does the AI become smarter?

[Rusty Edge]

Reading through the betterAI forum should represent just how difficult it is to create dynamic AIs...

I'll put it on my "to do list" , but my wife is home these days and I find that I have little control over the list now

I went overboard, but I guess my point is that difficult as they are, I think that the better A.I. efforts should focus on making the human re-act to the A.I., rather than training the A.I. to react to the human- that just opens the door for manipulation.

The A.I. amphib assaults come with escorts and more than one unit per ship now. The enemy spies do an effective job of destroying things, especially oil. The SODs arrive in mixed forces. With improved objectives and improved misdirection, they would be more challenging.

 

[BibiTref]

Does the AI play different depending on the difficult level, or they just gets bonuses.

The higher the difficulty, the more bonuses they get, but is the AI at noble as smart as the one at deity, just with less bonuses? Or do they actually make smarter decissions at higher difficulties?

Well if you highlight Monarch when you are starting a new game it states the AI is "...just plain smarter," so the answer to your question must be absolutely yes. I just got Game of the Year edition and it is inconceivable such a product could have confused, inconsistent, or incorrect documentation. Right?

 

[PieceOfMind]

I'll put it on my "to do list" , but my wife is home these days and I find that I have little control over the list now

I went overboard, but I guess my point is that difficult as they are, I think that the better A.I. efforts should focus on making the human re-act to the A.I., rather than training the A.I. to react to the human- that just opens the door for manipulation.

The A.I. amphib assaults come with escorts and more than one unit per ship now. The enemy spies do an effective job of destroying things, especially oil. The SODs arrive in mixed forces. With improved objectives and improved misdirection, they would be more challenging.

I don't know what you've been reading but Better AI is not being designed in the way you describe. In fact, for the most part the AI changes do not depend on the human player at all, but rather make it better against both human and other AI players. Most of the changes are logic-centered or "objectives" improvements, making it hard to argue the case it is any more able to be manipulated than before.

If you think the AI does an effective job of using spies now, wait til you try out the new espionage tactics by the AIs in the latest Better AI release v0.82. Or were you actually describing Better AI already, and I misunderstood?

 

[phungus420]

Reading through the betterAI forum should represent just how difficult it is to create dynamic AIs...

While true the general concept isn't that hard. It would be relatively easy to block certain high level strategies for lower difficulties, just a simple if check to see if the difficulty is at a certain threshold before running the dagger code for instance. Iv'e never understood why the devs didn't do this, or why jdog is against it for better AI. This isn't something that's difficult to code, it's a design decision.

 

[PieceOfMind]

You don't understand why the devs didn't do this? I think it's pretty good design to make the same AI accessible at all difficulties. For one, it would not be very smart to set the AI to be arbitrarily stupider than it could be for difficulties where many players (especially those who don't hang around forums) play the game for most of the time. Why put effort into AI at all if only 20%, if that, of players will get to use it? It's about as illogical as those who claim tribal villages should have been a feature that was disabled by default in the vanilla game.

Also, if AIs used different strategies based on difficulty, it would make any strategy discussions (e.g. on the forum) more difficult because everyone would always have to say exactly what level they were playing.

I think jdog made the right decision in this case. What do you think people would do if you gave them an optional setting called "Lower levels have less sophisticated AI". Would anyone actually play that? I know I wouldn't.

 

[TheMeInTeam]

While true the general concept isn't that hard. It would be relatively easy to block certain high level strategies for lower difficulties, just a simple if check to see if the difficulty is at a certain threshold before running the dagger code for instance. Iv'e never understood why the devs didn't do this, or why jdog is against it for better AI. This isn't something that's difficult to code, it's a design decision.

I think you misread or I mis-stated. What I meant by dynamic is AIs that actually adapt based on their situation in-game to successful execute some of those strategies (choking, wonder gambits, tech beelines, tile improvement choices based on civic plans). I was not implying that it would be hard to make it use these strategies or not if they already existed based on difficulty...only that it would be hard to create well-executed versions of these tactics at all. Just city specialization is on the to-do list still (though coming up), and betterAI has been around for years! They did get rid of the cottage--->workshop loop and such. Eventually Jdog and co will get the AI to be markedly better than it is now...but having people with both high-level playing ability and programming ability is rare, and then only some high-level tactics should apply (making every AI beeline aesthetics to trade would be stupid. Having it prioritize its best yield tiles and maximize it's whipping/city specialization would always help it).

