Comparison to a pretty similar game that came out just recently - Warlock - shows neatly the flaws and limitations of the Civ5 system.
And, to be fair, only parts of it can be blamed on the AI as such. A big part of the problem is caused by the 1UPT approach itself - as sound as the approach is by itself, it's just the wrong pick for a global scale civ game (at least in my opinion, many here disagree).
Most units, at least until fairly late in the game, have pretty low mobility (in Warlock, units can move much farther, and this alone makes a huge difference), yet - especially with the recent changes - you must bring a fairly large number of units to a war if you want to have any hope of winning (and capturing cities along the way).
This means, rather than Civ4's supposed (but in reality pretty rare) stack of doom, you now need a carpet of doom, with every unit taking part always standing in the way of other units, especially when there's need to rearrange due to enemy action or navigational hazards. Even a human player will constantly struggle with this, at least to a degree. Thus it's of course a big challenge for th AI as well.
Now, I have to admit, since the early versions, the way the AI handles this has massively improved. As PhilBowles explained, it's now capable of assembling it's carpet of doom and often manages to stage actual, fairly well planned assaults with it. It can move the units to a destination in tickles, amass them, and then launch a fairly coordinated strike.
From there, it often falls apart a bit, it often picks not the best priorities, concentrates on stuff it should ignore and vice versa, doesn't retreat units in time, positions units in a way that prevents other, better suited units from attacking or that puts/leaves vulnerable units in harms way.
Like I wrote above, the low mobility of the individual units probably plays a big role here, though the AI isn't really a master of cavalry warfare either.
To sum things up, the AI is much better than in vanilla/early versions, but any halfway seasoned player will have no trouble beating it on even remotely even grounds. The game relies on HEAVY cheating to make things challenging for the player, and this does work, at least partially. If you have no issue of playing against enemies that are playing the same game by a different set of rules, things are at least okayish now. Not splendid, but okay.
The "strategic" warfare AI is still pretty horrible, or so i found. You can still be offered massive compensation for a peace treaty in a war the AI started and in which you never actually met let alone fought the offending force, and at other times the AI will still stubbornly refuse unfavorable terms even if you just wiped out their army and captured every single city of theirs except one. It's just braindead.
The good thing is that Civ5 can be had for pretty cheap these days. Pick up a copy of the vanilla game without any DLC or the expansion during a sale and try it out running the newest version. You'll get a pretty good idea about how the game plays.
If you like what you see, you can consider the addOn, which features some mostly non-essential improvements and is generally MOTS, and if you're really fond, you can get the DLCs as well.
This has always tended to be the case in Civ games. And many others - Medieval II Total War seems to use basically the same calculation (military strength alone, regardless of contact with or losses to the enemy force, and counting allies' strength towards its own) to determine whether to accept peace, which routinely results in refusals to capitulate by powers that have lost entire armies en masse and cities to your forces.
This may not be wholly reflective of the current AI, since as I noted I believe some of these AI changes are in coding only available through the expansion, not through patches (I've also had a sneaking suspicion that 'stealth' AI patches may have been introduced in DLC, since I experienced drastic improvements in pre-G&K AI over time without any official patches announced).
In earlier Civ games, completely irrational peace offers were ... at least FAR less common than in Civ5.
Still doesn't explain away why an AI having lost it's entire empire, complete with armies and cities, still somehow values itself in a position to make or even resist demands.
_____
rezaf
I don't really now how to respond Phil, I'm not sure why you even decided to pick apart my post in the first place.
I was trying to be as fair as possible towards Civ5 - of course an AI will always have problems, a simple fact is that Civ5's AI was VERY bad early on, but has massively improved since. It's still not really good - even for an AI - but it gets the job done now, which is sufficient for the most part.
Comparisons to weaknesses of other (non civ) games AI's are pointless in my opinion, because we're talking Civ here.
About Civ4, of course it's AI wasn't remotely close to perfect, either. It played the game better on release, though, and was also improved a lot on the way (like Civ5's now), and many improvements came from work players started with the DLL release, a venue which is still closed for Civ5. Also, I feel the Civ5 AI is cheating much more blatantly and obviously, essentially - like I wrote - playing by a totally different set of rules. That's disappointing to me.
Regarding your Warlock question, first playing it my initial reaction was the notion that Warlock's AI is LEAGUES ahead of Civ5's. Later on, I realized it hat its weaknesses also, and a lot of the advantages to Civ5s AI stem from the fact that all units are much more mobile in that game - as it should be in a 1UPT game.
