Discussion in 'Civ4 - General Discussions' started by ianfuture, Jul 23, 2009.
30 MG's and 30 infantry in a city can defeat a stack of 100 units too.
The 14th legion versus Boudica?
That's the key right there!
I didn't say it was impossible, I said it was hard. Big mixed armies don't need just a few capable commanders, there's a long line of command to get things done right. Messengers need to find their way and units need room to maneuver, too.
Sure there are historical examples of well-led big armies, but that doesn't mean there wasn't a challenge to overcome. No doubt it was harder that just stacking heaps of independent units like you do in Civ.
I have no problem with SODs in Civ 4.
It's definitely preferable to the one kill kills the stack method from civ 1 and 2. If you ever play turn based MP (as opposed to simultaneous turn MP) you'll find SODs are not really all that effective anyway, past a certain size.
There's always going to be the advantage of having a mix of units in a stack, but once it gets too big it's just too much at risk of taking collateral damage.
As for flanking attacks, yes they'll wipe out siege units if you're not careful. Part of to this is to use forts where siege units cannot be flanked.
The AI is never going to have a good tactical mastery compared to a human player, and the way combat works now is probably the least bad, if you know what I mean. Making it spread its units out more in a deeper tactical way is just going to make the game easier in SP.
I would argue that any AI has the potential to be stronger than a human tactically, but will pretty much always be strategically weaker. Allowing for deeper tactical play may be a way to make the AI more challenging without adding handicaps.
Although chess and Civ are not the same thing, I think some parallels can be made. A decent chess engine (even a quick one that computes moves instantly to human time perception on modern hardware) will rarely make tactical mistakes. To beat a computer at chess, you can't try to fool it into making short term mistakes, you have to come up with a long term strategy.
I would propose a combination of ranged bombard + a penalty for being flanked by enemy units.
Tactical situations in Civ4 are infinitely more complex than tactical situations in chess.
It seems to me that Chess is infinitely more complex than Civ4, at least in terms of the kind of calculations required and the number of moves to think ahead, although I am not a Civ4 expert by any means.
Remember also that expectations are different. You would not expect a Civ4 AI to never make a tactical mistake, in fact that probably wouldn't be much fun.
It might seem more complex because it is better suited for analysis. Especially the closed world assumption in chess allows sophisticated reasoning to determine positively optimal strategy. Chess might actually be solved in the near future. Like tic-tac-toe, there might well be a sequence of moves that makes white absolutely impossible to beat.
Anyway, this closed world assumption entails that there is a clearer definition and measure of skill and performance in chess, yes. But it is precisely because of the limited scope and certainty that this is possible and not because of "complexity".
In Civ, there are so many options and configurations, so many possible moves and above all so much uncertainty about the state of the world at any turn, that it is utterly and completely impossible to do an exhaustive search for optimal moves. It is so complex that any player, be it human or computer, will have to do with vague generalizations and guesses. Those might look simple, but the real game/problem state is not.
No. In chess you have a board with 64 squares and 32 pieces of 6 different kinds. In a given situation there is a very limited number of pieces which can make only a few different moves. Over the course of a chess game the combinations become so many that they are still impossible to calculate beforehand (which would "solve" chess because you could calculate the combination of moves which leads to unfailing victory), but for a tactical timeframe, say 5-8 turns, modern computers can calculate the optimal moves.
In Civ, there is a board with thousands of squares and hundreds of units of dozens of different kinds. There are also cities, improvements, rivers and the squares (tiles) are different. If a unit attacks another, there is chance involved. Different units interact in different ways (whereas in chess any unit always beats any other unit so long as it can do a legal move onto the other unit's square). The list goes on. Civ has thousands of variables where chess has only a few. Just look at the rules: You can tell someone every existing chess rule in less than 5 minutes. Try to tell someone every existing rule for everything there is in Civ... you'd need hours and hours.
Note that I am in no way implying that becoming a good Civ player is harder than becoming a good chess player. But chess is suited very much to computing because it is a simple game at its core. The complexity arises because there are so many possible combinations of simple things. Computers are very good at doing simple tasks extremely fast.
From a Civ perspective this is true (Ctrl/Alt-Click and go)
From a real life perspective this is simply not true.
Good points Junuxx, I was referring more to the limited-scale tactics like the OP, for example: the AI senses certain requirements (enemy stack of 4 or more units with 2 or 3 moves of attacking a friendly stack, or something like that) and then switches into a limited tactical style thinking that might compute 3 or so moves ahead.
As for which is more complex, I think that is debatable. You mentioned uncertainty about the world. Uncertainty can actually lead to simplicity because you ignore what you don't know when calculating tactics, or you assume a specific state unless you see otherwise. In chess, you always see everything, and that is in a way more complex.