Anyway I wouldn't mind seeing the AI not willing to execute some strategies below noble. However it should *not* only execute them on the highest difficulties...doing so runs the risk of a) many players never even seeing the AI at its potential and b) using strategies that exploit bonuses directly rather than improving how the AI plays (the incentive to do this would be enormous).

But pulling some tactics below noble doesn't go against the spirit of current civ. The AI is already less likely to dow below noble than noble+ for example.

ALTHOUGH:

Also, if AIs used different strategies based on difficulty, it would make any strategy discussions (e.g. on the forum) more difficult because everyone would always have to say exactly what level they were playing.

I don't think this argument holds any water. It's already a problem. You can't reliably warrior rush on high difficulties ---> it's extremely easy below monarch. On noble teching aesthetics as your first classical tech is a pretty poor play...but on deity it is frequently your best bet for staying in the tech race. On noble you can hit with catapults and swords against archers for 100+ turns. On IMM or deity you might get 0 turns against archers because they got longbows in the BCs.

That said, I agree with your point overall for the most part. I don't see a problem with some kid gloves on the lowest difficulties, graduating quickly to "AI can do everything".

One thing that is very hard is that if the AI is improved, its ridiculous bonuses on the highest levels are too much...Jdog is on the brink of creating an AI that is unplayable at deity and possibly even immortal if he gets city specialization and building choices going the way he wants.

 

[r_rolo1]

One thing that is very hard is that if the AI is improved, its ridiculous bonuses on the highest levels are too much...Jdog is on the brink of creating an AI that is unplayable at deity and possibly even immortal if he gets city specialization and building choices going the way he wants.

That is pretty much expected. It already happened with Blake's project in Warlords/Vanilla and the solution was obvious and simple: cut the handicaps down.

Well, it looks like we have here another round of the handicaps vs level scalable AI abilitites debate In the end , if the the two options are open ( and to say the truth, in Civ IV they aren't ... unless any of you has a AI that can beat Deity players with ease with Noble handicaps hidden under the bed ) I would probably prefer the scalabale version, in spite of not being allergic to the handicaps system ( it has the great advantage of being more easily moddable, that is given ; ) ). Either way, it should had been made clear that the game uses handicaps since day 0, a thing that was not done ( in fact, like a poster said above, the game suggests otherwise ... but the game also says that Ballista Elephants are powerful melee units, so why bother with this minor details ? ) , a thing that was probably not done for marketing reasons ( well, part of the gaming communities sees games with a highly nitroed AI like "playing chess against a 6 year old and replace all it's pawns by queens: there is no shame in losing against it and no glory in winning" ( this is actually a memory quote from a post about Civ III with some years on top of it that I wasn't able to find again ) )

 

[PieceOfMind]

I don't think this argument holds any water. It's already a problem. You can't reliably warrior rush on high difficulties ---> it's extremely easy below monarch. On noble teching aesthetics as your first classical tech is a pretty poor play...but on deity it is frequently your best bet for staying in the tech race. On noble you can hit with catapults and swords against archers for 100+ turns. On IMM or deity you might get 0 turns against archers because they got longbows in the BCs.

That doesn't contradict my argument at all. I said strategy discussions would be more difficult. Of course I agree there are strategies that depend highly on the difficulty level, but I'd say they tend to stick in your mind more because they are the ones you have to be more aware of. There are, of course, many strategies/tactics that work at all difficulties (e.g. using siege units properly, building enough workers and working improved tiles, using diplomacy to prevent war, using a bureacracy capital, abusing slavery, etc.)

If for example, the AI was taught to use a lot more mounted units and do its best to attempt to flank attack enemy catapults only at Immortal+, strategies for using siege units would deviate based on difficulty.

 

[TheLazyHase]

One thing that is very hard is that if the AI is improved, its ridiculous bonuses on the highest levels are too much...Jdog is on the brink of creating an AI that is unplayable at deity and possibly even immortal if he gets city specialization and building choices going the way he wants.

I agree, handicap may need to be changed if the AI improve past a certain point. The handicap are big because AI is dumb ; it would make no sense to keep the bonii if nobody can (reliably ?) beat the deity AI.

 

[Badtz Maru]

You don't understand why the devs didn't do this? I think it's pretty good design to make the same AI accessible at all difficulties. For one, it would not be very smart to set the AI to be arbitrarily stupider than it could be for difficulties where many players (especially those who don't hang around forums) play the game for most of the time. Why put effort into AI at all if only 20%, if that, of players will get to use it? It's about as illogical as those who claim tribal villages should have been a feature that was disabled by default in the vanilla game.