Compared with the most recent Civ5 AI, Warlock doesn't play much better I guess, but it hides its weaknesses better, until it falls completely apart when faced with certain situations. But I guess the Warlock devs are also working on fixing the issues it still has - haven't played with the latest patch yet.
In earlier Civ games, completely irrational peace offers were ... at least FAR less common than in Civ5.
Still doesn't explain away why an AI having lost it's entire empire, complete with armies and cities, still somehow values itself in a position to make or even resist demands.
Well, but wouldn't you agree that it allows someone unfamiliar with Civ5 to get an idea what he'll be getting into? A few more improvements might be in the books when getting the expansion or the DLCs, but the base game impresson remains unchanged, no?
The last couple of Civ IV - unmodded - games I played I was reminded just how commonly the AI will randomly throw three-unit stacks of Jaguar Warriors at your larger army, run around the map with armies consisting entirely of siege units, fail to respond appropriately to the presence of particular unit types in your stack, or use cavalry-heavy stacks to defend cities - all using BtS AI. These aren't isolated examples; in those games I rarely went to war, but when I did I encountered this behaviour consistently.
And yes, Civ IV's most ardent defenders won't claim it had the world's best AI by any means, even post-BtS, but my point isn't that the AI isn't any better than Civ V's, or that it's worse - it's that it gets things wrong in exactly the same way, irrespective of the differences in combat system - poor unit selection, poor target selection (in the sense that in stack play you can choose the order of attacks based on the way the system selects defenders, to limit your units' exposure to their counters), weakness on defence etc. That suggests strongly it's not an issue attributable to the 1UPT system.
Civ V certainly did have 1UPT-specific problems in vanilla which have no equivalent in stack play (unit placement relative to one another, coordinating attacks from multiple units, moving armies as a unit)
That's what I mean by describing G&K AI as a bad player, and vanilla AI as "playing like an AI". An AI doesn't know the rules; a bad player knows the rules but not how to apply them to best effect. The AI now "gets" 1UPT in the same way the earlier AIs "got" stacks - but it's never going to produce the best armies or use them to best effect, any more than Civ IV will do with its stacks.
I'd agree, although I came onboard with Civ V after it's early "VERY bad" phase, after the first big patch while I'm told did a lot to help the AI. Though I'd estimate its AI as being better than you give it credit for in comparison with other games - Civ games and non-Civ games alike; it handles more complex decisions than these previous incarnations in issues like combat and diplomacy, and for the most part produces comparable results.
Didn't you start with a comparison with Warlock...? As above, the point is to identify common elements where AIs err regardless of mechanics. Arguing that Civ AI isn't comparable to Total War AI because the games are mechanically different is no different from claiming Civ V AI is immune to comparisons with Civ IV AI because the combat mechanics are different, and plainly you don't accept that premise any more than I do. And in cases like the diplomacy system, the mechanisms the AI uses to determine whether and on what terms to accept peace are essentially identical between older Total War games and Civ.
Describing an AI as "cheating" is nonsense. It has no motives, and didn't ask for any special favours. It simply gets bonuses to compensate for being a weaker player than its likely human opponent, and does so for the benefit of the human, not the benefit of the computer - if that's cheating, getting a handicap in golf is cheating. And whatever the "feel", Civ V's AI-bonus-by-difficulty setting isn't any different from the systems in the previous Civ games.
We can hope that this is also the case with Civ V. The game may yet have one more expansion in it, and almost certainly further patches are planned.
But I haven't noticed much difference in the situations where the AI would offer or accept a silly deal across Civ's incarnations.
I've seen them resisting demands in that situation, but even in Civ V vanilla I didn't often find them making demands if they were down to their capital or if their army was gone. The positive thing about a system based entirely on military strength assessment is that the AI will usually be willing to accept peace once it perceives its military is weaker than yours, and that's been my experience.
I'm not so sure - for a start the 100 HP system isn't in the patch (which I've played in multiplayer with G&K disabled before my group members all had a copy), it's in Gods & Kings, and I think that does make a substantial difference to the AI's combat performance. Also, while the AI plays diplomacy in much the same way in both versions, it feels very different in Civ V (EDIT: In G&K, I mean of course), not because of AI improvements, but because the types of game-long relationships, tripartite relations etc. that you could routinely form in vanilla with sufficient effort are much more accessible and easier to form for less experienced Civ V players following the expansion.
