orbital. I'm not sure how you could implement this idea in a non-scenario context, but you could at least use it to test the AI behavior. I'm too busy to spend the needed time playtesting, but I thought I'd pass the concept on to you. I'm also posting about this in the AC-Creation forum, if you want to check that out.
[This message has been edited by Helium Pond (edited August 15, 2000).]
|
|
| |
|
|
|
Helium Pond
|
|
Los Anheles, California, Good Ole U S of A
Nov 1999
time: 05:13
|
|
Chowlett, that's true, and it's probably what anyone who wants to test this out should do, but it's not a great solution, just because it involves a lot of messing with faction files each time you want to play a different faction, and also renders SMAX's randomization feature un-usable. I'm trying to figure out a way around this, which I describe in greater detail in the AC-Creation forum. I want to avoid messing with faction files, and find something purely alpha.txt based. Had a couple interesting results, check it out.
|
|
| |
|
|
|
Black Sunrise
|
 |
Buffalo, New York, USA
Aug 2000
time: 05:13
|
|
I just had an opponent chop-and-drop really viciously on me.
Year is around 2300, mid-to-late game. Everyone has a continent, pretty much. Huge map. Shared tons of technology, so we're all fighting nasty wars with shard interceptors and tanks. Everyone was at war with Miriam but me (she used a couple planet busters on Marr) when she decided the time had come for a double cross. Renounced treaty, Vendetta'ed, brought in about 8 or 9 shard choppers, wiped out every unit in half my continent(!!!!), airdropped on about four cities. Her choppers each came in, attacked, attacked, attacked. At least 4-6 attacks per chopper. Thank god I had 5 good cities producing shard tanks and interceptors, as well as a nice mag-tupe network, she couldn't hold her gains.
Personally, I think the AI spends a LOT of time building up, more than a human, because if it attacked prematurely, you'd beat it and complain that it was too easy. So they only pull the wicked many-probe, many-chopper, mass invasions when they have a LOT of power. Let them build a good infrastructure, keep them up to par on tech, and they get good.
I think the reason the AI doesn't build tree farms is because it only builds them if there are trees. It doesn't plant trees because they aren't as good as farms/mines/solar collectors untill you have tree farms.
I've had some interesting luck with home home scenarios by planting a forest next to each AI players start location. By mid game, that forest has usually become quite huge, and the AI has build the roads/mines they need and is playing a sherwood forest strategy on me.
In short: AI behaves badly because a) is small, b) has poor infrastructure and c) has low tech.
Share tech (benefits you, they have more to steal), give them a small edge on infrastructure (like my forest/start idea (which could make a nice faction, start faction with 1 former unit...)), and either hope he expands or choose yang, sven, etc.
Then you get wicked wicked opponents, and it gets really hard to roll over them.
Black Sunrise
|
|
| |
|
|
|
DilithiumDad
|
|
This scenario sounds a lot like "AI's Revenge" which is popular on ACOL. I set up Zak, Yang and Miriam on the Ultimate Builder Map in unbreakable pacts with pre-forested baes with prebuilt facilities. They never made choppers or probe foils, but they did probe effectively and shared tech with 100% efficiency. The only loophole I left was they were still willing to sell tech --Miriam sold Air Power for 100 credits. Even with that loophole, the only way for the 4 humans to win was to combine forces and coordinate attacks. The AI factions were able to maintain a stalemate in 1-on-1. AI's Revenge 2, 3, and 4 are currently running on ACOL with the loophole eliminated --all AI factions are in vendetta with the humans. Some very interesting stragizing among the pitiful human factions.
I also created a single-player scenario based on the same principles and neither I nor Mongoose could beat.
------------------
Creator of the Ultimate Builder Map, based on the Huge Map of Planet
|
|
| |
|
|
|
Commy
|
 |
St. Louis, MO, USA
Mar 2004
time: 05:13
|
|
Things the AI does rarely, or never...
Build clean reactor units--although an obvious to humans, apparently not to the AI
Build rover formers--seen this one a couple of times, but for only one reason, to build a road to a pactmate, then the unit apparently dissapears, probably disbanded...
Build satellites--excluding OPDs, only seen this once, on transcend against Zakhorav, who had the entire map except my OCC city...
Terraform bunkers--Never seen the AI do this
Win an economic victory--I have even given away enough energy to the AI to do this, but they never do...
Use crawlers--Only in pictures on this site, but not in my experience...
If any of you have seen this weird stuff, I'd like to see it too...
|
|
| |
|
|
|
Commy
|
|
St. Louis, MO, USA
Mar 2004
time: 05:13
|
|
Note the key words "right circumstances". I really don't care about the AI doing it, I want to see the reason, the conditions, that have caused the AI to do it. This way I can learn more about how the AI works.
Oh, and I have seen Zeiters game already with the dominant Hive, but didn't realize that it was going for an economic victory...
As for the satellites, I did say I have seen them once...
Last edited by Commy on 28-06-2004 at 08:19
|
|
| |
|
|
|
Natalinasmpf
|
|
I'm not sure if its might. I was playing as Svensgaard, unsurpassed, but I think its when the AI feels secure to "invest" in new, radical, things. Ie. Domai built THIRTY FOUR sky farms in a game, but only when it had a treaty with me, and was considerably advanced, even though it was just "potent" (oh fine I conceded unsurpassed a few moments :P).
When I declared vendetta, he stopped building them. Funny thing, the AI started to overdo it too, because not even I had reached supertensile solids yet for habitation domes, (he had Ascetic Virtues, nevertheless) and it wasn't doing a lot of good for new satellites after a while. But he kept building them. Oh, yeah, we were jockeying peacefully for top position.
Only when Domai decided it was time for the double cross, did his downfall start, and stopped building pods. Then I came out with orbital defense pods (once I had taken his HQ - which he managed to evacuate (!!) , but his bloody Cloudbase Academy was preventing MY air drops) to finally kill his last dissenting bases who wouldn't die because of all the food they were releasing, but would take forever to fight the garrison..
Too much might perception seems to be bad in this case. Overestimation of himself and his arrogance of his position led to his downfall 
Same with Roze. She was in fact, in a "sufficient" position, but she had a treaty with five other people. She build her own countless orbital defense pods. She was also with a treaty with me, and I dominated the sea, perhaps it elevated her perception of security? Her last vendetta was against an eliminated faction.
The unbreakable pact has something to do with it. Perhaps might too...its a question of, does the AI feel secure enough to take a gamble?
Does intense rivalry have something to do with it? Perhaps it affects the AI's perception of what is a gamble, and what is not? After all, aggressive doesn't mean smart.
Last edited by Natalinasmpf on 13-07-2004 at 18:14
|
|
| |
|
| |
All times are GMT. The time now is 05:13.
Apolyton Time is 00:13. |
top of page
|
|
|
Forum Rules:
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
|
HTML code is ON
vB code is ON
Smilies are ON
[IMG] code is ON
|
|
|
|
|
|
10-08-2000 23:40
Attachment: 
). And their inherant base perimenter defense. It hasn't been easy. But I have been using infantry for that addition 25%. I've had to sacrifice some, but I'm getting the job done. I'm not sure if they will surrender. If not, I may just annhilate them. 
|