Galactic Civilizations 2 is pretty successful with a scaling AI. The AI doesn't use all the strategies it is programmed for until you are playing at a level that's one or two above the "default" difficulty that is the equivalent of Noble in Civ4. If I recall correctly, you can set the AI difficulty independent of the handicaps for or against the AI players as well, so you can play a dumb AI that has big production and research bonuses or a smart AI that has penalties.

 

[PieceOfMind]

Galactic Civilizations 2 is pretty successful with a scaling AI. The AI doesn't use all the strategies it is programmed for until you are playing at a level that's one or two above the "default" difficulty that is the equivalent of Noble in Civ4. If I recall correctly, you can set the AI difficulty independent of the handicaps for or against the AI players as well, so you can play a dumb AI that has big production and research bonuses or a smart AI that has penalties.

I haven't played galciv2. Can you say what these strategies are that turn off at lower difficulties?
Depending on the rules of the game etc., it's quite possible changing AI behaviour based on difficulty might be the better way to go. Can you tell the game is better because it has changing AI, or is the game's success attributed to other reasons? I don't think you can say that something that works for galciv2 will necessarily work for civ.

If I had a guess, I'd say the AI features that turn off at the lower difficulties are along the lines of "be nice to newbies". If those strategies that it would have employed are particularly ruthless, it's probably a good idea to disable them to avoid frustrating new players to the game.

 

[Badtz Maru]

I haven't played galciv2. Can you say what these strategies are that turn off at lower difficulties?
Depending on the rules of the game etc., it's quite possible changing AI behaviour based on difficulty might be the better way to go. Can you tell the game is better because it has changing AI, or is the game's success attributed to other reasons? I don't think you can say that something that works for galciv2 will necessarily work for civ.

If I had a guess, I'd say the AI features that turn off at the lower difficulties are along the lines of "be nice to newbies". If those strategies that it would have employed are particularly ruthless, it's probably a good idea to disable them to avoid frustrating new players to the game.

No, it's not just "be nice to newbies" type stuff. I just checked Wikipedia and found out that the AI is fully enabled at two difficulties above "Normal", but still does not receive bonuses at that level. At higher difficulties the AI is the same but it receives bonuses.

The AI will build units specifically to counter units that the player has at the "Intelligent" setting, and plays a more long-term strategic game with it's placements of bases and it's counters to the player's development. But even on Normal difficulty it was really challenging and surprising at how smart it could be at times.

Edited to add: Another thing the AI will do on Intelligent that it won't do on Normal is recognize signs of betrayal. On Normal, you can trade and be friendly with a neighboring AI while building up your military starbases and fleet of military ships, and they will continue to be friendly with you. On Intelligent, it will recognize that it's likely to be back-stabbed, so they will start building defenses as if you are a hostile power, and may even make pre-emptive attacks if they see the opportunity. Essentially, the equivalent of a Civ that never declares at Pleased will start acting as if you are hostile if you are building up a large offensive force and there is no other likely target other than themselves. It won't do this at lower difficulty levels that have the same handicaps.

 

[PieceOfMind]

At higher difficulty levels these ships are tailored to the opposition, so players sending fleets of laser-armed ships may find them repulsed by shielded defenders, while those preferring to employ mass drivers will find that they come up against heavily-armored opposition instead.

That sounds almost like a cheat that is disabled at the lower difficulties. Anyway, I still don't really see the point. Would someone actually feel better about themselves by playing a less sophisticated AI and giving the AI greater handicaps than just playing the more intelligent AI with fewer handicaps?

To me it looks like the "intelligence setting" (whatever it's called) is there pretty much just to be set to full as soon as someone has learnt how the game works because there are still many difficulty levels above the one where the AI starts playing at its most intelligent.

I suppose the idea is that if the AI is good enough (maybe the game rules in galciv are simpler than in civ4) then it might really need to have features that can be turned off. I don't think better AI for civ4 can hope to get as good at its game than galciv's AI can at its game, by the sound of things.

Anyway, it's all just mostly useless musing. Most of the people desiring BtS's Better AI mod are not playing at lower difficulties and so to disable features at lower difficulties would not only be a waste of time, but pointless as well.

 

[TheLazyHase]

Anyway, it's all just mostly useless musing. Most of the people desiring BtS's Better AI mod are not playing at lower difficulties and so to disable features at lower difficulties would not only be a waste of time, but pointless as well.

