Major AI weaknesses (some correctable)

  Alpha Centauri Forums
  The Game
  Major AI weaknesses (some correctable)

profile | register | prefs | faq | search

Author Topic:   Major AI weaknesses (some correctable)

Piglet posted 02-24-99 04:19 AM ET
As I've now got hold of the full game and completed one, now on my second, I thought it would be worthwhile starting a thread on the AI's weaknesses. Some of these seem quite severe and make the game somewhat too easy (but I'm looking forward to MP!), and some appear just to be sillinesses that should be readily correctable.
Anyway, I played my first game as UOP on Transcend level all Planet parameters and rules standard. I got a reasonable sized (supported 8 or 9 bases eventually) island to myself (had to expell Yang at one point but it wasn't his home island). Essentially I found I was able to just major on technology and cruise through to a victory by transcendence. As this is the highest form of visctory at the highest game level and it was my first full game I didn't think this should have been so easy! Below is a rather lengthy diagnosis of why I belive this occured:
1) Once you have an island to yourself the AI isn't very good at amphibious invasions. This is a general point and it's not easy to address. In part it's down to massing troops. Compared to Civ II the AI is much improved in this area for land attacks (tends to attack in good sized stacks), but doesn't seem to have the same improvements for attacks across water (should amass several transports).
2) My strategy was reliant on a super-base in which I maximized energy production (lots of supply crawlers to remote boreholes etc) and built every lab (and to a lesser extent economy) multiplier I could (eventually including both Super Collider and Network Backbone). By the end of the game this base was producing over 800 labs per turn! I also built about 5 other wonders here (which was foolish of me in fact). The point is that this information SHOULD be available to the other factions (special projects report gives most of it away) and that serious damage to this base would have crippled me. Therefore this should have been a major strategic goal (and I'm sure a human player would have exploited it).
3) There was a lengthy period during which I lacked a navy (didn't get doctrine:flexibilty for a while) during which time my coasts received lots of visits but no attempted landings. This probably isn't indicative of a problem but may be.
4) There was another extended period during which Yang (who I was at war with) had air units and I did not (and therefore no way to attack his). Use of his air power to support an amphibious assult would have been hard to resist, but all his aircraft did was overfly me essentially (and attacks non-combat units which is sensible).
5) There was a VERY extended period in which several of my enemies had missiles and I had no missile defenses. During this period I had a couple of sporadic conventional missile strikes from Santiago but these were ineffectual because they were not concerted (so they just wound up trading (expensive) missile units for (cheap) defensive units without ever actually overcoming a defense). Again using missiles in a concerted fashion to support an amphibious result would have made life difficult. All this time I was scared witless of a planet-buster attack on my super-base. This was such a strategic imperative that I'm sure any human player would have done it. The two enemies with the ability (in range and at Vendetta with the appropriate techs) at the time were Lal and Santiago. I can (just about) buy Lal's reluctance on character grounds, but why didn't Santiago do this!!?
In my second game (same setup for level and Planet) I'm playing Morgan. This is proving harder, but I'm pretty sure I'm still going to win (currently around 2240 and I'm about joint top having wiped out Dierdre and taken complete control of her area which is the monsoon jungle). I am (and have been for a while and probably will be for a while) at war with Yang and his units have done some really stupid things which I'm sure could easily be corrected:
