when the ai wont accept 7 techs for 1 gpt

[TheNiceOne]

Originally posted by Pembroke
Oh, it's logical (as in "predictable") when you know the rules. It's still dumb, though.

It's dumb of course, but I have absolutely no problem with the fact that it won't accept the deal you give in your example. If you want to give the AI 1gpt for free in addition to what it accepts, then do it independant of any other trade.

But I agree with your idea, it would be better if the AI calculated you per turn offers by multyplying it with a percent based on how much it trusted you, ranging from 0 - 100%.

 

[Quandary]

TheNiceOne said:

quote:Originally posted by Pembroke
Oh, it's logical (as in "predictable") when you know the rules. It's still dumb, though.

It's dumb of course, but I have absolutely no problem with the fact that it won't accept the deal you give in your example.

I, on the other hand, do have a "problem" with it, although there isn't a blessed thing I can do about it other than stop playing, which I won't yet.

Pembroke, as I read him, was certainly not talking about giving the AI a gift. His example was the furthest logical extension of the AI's illogic: That you can give it enough to make the deal with you, but reputation or whatever other factors come into play can actually go so far as to destroy a deal when GPT is offered. Or sought, for that matter.

Sorry, everything you want in advance plus something else you think I might not be able to provide in the future, is still everything you want in advance. A 100% no lose situation.

Back to the original topic, I've gotten used to some Civs never, ever agreeing to pay GPT. I don't think I've ever gotten Rome or the Zulu to pay a single gp per turn even if I was offering them 30 cities in return.

Sorry, this is not logic in the traditional sense. This is hard coded idiocy where the limitations of the program can't conceive of selling one 20 GP improvement or disbanding one single over the limit mobile unit, to pay 0.0001% the value of what it's being offered.

 

[wilbill]

Originally posted by Quandary
Back to the original topic, I've gotten used to some Civs never, ever agreeing to pay GPT. I don't think I've ever gotten Rome or the Zulu to pay a single gp per turn even if I was offering them 30 cities in return.
Sorry, this is not logic in the traditional sense. This is hard coded idiocy where the limitations of the program can't conceive of selling one 20 GP improvement or disbanding one single over the limit mobile unit, to pay 0.0001% the value of what it's being offered.

I've certainly never noticed any specific AI's who never engage in gpt deals. It may be that some are more likely than others, but I've never noticed it.
What I do see in my games, though, is that the AI's seem to go through large stretches of the game without a large net gpt income. In the mid and late Ancient era, the wealthiest AI's may have a little disposable income to put into gpt deals. But generally it's mid or late Industrial before I see them willing/able to spend gpt.
I'm pretty sure that the "hard coded idiocy" is written into the AI to avoid setting up exploitable behavior. Humans can balance current and future income, present and future value of a resource or a tech, all the difficult to quantify plusses and minuses of a trade. The AI needs rules and fixed logic paths to make decisions. Yeah, it's limited, but it's what we've got. It is possible for the human player to learn how the AI thinks and adjust strategies and tactics accordingly

 

[Quandary]

Wilbill said

I've certainly never noticed any specific AI's who never engage in gpt deals.

It could very well be just coincidence that the Romans and Zulu seem to never agree to any GPT payments in my games.

 

[Speaker]

I've definitely gotten gpt out of Rome and Zulu, but like Willbill said, most of the AI civs rarely have gpt to spare until the industrial age. They usually run science at a very high rate, which makes little sense anyways since they all trade so much at such a discount. Why not keep some cash around? I found it that much weirder when in my last solo game, a Deity OCC, Greece had over 10,000 gold in their treasury from the middle of the Middle Age onward. Unfortunately, I had nothing to trade them.

 

[wilbill]

Originally posted by Quandary
It could very well be just coincidence that the Romans and Zulu seem to never agree to any GPT payments in my games.

I imagine that's it. Of course, Romans and Zulu rarely make it to the Industrial Age with a decent, productive civ in my games - it's happened, but rarely.

 

[Hurricane]

Originally posted by Quandary
Back to the original topic, I've gotten used to some Civs never, ever agreeing to pay GPT. I don't think I've ever gotten Rome or the Zulu to pay a single gp per turn even if I was offering them 30 cities in return.

1. You can't sell or buy cities.

Sorry, this is not logic in the traditional sense. This is hard coded idiocy where the limitations of the program can't conceive of selling one 20 GP improvement or disbanding one single over the limit mobile unit, to pay 0.0001% the value of what it's being offered.

2. The AI follows exactly the same rules as you in this case. They can not change their tax allocation, sell buildings or disband units DURING YOUR TURN. Is that so hard to understand?

 

[Pembroke]

Originally posted by Hurricane

1. You can't sell or buy cities.

Actually you can. Two humans can agree to trade cities in a multiplayer game. The game allows it. However, to close an exploitative culture flip loophole the AI is programmed to outright refuse all trades where a city changes owner except as a gift/tribute.