It's worse than that : they are people like me that con't enjoy the highest difficulty because of the handicap (it look like you must "cheese" your way to win in deity : see for example the aesthetics trade and others tricks that work only because you're against AI with massive bonii), and definitely want more challenge. So disabling feature would be the most pointless thing to do for better AI : it will completely defeat the purpose of better AI.

 

[Badtz Maru]

That sounds almost like a cheat that is disabled at the lower difficulties. Anyway, I still don't really see the point. Would someone actually feel better about themselves by playing a less sophisticated AI and giving the AI greater handicaps than just playing the more intelligent AI with fewer handicaps?

To me it looks like the "intelligence setting" (whatever it's called) is there pretty much just to be set to full as soon as someone has learnt how the game works because there are still many difficulty levels above the one where the AI starts playing at its most intelligent.

I suppose the idea is that if the AI is good enough (maybe the game rules in galciv are simpler than in civ4) then it might really need to have features that can be turned off. I don't think better AI for civ4 can hope to get as good at its game than galciv's AI can at its game, by the sound of things.

Anyway, it's all just mostly useless musing. Most of the people desiring BtS's Better AI mod are not playing at lower difficulties and so to disable features at lower difficulties would not only be a waste of time, but pointless as well.

It's not a cheat, as it's pretty easy to determine what kind of units your opponent is building in the game. In Civ terms, if the AI saw that you were building a stack that was mostly knights and figured they were your likely target, they would start building tons of pikemen instead of a more balanced defense force.

I personally did not choose to play a dumber AI with the AI getting bonuses. After a couple of learning-the-ropes type games on the equivalent of Chieftan, I went to the "Intelligent" setting and repeatedly was beaten. It was just way too hard for me at that point. It was virtually impossible to surprise the enemy, whenever I went on the offence it would be ready with a defensive fleet specifically designed to tear apart my designs with minimal losses. The AI played as smart as it could, and it had the advantage of being able to micromanage it's entire empire perfectly and exploit any minor mistake I made. I ended up bumping it down to the Normal, which was still pretty tough, but the opponents seemed more human in that they made mistakes sometimes. Intelligent was what I'd imagine playing against the best Deity-level players in multiplayer would be like, if they had unlimited time to do their turns. I eventually did do the full AI with a handicap to give myself production and research edge, I still rarely won, but it was more interesting playing as a more powerful and advanced civilization being brought down by super-intelligent and conniving AIs.

 

[Badtz Maru]

GalCiv2 is less complex than Civ4 in a few things that may explain why it's AI is so much smarter.

Terrain is a lot less complex. Essentially, you have space that is all the same movement cost, filled with star systems with planets that can be colonized. There's also wormhols that you can travel through as shortcuts and things like black holes in places that you have to go around, but it's not nearly as varied as terrain in Civ4. I don't think that has too much to do with it as it's not pathing issues that GalCiv2 is superior in, the AI in Civ4 doesn't seem to have much trouble navigating to where it needs to go, it's weaknesses are elsewhere.

Combat is simpler in GalCiv2. Units are custom designed, and there is no terrain bonus other than some bonuses for being near certain kinds of starbases. There are essentially three kinds of offense and three kinds of defense that work in a paper/rock/scissors type system, and advancing miltiary techs just increases the powers of these different offenses and shields. This makes countering the player easier, if they see you using lasers, they build shielded units, if they see you using missiles, they build ECM protected ships, if they see you using mass drivers, they build armored ships. You CAN mix defenses and offenses in your ships, but that is generally very inefficient and your ships will be weak at everything compared to what you could do if you specialized.

The real edge GalCiv2 AI has over Civ4 is in it's adapting to the player's strategies and setting goals. If it sees you going for one kind of victory, it will either try to beat you by going for one you aren't trying for that it knows it can get quicker, or it will directly do whatever it can to hinder you. Are you going on a spree of conquest, picking off the weaker civs first? Expect those weak civs to be getting lots of technology and ships gifted to them by your stronger opponents, even if they are not necessarily friendly. If an AI civ KNOWS it will lose, it will surrender to an ideologically similar strong civ, and it won't wait until it's totally wrecked - frequently I would declare war on another civ that was easy pickings, and it would immediately surrender to someone I had no chance of beating. I think in some cases they will even surrender preemptively, i.e. before you even declare if they think you are about to. It works both ways, it's not uncommon at all to have an AI surrender to you out of the blue because they think you are the one most likely to stand up to the aggressor they fear.