Thanks for the clarifications Phil.
Let me put it this way then, I felt many of Civ4's weaknesses only surfaced when you played on very high difficulties. The often lamented SoD was practically never to be seen unless you played at King or higher. And if the AI had smaller armies and stacks, the problems you describe were much rarer, thus the issues were better concealed.
I gotta say, though, that I felt a catapult in a stack attacking at the wrong time in Civ4 was much more bearable to me than the AI maneuvering the largely defenseless catapult right in front of my knight or something like that in early Civ5.
An AI always must know the rules of the game, if it does not, it's merely broken.
Early Civ5 AI was. By know, this has largely been addressed.
Combat is now more complicated, I'll give you that, but older Civ AIs had to deal with more complicated mechanics in other areas, imo.
I just think a StarCraft II AI is faced with mostly very different problems than the Civ5 AI. Also, it manages to solve it's problems in a much more timely manner - one of the biggest issues I still have with Civ5 are it's IMMENSE turn times in the later ages.
And giving it a few extra resources is one thing, but making cornerstones of the game design not apply to it is just over the top in my book.
In Civ5, the 1UPT meant there should be less units present on the map. A lot less, being ground combat units can no longer be stacked in any way.
So, to address this, and to prevent ICS, the global happiness was introduced.
Fair enough.
But as early as in Prince difficulty, you can notice in Civ5, that the same restrictions placed upon the player are OBVIOUSLY not present for the AI.
it can build and support more troops,
it can build more improvements,
Well, I have. Though one has to admit this was always a pretty weak area in Civ games, I feel it's worse in Civ5.
War resolutions aside, another example is how the AI often makes totally outrageous suggestions for research treaties from which it will benefit just as much as the player. What, you want half my per-turn gold, half my iron, a third of my horses and all my coal to sign up for this? Lemme show you to the door...
This kind of behavior can actually help many a game, though - when wars are going on in which the player participates is where Civ5 really shines. As an empire building game, it's pretty boring and dull, and a giant leap backwards from previous civ games.
So more battles (longer wars) usually mean, more fun to be had.
Only a over-hating AI or an over-loving AI. The rest are just Neutral or most of the time
I can't remember if I was playing on Prince or Noble when those came up, but they were certainly noticeable.
That depends whether you're defining "complicated" from a human or an AI perspective, I suspect. Civ IV's mechanics tended across the board to have more binary answers - "I'm specialising for production, do I want a windmill on that hill? No", "Does X share my religion? Yes", "Is my unhappiness nearly equal to my happiness? If yes, build appropriate building" - which suit an AI. In diplomacy, for instance, there was nothing resembling the denunciation mechanic that made the AI need to consider who it should denounce rather than just whether it should. Very often these are decisions that will be trivial to a human, and many people feel that this is more generally the case in Civ V than Civ IV, but they're harder for an AI to process.
I'm used to playing Total War games (with animations enabled), so even later-age Civ turns seem lightning-fast...
Which elements are these? I know the AI gets extra gold, happiness, units and an early tech or two at higher levels, but I'm not sure what other bonuses it receives.
I don't think this is a case of "rules don't apply", (...)
the AI would actually need to play as though those bonuses give it enough of an advantage to be challenging, which many players feel isn't the case.
And to be honest I haven't noticed the AI having large numbers of gargantuan cities - in my Emperor games the AI capital routinely level-pegs with mine in the early game in its population size, and G&K has either reduced happiness bonuses or curtailed the script that prompts the AI to found cities. Sure, in my last duel game Gustavus had far higher happiness than I did - but until 1810 AD Stockholm was his only city, and that was smaller than Gao (my capital), while I had 3-4 other moderate-size cities.
So, I've noticed the bonuses (how do you get 54 happiness with just one city and no CS allies, again?), but not at a level drastic enough in terms of the way the AI plays to make it feel it's "playing a different game".
Which it needs to compensate for losing more of them.
Although it rarely does...
In fairness, I've mostly noticed this when there's a tech imbalance in which the player will gain more from the RA than the AI because the player's in the lead technologically to begin with. The AI will usually offer a more favourable deal if it's the tech leader because it similarly considers it will benefit more than the player.