6) Use of artilary. The AI (sensibly) uses artilary extensively to soften up bases prior to ground attack. However it uses it ineffectively. Since artilary will only take ground units down to 50% damage there is no point in continued bombardment after that point (choose another target). I have had Yang pound away on an empty base (my units are deliberatly sitting just outside of it just away from artilary range) with twenty plus units each turn. This is pointless because the attacks achieve precisely nothing (he cannot reach the base with ground units in one turn anyway, so there would be limitted purpose even if I had units there). Worse still if I move one defender outside a base but near his artilary he preferentially chooses to target this - with all 20 bombardments!! (good way to save a base from bombardment).
7) Stacking - as I said Yang has been bringing massive quantities of artilary to bear. However, he tends to do this in a single massive stack. I then attack this with a few rovers and can generally destroy the entire stack (27 units in one case!) with 3 or 4 units attacking (first few attacks destroy a single unit and give colateral damage, then eventually they all fall). This is a MAJOR weakness. The AI should avoid over-stacking and split into a number of stacks on separate tiles. It should at least put an ECM unit in as defense (he does have the tech). In cases like the one Yang was involved in he could also have built a sensor array in his territory that would have extended far enough to protect these stacks.
8) The AI should note when its enemies possess good air defenses and stop using their air power in the defended area (except for consolidated attacks maybe and for reconisance). I have found that if I base a good interceptor unit in my bases Yang will still try to attack my formers and most of the time lose his air units when my interceptors scramble.
I'm sure I'll have more to say as I continue to play! Anyway - it's still a great game and I'm really looking forward to some multi-player sessions!
Piglet posted 02-24-99 04:30 AM ET
Oh yeh - there were a couple of other things from my UOP game that I forgot to mention:
1) Lal beat me to the Hunter Seeker Algorithm but I was never troubled by probes! I guess this is part of the being on an island thing. Anyway if the AI had built some probe foils (I've never seen it do this - does it?) it would have been effective against me (especially in the period when I didn't have much navy or air power).
2) I had two sea bases miles from my island (in the new sargaso) which Lal knew about and came by every now and again. There was no way I could possibly have held onto these if seriously attacked but they never were! Lal would come by with a few battleships and kill any units not in base (sea formers mostly) but never attacked the bases themselves!
3) If the island thing is a big problem why not get the AI to terraform a bridge? Yang could have easily achieved this in this game. In my other game I did this to connect my original Morganic island with my holdings in Dierdre's original area so I could support it effectively (mind you - I have the weather paradigm, but Yang had the techs anyway at the game stage that mattered).
Fenris posted 02-24-99 11:54 AM ET
I've seen the AI do some incredibly stupid things also, in the game I'm playing now, Miriam moves a stack of 10-15 needlejets around my bases. They never attack, just over fly. I've also seen the same stack move throughout her territory. They didn't even respond to my attacking their home base. On the otherhand I've been isolated on an island and felt relatively secure only to have land bridges built and massive invasions started by the AI. It can use some tweaking, but it is better than most I've seen.
will posted 02-24-99 01:12 PM ET
I agree that the AI occasionally does some dumb stuff. However, I've also seen it use some really clever strategies -- massing units for a big attack and then spreading the attacking units over several squares so it was difficult to knock them all out. As people have said, it's better than anything else I've seen. Frankly, I'm kind of glad to know I can still beat the computer when the odds are even.
micje posted 02-24-99 01:18 PM ET
I could have written the scenario you described, Piglet. Exactly the same thing happened. I was on an island, at war with nearly everyone. I had 7 cities, my capital was in the south and was the only one responsible for my 1 tech/2 turns. In the north my cities were empty. As soon as I built a unit, the Hive (with 40+ cities) destroyed that unit. But they never occupied it. One time over 15 high-tech needlejets landed in a sea base in the north. I destroyed all of them with 2 choppers. Stupid! Or the AI keeps bombing a former of mine for 20 turns in a row with a gatling cruiser, which I'm stealing energy behind his back with a probe cruiser.
To close this thread, click here (moderator or admin only).