Some call this a solution. I call it lousy programming.

You should only disallow an otherwise reasonable something in a game if it is absolutely necessary. A far better solution is to make it harder or more expensive turning the exploit into a viable strategy.

A good game AI is one that emulates human behaviour as well as it can. Obviously presently no AI is so good it can pass as a human player in civ, and I don't expect that either, but every time the AI behaves in a way that rubs your nose in its AIness causes a "suspend belief" problem: the game experience gets interrupted.

If a better AI simulation in some part of the game is hard to implement then that's understandable and I can live with it. However, if a small tweak or a minor fix would change the AI behaviour into something that at least looks sensible then not do that is unexcusable. Ok, if it costs money and Firaxis is a business then release the improvements as an expansion pack and charge us. In fact here's a business idea for you: continue to improve, and I mean really improve the game and then charge us $1 via paypal for the downloads.

Heck, if Firaxis would do a major revamp of the AI concentrating to make it a really good opponent similar to what chess engine programmers achieve to do and released a true Deity AI Expansion Pack, and it would really be that good I would gladly pay another $30 for it. If it included an open interface AI double that. Probably wouldn't be the only one, either...

2. The AI follows exactly the same rules as you in this case. They can not change their tax allocation, sell buildings or disband units DURING YOUR TURN. Is that so hard to understand?

True, true. However what I *could* do is that if the AI contacts me during his turn and I had a sizeable treasury but with a zero or negative turn income, I could still pay the first turn gpt amount from the treasury running a deficit and then on my turn adjust the sliders.

Except the AI refuses, because of the hard-set rule that gpt offers more than per turn income are not to be allowed. Even if you had a treasury chest that _alone_ would pay for the entire 20 turn deal.

This sounds like another simple thing to fix to me...

 

[Hurricane]

Originally posted by Pembroke
True, true. However what I *could* do is that if the AI contacts me during his turn and I had a sizeable treasury but with a zero or negative turn income, I could still pay the first turn gpt amount from the treasury running a deficit and then on my turn adjust the sliders....

The point is that you can't do that in the game. The game won't let you offer more gpt than you earn at that particular moment.

 

[BiatchGuy]

what's even worse than that, is when the AI refuses to give up one tech in exchange for three or four..

 

[TheNiceOne]

Originally posted by BiatchGuy
what's even worse than that, is when the AI refuses to give up one tech in exchange for three or four..

Worse in what respect? If something like this happens, this is simply because the one tech the AI holds is worth more (for the AI) than the four you hold. Simple as that. This may be because the 4 you hold is also know by other civs, but makes perfect sense.

 

[TheNiceOne]

Originally posted by Quandary
Pembroke, as I read him, was certainly not talking about giving the AI a gift. His example was the furthest logical extension of the AI's illogic: That you can give it enough to make the deal with you, but reputation or whatever other factors come into play can actually go so far as to destroy a deal when GPT is offered. Or sought, for that matter.

Sorry, everything you want in advance plus something else you think I might not be able to provide in the future, is still everything you want in advance. A 100% no lose situation.

I agree that it would be much better if the AI accepted such a deal, by multyplying the per turn part by a number between 0 (if its trust in you is 0%) to 1 (if its trust in you is 100%), but if you want to give the AI something in addition to what it already accepts, you are free to do it now, you just have to give the gift sepearately. I fail to see that the game becomes any better by allowing me to give that gift as part of a trade as long as the AI treats the per turn part as worthless.

Back to the original topic, I've gotten used to some Civs never, ever agreeing to pay GPT. I don't think I've ever gotten Rome or the Zulu to pay a single gp per turn even if I was offering them 30 cities in return.

I think you actually have seen something that is part of a pattern. In the editor, different civs have different settings for what it builds often or never builds. If a civ is set not to prioritize wealth, then it will generally have less income than a civ that is set to build wealth often.

 

[Quandary]

The part of this that you might be overlooking, TheNiceOne, is that people don't necessarily know it's the GPT that's keeping a deal down. I'm sure that many have never noticed or realized that they could have a viable deal if they removed (the perceived value of) the gpt.

If someone doesn't know about the idiosyncratic approach the AI takes toward gpt, they could start a deal with a gpt value on their side and keep adding other values well beyond what the AI needs to agree, but still never get what they want.

I saw the gpt trap because I added that value after I'd gotten close to a deal. If I'd put it up first I might never have known that I could get the AI to agree on a trade by removing the gpt from the table.

By adding one aspect of quasi realism to trading, without having the advisor specifically telling the player why "they would never accept this (gpt) deal", they made trading more clumsy all the way around.

Your advisor is a semi omniscient being who already knows things he really shouldn't know. This makes life easier than having the hit "will you accept" with every change we make.

