AI question

[sapi]

I'd have to disagree.

For a modder, that's a good solution; but a dev should focus on fixing the problem (the way the campaign AI spends money and its inability to save) rather than the symptoms (low level cities).

It's not a bad idea, but imo it's not the right one for the devs to focus on.

[Foz]

I'd have to disagree.

For a modder, that's a good solution; but a dev should focus on fixing the problem (the way the campaign AI spends money and its inability to save) rather than the symptoms (low level cities).

It's not a bad idea, but imo it's not the right one for the devs to focus on.

I agree. If it's possible for a human to play the game well, then it is largely possible to make AI play the game on the same level. After all your decisions are based on something, and in a game like this that something is largely tangible - some surefire method you follow, or some particular way you react to a given situation. These reactions and decisions determine what and how you build, recruit, and in general how you go about successfully playing the game. The idea, then, is not to help the AI by giving it freebies, but to actually make it play more like a human by determining how players make the decisions they do. In the end this all boils down to mathematics. Somehow, when I have to decide what to build, I arrive at a clear choice by weighing different factors and ultimately deciding the relative importance of each possible choice using the information I have. So in theory the AI can mimic this by applying a formula that takes factors, weights them, and uses them to determine an importance value for each option. The whole thing sounds easy I know... but it's really not. Human thought is REALLY complex stuff! In the end though, I'd much rather have developers spending time simulating player logic than assigning the AI arbitrary perks to help it along. Those perks are simple enough to add at any given point, and if it falls on the members of the modding community to do that, I'd still much rather that than having the community try to invent AI decision-making logic. It seems only natural for devs to work hardest on the things that the community will not be able to change at all, and the things they can do that the community largely cannot...

[Katana]

Somehow, when I have to decide what to build, I arrive at a clear choice by weighing different factors and ultimately deciding the relative importance of each possible choice using the information I have. So in theory the AI can mimic this by applying a formula that takes factors, weights them, and uses them to determine an importance value for each option.

indeed, assuming the values are weighted correctly, the ai would have perfect information and hence be able to make perfectly rational choices, and wax the average player.

[ieyeasu]

indeed, assuming the values are weighted correctly, the ai would have perfect information and hence be able to make perfectly rational choices, and wax the average player.

Depends on what's deemed "rational".

I think what you'd do is set certain goals for the computer, the way a player does. A player builds a building because "Hey, I need to plan for my order situation when this city gets big" or "Hey, I want to build Janissary Musketeers here". So you set up those goals dependent on a myriad of factors: income, military situation, who your generals are, what faction you are, and go from there...

Excuse me while I wax vaguely on into the night...

[sapi]

I've said this before, but I think the major problem with the AI is not a lack of planning but an inability to plan.

A human, while saving to create, say, janissary musketeers, would put off other construction and recruitment tasks in order to raise sufficient money.

The AI always seems to find something to spend its cash on, and thus only lives from turn to turn.

[Foz]

I've said this before, but I think the major problem with the AI is not a lack of planning but an inability to plan.

A human, while saving to create, say, janissary musketeers, would put off other construction and recruitment tasks in order to raise sufficient money.

The AI always seems to find something to spend its cash on, and thus only lives from turn to turn.

I don't remember where I saw it, but I think there's a slider in the game somewhere that controls the AI's disposition toward spending vs. saving. It defaults to 100% spend. I haven't really played with it much, but it's entirely possible that its default position is what prevents the AI from ever having the cash it would need to make really great buildings or plan for the future. In fact I don't think I've ever heard ANYONE talk about that particular slider... so it should probably be something we try to tinker with extensively ASAP, since it's quite likely in my estimation that it will have a very noticeable impact on AI behavior. Whether that translates into better development and planning, or will just make the AI sit on piles of cash, though, is hard to say without trying it.

[Gaius Terentius Varro]

That's because the slider only works if you allow the AI to manage YOUR cities