When they developed GalCiv2, they looked at where the AI in other strategy games was dumb or easy to exploit, and designed the game and AI around making it really try harder to win, or at least survive.

I still prefer Civ4 to GalCiv 2 for a number of reasons, but the AI in Civ4 seems ******** in comparison. If Civ4 had an AI like GalCiv2, I'd probably have to go down at least two difficulty levels to have a chance of winning.

 

[Rusty Edge]

I don't know what you've been reading but Better AI is not being designed in the way you describe. In fact, for the most part the AI changes do not depend on the human player at all, but rather make it better against both human and other AI players. Most of the changes are logic-centered or "objectives" improvements, making it hard to argue the case it is any more able to be manipulated than before.

If you think the AI does an effective job of using spies now, wait til you try out the new espionage tactics by the AIs in the latest Better AI release v0.82. Or were you actually describing Better AI already, and I misunderstood?

I appollogize for any confusion I may have caused. I try to avoid it by putting stuff in my sigline. Sometimes it's caused because I don't know what I'm talking about. Sometimes it's caused by my medical issues. I'm not offended if you ask what I'm talking about.

I was contemplating the changes in the A.I. from Civ IV vanilla onward to summer versions of Legends of Revolution and History in the Making( which has a "cunning " leader trait ) before I made that particular post. I used to read the Toward a Better A.I. thread daily in the Warlords days, but haven't since.

I was recalling the changes over time. There used to be a worker-stealing festival. Then it was patched so that the workers ran home and stayed there whenever a warrior came to the fat cross. It became a well publicized strategy/exploit to park a warrior or archer on a wooded hill in the neighbor's fat cross to stagnate the A.I. until you could repare your axmen . Then it was patched again.. This is my classic example of designers reacting to the humans manipulating the A.I, rather than forcing the human to react to the A.I. I had another example in mind, however, I forgot it. Something else that gets debated as an exploit.

I like to forget that kind of stuff. I watch movies and play Civ for the immersive qualities. For the enjoyment. I don't like to be thinking about special effects and underlying equations. I've been described as totally role-playing civ. I'm OK with that.
I don't read or change code unless I'm trying to fix airships to make them more historical so that they don't interfere with my sense of immersion- or something of that sort.

I still enjoy playing with improved combat odds, or hearing your opinions and insights, and those of TheMeInTeam.

 

[cardgame]

GalCiv2 is less complex than Civ4 in a few things that may explain why it's AI is so much smarter.

Terrain is a lot less complex. Essentially, you have space that is all the same movement cost, filled with star systems with planets that can be colonized. There's also wormhols that you can travel through as shortcuts and things like black holes in places that you have to go around, but it's not nearly as varied as terrain in Civ4. I don't think that has too much to do with it as it's not pathing issues that GalCiv2 is superior in, the AI in Civ4 doesn't seem to have much trouble navigating to where it needs to go, it's weaknesses are elsewhere.

Combat is simpler in GalCiv2. Units are custom designed, and there is no terrain bonus other than some bonuses for being near certain kinds of starbases. There are essentially three kinds of offense and three kinds of defense that work in a paper/rock/scissors type system, and advancing miltiary techs just increases the powers of these different offenses and shields. This makes countering the player easier, if they see you using lasers, they build shielded units, if they see you using missiles, they build ECM protected ships, if they see you using mass drivers, they build armored ships. You CAN mix defenses and offenses in your ships, but that is generally very inefficient and your ships will be weak at everything compared to what you could do if you specialized.

The real edge GalCiv2 AI has over Civ4 is in it's adapting to the player's strategies and setting goals. If it sees you going for one kind of victory, it will either try to beat you by going for one you aren't trying for that it knows it can get quicker, or it will directly do whatever it can to hinder you. Are you going on a spree of conquest, picking off the weaker civs first? Expect those weak civs to be getting lots of technology and ships gifted to them by your stronger opponents, even if they are not necessarily friendly. If an AI civ KNOWS it will lose, it will surrender to an ideologically similar strong civ, and it won't wait until it's totally wrecked - frequently I would declare war on another civ that was easy pickings, and it would immediately surrender to someone I had no chance of beating. I think in some cases they will even surrender preemptively, i.e. before you even declare if they think you are about to. It works both ways, it's not uncommon at all to have an AI surrender to you out of the blue because they think you are the one most likely to stand up to the aggressor they fear.

When they developed GalCiv2, they looked at where the AI in other strategy games was dumb or easy to exploit, and designed the game and AI around making it really try harder to win, or at least survive.