Adding an ability to have him say "Boss, the Atlantians won't believe our promise of gpt because they know we can't afford it (or they don't trust us, if that's the reason)" would maintain the borderline realism while telling the player exactly what is that's keeping the trade from getting done.

 

[Iulianus]

Refusing to buy tech for 1 gold may be "reasonable" but can anybody tell me why the AI never trades cities? I tried to sell my second to capital city for ONE gold several times even when it was gracious to me and it never accepted the deal (it was an insult many times)

 

[etj4Eagle]

Originally posted by TheNiceOne
I agree that it would be much better if the AI accepted such a deal, by multyplying the per turn part by a number between 0 (if its trust in you is 0%) to 1 (if its trust in you is 100%), but if you want to give the AI something in addition to what it already accepts, you are free to do it now, you just have to give the gift sepearately. I fail to see that the game becomes any better by allowing me to give that gift as part of a trade as long as the AI treats the per turn part as worthless.

I had thought about that solution the last time this topic came up a few months ago, but I think it was Killer who pointed out a substantial flaw with that plan. That factor for GPT would need to be added not just to the deal stage but retained for as long as the deal remains. By accepting GPT deals that have a good chance of being broken, the AI can no longer take its income rate at face value. It now needs to do more complicated computations in deciding how to manage its income and what kind of GPT deals that it can offer.

 

[etj4Eagle]

Originally posted by Iulianus
Refusing to buy tech for 1 gold may be "reasonable" but can anybody tell me why the AI never trades cities? I tried to sell my second to capital city for ONE gold several times even when it was gracious to me and it never accepted the deal (it was an insult many times)

In the original version it did, but this proved to be a major exploit. For one the AI put a very high value on cities and two it couldn't judge the risk of loosing a city well. A popular thing to do was to found a size one city right next to your capital. Then sell it to one of the AI and a couple of turns later it would flip (due to proximity to your capital, and that back then there was no way to completly eliminate a flip with a garrison). Then go and sell that city to another AI. As you can see this was an easy way to rack up tech, gold, and resources.

Even if some how that could be fixed, I think programing reasonable city trades would still be very hard. Much of the value of a city is quite subjective. I now I would find it hard trying to place a value on a city and don't know how one could write a deterministic set of rules for city trading that are reasonable under all circumstances.

 

[Pembroke]

Originally posted by etj4Eagle

That factor for GPT would need to be added not just to the deal stage but retained for as long as the deal remains. By accepting GPT deals that have a good chance of being broken, the AI can no longer take its income rate at face value. It now needs to do more complicated computations in deciding how to manage its income and what kind of GPT deals that it can offer.

But why would this be a problem? I mean, if a human player first got, say 100gpt from one civ bumping his total net income to 150gpt and then accepted another deal where the human player gives 150 gpt to someone else then I would say the human is definitely taking a high risk if not actually being dumb. You *should* anyway discount your per turn income with the risks whether AI or human. You should do so even with the current gpt implementation, because not having been betrayed or hit by bad luck in the past doesn't mean you won't be in the future. In fact one of the strategies you can use against the AI is to make him broke with forced deal breaks...

One of the current problems in the AI strategy is that it doesn't build gold reserves whereas any reasonable human player will do so. For example, in the Middle Ages when my core empire is about "ready" I habitually keep my treasury in the thousands and growing towards the end game. It can be a definite advantage if you can do a quick mass upgrade of units, rush build lots of improvements in a very short time, or have a research rush to something important with science at 100%.

Not keeping substantial reserves is an AI flaw whether the risk-based gpt deals are there or not, but if the AI gold reserve strategy was improved then the gpt discounting would be solved too as a by-product.

 

[Park Ranger]

Originally posted by etj4Eagle
Even if some how that could be fixed, I think programing reasonable city trades would still be very hard. Much of the value of a city is quite subjective. I now I would find it hard trying to place a value on a city and don't know how one could write a deterministic set of rules for city trading that are reasonable under all circumstances.

One time I would like the game to allow trading cities is when you and your ally are both conquering cities and want to reorganize the "spoils of war" to keep the conquests as close to capitals as possible. So I think if the AI could recognize, say, cities of all foreign nationals as tradable then you might be able to make something work.

 

[Sgt.Hellfish]

I once got over 100gpt from the romans for a MPP (odd huh )

 

[Skyer]

What the hell are you talking about? Trading with AI is so easy! You have a technology comp doesn't have. You negotiate with him and place the technology in column "We offer". And ask AI: "What can you offer me in exchange for this technology?". And AI may give you a lot of stuff: 30-100 gold per turn, hundreds of gold lump sum, world map, luxuries, contacts - and this all for one technology!

My record was to make +431 with 100% taxes on science, and all the money was only from AI for my technologies!