Hop to:

Contact Us | Alpha Centauri Home

Powered by: Ultimate Bulletin Board, Version 5.18
© Madrona Park, Inc., 1998.

Author	Topic: Major AI weaknesses (some correctable)
Piglet	posted 02-24-99 04:19 AM ET As I've now got hold of the full game and completed one, now on my second, I thought it would be worthwhile starting a thread on the AI's weaknesses. Some of these seem quite severe and make the game somewhat too easy (but I'm looking forward to MP!), and some appear just to be sillinesses that should be readily correctable. Anyway, I played my first game as UOP on Transcend level all Planet parameters and rules standard. I got a reasonable sized (supported 8 or 9 bases eventually) island to myself (had to expell Yang at one point but it wasn't his home island). Essentially I found I was able to just major on technology and cruise through to a victory by transcendence. As this is the highest form of visctory at the highest game level and it was my first full game I didn't think this should have been so easy! Below is a rather lengthy diagnosis of why I belive this occured: 1) Once you have an island to yourself the AI isn't very good at amphibious invasions. This is a general point and it's not easy to address. In part it's down to massing troops. Compared to Civ II the AI is much improved in this area for land attacks (tends to attack in good sized stacks), but doesn't seem to have the same improvements for attacks across water (should amass several transports). 2) My strategy was reliant on a super-base in which I maximized energy production (lots of supply crawlers to remote boreholes etc) and built every lab (and to a lesser extent economy) multiplier I could (eventually including both Super Collider and Network Backbone). By the end of the game this base was producing over 800 labs per turn! I also built about 5 other wonders here (which was foolish of me in fact). The point is that this information SHOULD be available to the other factions (special projects report gives most of it away) and that serious damage to this base would have crippled me. Therefore this should have been a major strategic goal (and I'm sure a human player would have exploited it). 3) There was a lengthy period during which I lacked a navy (didn't get doctrine:flexibilty for a while) during which time my coasts received lots of visits but no attempted landings. This probably isn't indicative of a problem but may be. 4) There was another extended period during which Yang (who I was at war with) had air units and I did not (and therefore no way to attack his). Use of his air power to support an amphibious assult would have been hard to resist, but all his aircraft did was overfly me essentially (and attacks non-combat units which is sensible). 5) There was a VERY extended period in which several of my enemies had missiles and I had no missile defenses. During this period I had a couple of sporadic conventional missile strikes from Santiago but these were ineffectual because they were not concerted (so they just wound up trading (expensive) missile units for (cheap) defensive units without ever actually overcoming a defense). Again using missiles in a concerted fashion to support an amphibious result would have made life difficult. All this time I was scared witless of a planet-buster attack on my super-base. This was such a strategic imperative that I'm sure any human player would have done it. The two enemies with the ability (in range and at Vendetta with the appropriate techs) at the time were Lal and Santiago. I can (just about) buy Lal's reluctance on character grounds, but why didn't Santiago do this!!? In my second game (same setup for level and Planet) I'm playing Morgan. This is proving harder, but I'm pretty sure I'm still going to win (currently around 2240 and I'm about joint top having wiped out Dierdre and taken complete control of her area which is the monsoon jungle). I am (and have been for a while and probably will be for a while) at war with Yang and his units have done some really stupid things which I'm sure could easily be corrected: 6) Use of artilary. The AI (sensibly) uses artilary extensively to soften up bases prior to ground attack. However it uses it ineffectively. Since artilary will only take ground units down to 50% damage there is no point in continued bombardment after that point (choose another target). I have had Yang pound away on an empty base (my units are deliberatly sitting just outside of it just away from artilary range) with twenty plus units each turn. This is pointless because the attacks achieve precisely nothing (he cannot reach the base with ground units in one turn anyway, so there would be limitted purpose even if I had units there). Worse still if I move one defender outside a base but near his artilary he preferentially chooses to target this - with all 20 bombardments!! (good way to save a base from bombardment). 7) Stacking - as I said Yang has been bringing massive quantities of artilary to bear. However, he tends to do this in a single massive stack. I then attack this with a few rovers and can generally destroy the entire stack (27 units in one case!) with 3 or 4 units attacking (first few attacks destroy a single unit and give colateral damage, then eventually they all fall). This is a MAJOR weakness. The AI should avoid over-stacking and split into a number of stacks on separate tiles. It should at least put an ECM unit in as defense (he does have the tech). In cases like the one Yang was involved in he could also have built a sensor array in his territory that would have extended far enough to protect these stacks. 8) The AI should note when its enemies possess good air defenses and stop using their air power in the defended area (except for consolidated attacks maybe and for reconisance). I have found that if I base a good interceptor unit in my bases Yang will still try to attack my formers and most of the time lose his air units when my interceptors scramble. I'm sure I'll have more to say as I continue to play! Anyway - it's still a great game and I'm really looking forward to some multi-player sessions!
Piglet	posted 02-24-99 04:30 AM ET Oh yeh - there were a couple of other things from my UOP game that I forgot to mention: 1) Lal beat me to the Hunter Seeker Algorithm but I was never troubled by probes! I guess this is part of the being on an island thing. Anyway if the AI had built some probe foils (I've never seen it do this - does it?) it would have been effective against me (especially in the period when I didn't have much navy or air power). 2) I had two sea bases miles from my island (in the new sargaso) which Lal knew about and came by every now and again. There was no way I could possibly have held onto these if seriously attacked but they never were! Lal would come by with a few battleships and kill any units not in base (sea formers mostly) but never attacked the bases themselves! 3) If the island thing is a big problem why not get the AI to terraform a bridge? Yang could have easily achieved this in this game. In my other game I did this to connect my original Morganic island with my holdings in Dierdre's original area so I could support it effectively (mind you - I have the weather paradigm, but Yang had the techs anyway at the game stage that mattered).
Fenris	posted 02-24-99 11:54 AM ET I've seen the AI do some incredibly stupid things also, in the game I'm playing now, Miriam moves a stack of 10-15 needlejets around my bases. They never attack, just over fly. I've also seen the same stack move throughout her territory. They didn't even respond to my attacking their home base. On the otherhand I've been isolated on an island and felt relatively secure only to have land bridges built and massive invasions started by the AI. It can use some tweaking, but it is better than most I've seen.
will	posted 02-24-99 01:12 PM ET I agree that the AI occasionally does some dumb stuff. However, I've also seen it use some really clever strategies -- massing units for a big attack and then spreading the attacking units over several squares so it was difficult to knock them all out. As people have said, it's better than anything else I've seen. Frankly, I'm kind of glad to know I can still beat the computer when the odds are even.
micje	posted 02-24-99 01:18 PM ET I could have written the scenario you described, Piglet. Exactly the same thing happened. I was on an island, at war with nearly everyone. I had 7 cities, my capital was in the south and was the only one responsible for my 1 tech/2 turns. In the north my cities were empty. As soon as I built a unit, the Hive (with 40+ cities) destroyed that unit. But they never occupied it. One time over 15 high-tech needlejets landed in a sea base in the north. I destroyed all of them with 2 choppers. Stupid! Or the AI keeps bombing a former of mine for 20 turns in a row with a gatling cruiser, which I'm stealing energy behind his back with a probe cruiser.