I still prefer Civ4 to GalCiv 2 for a number of reasons, but the AI in Civ4 seems ******** in comparison. If Civ4 had an AI like GalCiv2, I'd probably have to go down at least two difficulty levels to have a chance of winning.

Haha, yes, you're right, the AI in GalCiv is pretty good compared to Civ's...

The second time I declared war on someone they folded to the already-strongest empire

 

[r_rolo1]

Haha, yes, you're right, the AI in GalCiv is pretty good compared to Civ's...

The second time I declared war on someone they folded to the already-strongest empire

Sorry to be a party breaker, but i find it hard to call that behaviour "inteligent". Call it "annoying to the human player", challeging to the human player" or something of that kind , but that is NOT inteligent behaviour by the part of that particular AI, especially in Civ IV, where you can win with one city and, barring PA, there is no way of making another player win my own ( as almost all of the GalCiv II wins are, atleast acording to the GalCiv II wiki ( since I never played the game )). So, there is no real incentive to a player to throw himself under the good graces of a protecting power just because it is being beaten to crap: being a vassal of someone who wins does not make you win in Civ IV That kind of behaviour can make the game more challenging , but that is in the exact oposite side of being inteligent, atleast seen in the prespective of the weaker civ in Civ IV that technically can win until the moment it loses the last city That said, in civ IV, where every civ can only win by themselfes barring PA, throwing themselfes into the mercy of a friendly civ or wait until vassaling to the agressor have the same value in terms of chances of winning... in fact vassaling to the agressor may even be a better strategy in a lot of cases This applies also to the argument that the AI should jump to burn UN/AP , the capital or one of the legendary cities to avoid a enemy win : ok, that might make the other guy not win the game, but it will make the attacker win it? Not necessarily ( and, given that those raids sometimes are pretty much kamikaze attacks, where you don't expect to see your soldiers back, in fact it can even make the player in question more prone to lose the game )

Other aspect to have in mind is that, unlike it seems to me from reading the victory conditions of GalCiv II, Civ IV victory conditions allow for very sudden jumps in the degree that a civ is close of winning via a certain VC and you can pass easily from having 50% of the win conditions fulfilled to 100% in one turn or even in the same turn for all the victory conditions. As it is easy to understand, gradual wins are far more easy to detect and counter by a AI than things that can sudently jump a lot. if you say to a AI " start preparing the ships to attack a civ when it reaches the second legendary city " ( for the sake of example .. it could be the first engine of the SS, or 40% of the expected votes in UN , or 30% of the land mass .... chose one, the logic behind is the same ), what the AI will do if I get the second legendary city at the same time as the third? How will the AI react to a player ( human or not .... ) that builds all the space parts in the same turn and relocates the capital to a spot inacessible in the turns that are needed to the ship to get to AC ? Or worse, what happens if the AI gets to be ready in time to burn the capital of a civ that built a slow SS just to be surprised by the launch of another SS in the other side of the world?

Bottom line, I strongly believe that the behaviour that people describe as being of GalCiv II AI is not adequate to Civ IV AI due to the diference between the victory rules. And , in another line of thought ( that I already used elsewhere regarding the same issue ), I have my serious doubts that making the life hard to a player that is in the way of winning can be described as inteligent behaviour in a enviroment that, besides one exception ( the already quoted PA ), doesn't allow comunal wins.

 

[Rusty Edge]

That is pretty much expected. It already happened with Blake's project in Warlords/Vanilla and the solution was obvious and simple: cut the handicaps down.

Well, it looks like we have here another round of the handicaps vs level scalable AI abilitites debate In the end , if the the two options are open ( and to say the truth, in Civ IV they aren't ... unless any of you has a AI that can beat Deity players with ease with Noble handicaps hidden under the bed ) I would probably prefer the scalabale version, in spite of not being allergic to the handicaps system ( it has the great advantage of being more easily moddable, that is given ; ) ). Either way, it should had been made clear that the game uses handicaps since day 0, a thing that was not done ( in fact, like a poster said above, the game suggests otherwise ... but the game also says that Ballista Elephants are powerful melee units, so why bother with this minor details ? ) , a thing that was probably not done for marketing reasons ( well, part of the gaming communities sees games with a highly nitroed AI like "playing chess against a 6 year old and replace all it's pawns by queens: there is no shame in losing against it and no glory in winning" ( this is actually a memory quote from a post about Civ III with some years on top of it that I wasn't able to find again ) )

I neglected to mention that I also enjoy yor input and European history lessons