Also, there are a lot more variables, factors in civ4, but the great majority of those are hardly relevant to limited scale tactics. The AI might be concerned about not getting flanked, trying to flank the enemy, maybe setting up appropriate ranged bombardment an appropriate distance away, things like that.
Hm yeah, knowing everything you could does make the search space larger. On the other hand, if you know very little about the world, you might have few options to choose from. But it will be harder to determine the value of each option. So I suppose it depends a bit on your concept of complexity.
If you just leave out everything you aren't sure about, the problem might become 'simple', but you're bound to make huge mistakes. Sending a single spearman up that hill to check whether the enemy has reinforcements coming up could have a big impact on subsequent decisions.
If, on the other hand, you want to make your model as complete as possible, considering the likelyhood of at least the most relevant unknowns/risks, the task becomes considerably more complex.
I'm an AI student myself and a recent school assignment had me build a poker program. Initially it was to see every card, i.e., all cards were face-up. A later demand was to make it able to deal with uncertainty as opponent and table cards where face-down. Trust me when I say it's a huge increase in (computational) difficulty.
Well terrain, promotions, XP, unit costs, fortification bonuses, the difference between attacking and defending, first strikes are among them.
You might be right about the limited-scale thing though... I think chess' limitation of one-move-per-turn vs. Civ's one-move-per-piece-per-turn has a huge impact here, possibly the decisive difference.
Has anyone considered a rule like in the defense mod. It has a ten percent decrease in strength per unit on same tile down to 10 percent.
I am not saying make it this severe, but it could work.
How's that not true? If your soldiers are spread out hundreds of miles, they are going to be harder control than if they were all in the same place...
Also, I'm no historian, but I don't think Boudica's army really tried to co-ordinate themselves. Also, they would have done even worse if they had fewer numbers... if they had split up into a bunch of different groups, and fed themselves piece meal to the romans, they wouldn't have suddenly become master tacticians.
But if you put too many people in one place they'll become more difficult to control than otherwise. I'm not talking about merging two 50 unit squads into a single 100 unit squad. I'm talking about merging fifteen 200,000 unit squads into a single 3,000,000 unit squad. It becomes impractical somewhere between 100 and 3,000,000.
It won't be a single 3,000,000 unit squad though. In CIV you are a king/president/emperor. It is fair to assume the units will have enough officers and NCOs to have a chain of command. Plus, it's not 3,000,000 units in my back yard. It's more like 3,000,000 units deployed in London and the surrounding areas, which wouldn't be that impractical.
Civ 4 is not a war simulator.
Hearts of Iron 2 is a war simulator. In that game big stacks of units suffer severe penalties, if not commanded properly. Example: a general can command 9 divisions without penalties, a field marshall 12. Command limit is doubled if a mobile HQ unit (commanded by a general or a FM) is present in one of the adjacent provinces.
But I don't think I'd like Civ 4 turned into a realistic war simulator. It's not the point of this game.
I always wondered why population count and the persons in the unit aren't unified. Or are we building robotic warriors in 4000bc?
I don't agree at all. Personally I'd prefer to not compare to chess because it's such a different game. For one, tactics in chess are easier for an AI because it can feasibly look ahead more than 1 move. I don't think anyone is anywhere near teaching the AI in civ brute force tactics that takes into account the situation 2 turns ahead - the number of combinations of possible "board" configurations is already mind boggling I bet.
The fact the AI can build a civilization at all almost suggests it has some strategic strength. But even if it didn't and building the civilization was pretty much by accident, it doesn't look like a failure from a human perspective, outside looking upon it. Tactics on the other hand are where people usually level their complaints against the AI.
I suppose tactical concerns are mainly warfare and things like assigning citizens in cities.
As for whether the AI could be stronger than a human tactically, that's very unlikely given the huge uncertainty involved (RNG) in many parts of the tactical game. Trying to teach the AI how many siege units it needs is really difficult but an experienced human player just has a feel for it, for example.
AIs have actually beaten many human players. I think we all at some point or other have lost or nearly lost to an AI culture victory. The strategy implemented there might not be all that complicated but at least it gets results. In Better AI, the conquering strategies of strong civs has been improved and it probably won't be long before the team can teach the AI to pull off space victories more effectively.
I guess I would argue the strategic considerations in civ are sometimes a little less complex than the tactical considerations. There are only a few ways to win and many basic principles like "having more cities = good" that need to be employed, whereas on the tactical level the composition of stacks and short term city goals is too varied IMO.
Separate names with a comma.