I don't think I'd agree. I don't think you can separate the war and empire-building elements - no, it's not Sim Empire, and it may be telling that a lot of Civ V's critics favourably compare Civ IV with games like Sim City which I found interminable and somewhat pointless - to me it feels more a game and less a management tool. War's part of that, diplomacy's part of that (and as weak as the AI may be at utilising it, the DOF/Denunciation system represents in my mind a great advance over previous games' almost wholly trade-based diplomacy system) and, of course, so is the actual building part.
I think it's a mistake to think of Civ V as a game that forces or encourages you to go to war rather than build peacefully throughout, in the same way that it's a mistake to think of chess as a game that forces you to move your pieces
Somebody rename this thread "Phil and rezaf talk Civ" already.
I guess it depends how you define SoD. Just a number of units stacked into an army? Yeah, happened all the time, even on Noble or lower.
A gargantuan stack that you basically know you have no chance against, no matter how many defenders you railroad in from other cities or how much you weaken it by artillery or whatnot? This was pretty rare in general, and I've only seen it on King and above. Except in mods, that is.
Well, earlier Civ5 showed just how much thought the AI puts into things like denouncations. I don't really see these evaluations as all that complicated.
It's still only a bunch of variables being computed.
How big is the other player? How powerful his army? Do we have a conflict of interest (close borders etc.)? Maybe stuff like "are we competing for the same City State"?
These are put into a calculation, and there you have it, denouncation.
I don't see this being very hard at all.
Heh, is that a serious remark? Just because someone else does an even worse job doesn't mean you should do a bad job, imo.
All you're saying there is that you don't mind the chess AI getting a full row of queens because it isn't using them to full effect. I say, I don't want the AI to have a row of queens, because I want to play chess, and I want the enemy I play against to ALSO play chess, by the same rules.
Suppose you've never had a game where a Civ stole EVERY wonder you attempted to construct AND built a larger army than you ever could hope to field?
Phil, I appreciate your opinions and the time you take to put them into writing, but I really feel we've taken this thread FAR too much off-topic, so I'll no longer respond to your posts here in-depth.
I'd love to discuss that stuff further, so if you want to do that, think of a good title and open up a new thread, and maybe a mod can then even move our posts from this thread to the new one, so this becomes less cluttered.
Anyway, I believe it boils down to that, in the end, we probably agree about many aspects in one way or the other, but ultimately possibly want the series steered in different directions.
I agree with many of your analyses, but I felt the solutions to the problems at hand Civ5 attempted were all wrong.
_____
rezaf
We'll have to agree to disagree on this point, I guess. Maybe cheating isn't the correct word to use (I believe it to be, but I'm not an english native speaker, as you probably have noticed), but I STRONGLY dislike it when an AI needs this kind of "support" to function in a challenging way.
And giving it a few extra resources is one thing, but making cornerstones of the game design not apply to it is just over the top in my book.
In Civ5, the 1UPT meant there should be less units present on the map. A lot less, being ground combat units can no longer be stacked in any way.
So, to address this, and to prevent ICS, the global happiness was introduced.
Fair enough.
But as early as in Prince difficulty, you can notice in Civ5, that the same restrictions placed upon the player are OBVIOUSLY not present for the AI. The computer can found tons of cities and grow them all to gargantuan size, and still end up WAY ahead of you in happiness polls.
In turn, this of course means the AI has much higher income (plus bonuses), it can build and support more troops, it can build more improvements, it generally plays a different game than you do.
To me, this is such an essential part of the game mechanics, it's as if you gave a Chess AI a full row of Queens instead of Peasants, to even the field a bit, because hey, it's just an AI and needs a hand.
_____
rezaf
Shogun 2 eh.... total different beasts compared to Civ and shogun 2 is really different. Its not really fair to compare the two.
I remember when I was playing as the pro imperials, my armies got decimated because my faction got singled out by four massive pro shogunate factions but I won because I pulled off heroic victories where it counted and prevented different factions from battling together in same battle against my armies by subterfuge.
There's agents that can delay your armies for like a turn, and that single turn can make a world of difference enabling you to fight like 3k vs 10k instead of fighting 3k vs 20k.
I could potentially throw two warriors into a horde of 10 barbarians and have them survive somehow. But terrain is a huge factor on that. Because of the crazy bonuses we get against barbarians.
And for rebels... rebels in total war series is actually rebels. Whereabout civ rebels is really just barbarians. =.= Bad firaxis, bad. That's just lazy.