The AI and the GCIII economy.

Something Paul said during the 9/4 stream has really stuck in my craw. I know Paul isn't an expert on AI, so there might be some misunderstanding, but Paul said that if they had the AI govern the planets at optimal efficiency using the planetary wheel, it would be so good at it that it would seem like it was cheating.

Isn't this exactly what players want when it comes to AI at higher difficulties?

I am no expert in game AI, but it seems to me if you have an AI that can play the game so well while following the same rules as a human player that it appears to be cheating, is't that exactly what players have been begging for since game AI existed? Isn't that like the holy grail of AI? If you could have made the AI that good at managing the economy and elected not to, I find that infuriating. What most players want is an AI that is impossible difficulty without any cheats, then is programmed to run at less than maximum efficiency to bring it down to normal difficulty. Paul seemed to imply that the AI was potentially so good at individual planet management was a reason to not have planetary management. Instead, it seems to me to be a massive untapped well of power. SD is getting rid of the planetary wheel, but if you can program the AI to use it properly, why not leave it in for the AI as a way to increase difficulty rather than giving the AI conventional bonuses and cheats?

33,259 views 19 replies
Reply #1 Top

Yeah, it did seem a bit bizarre to say 'oh, if the AI played properly everyone would think it was cheating' the day after releasing a patch which buffed AI handicap bonuses on high levels and while the AI is reliant on cheating (oh, sorry, 'handicaps') anyway.

 

Again, a far more satisfying method of AI difficulty would be letting it micro manage increasing numbers of planets depending on the setting - say, 5 for normal, 15 for gifted, 30 for genius and all planets for godlike. Then ditch the unnecessary handicap bonuses. This would mean you had an AI that got smarter as you increase it, rather than just getting huge bonuses.

+1 Loading…
Reply #2 Top

Had exactly the same thought about pauls comment.  Would like to hear Brad's thoughts

Reply #3 Top

I don't agree. 

In previous GalCiv's, a lot of the AI quality came from the AI's use of planetary optimization features.  

My problem with the per planet wheel is that it is too gamey.  I don't mean in terms of it being exploitive. It just totally rubs me the wrong way. 

If you've ready any of my AAR's you know how important it is to me when I play to think of my civilization as being full of real people. Not robots that I can command in lockstep.

Reply #4 Top

I under

Quoting Frogboy, reply 3

My problem with the per planet wheel is that it is too gamey.  I don't mean in terms of it being exploitive. It just totally rubs me the wrong way.

I understand, and though I've posted a lot in support of the planetary production wheel, that isn't what I'm talking about here. That just happens to be the context in which the comment was made. In the stream on Friday, Paul basically said that the AI could have governed planets more efficiently than a human could, but you didn't let it do that because it would feel like cheating to the players. If this is true, I was saying that if you can make the AI play the games so well that it feels like cheating, that seems like a better solution for higher difficulty levels than the current solution, which is that the AI is actually cheating.

Reply #5 Top

On the higher levels program the AI to be better at managing the planets than at lower levels.  

Reply #6 Top

Quoting peregrine23, reply 4

I under

Quoting Frogboy,

My problem with the per planet wheel is that it is too gamey.  I don't mean in terms of it being exploitive. It just totally rubs me the wrong way.



I understand, and though I've posted a lot in support of the planetary production wheel, that isn't what I'm talking about here. That just happens to be the context in which the comment was made. In the stream on Friday, Paul basically said that the AI could have governed planets more efficiently than a human could, but you didn't let it do that because it would feel like cheating to the players. If this is true, I was saying that if you can make the AI play the games so well that it feels like cheating, that seems like a better solution for higher difficulty levels than the current solution, which is that the AI is actually cheating.

The AI can do better with the production wheel than a human.

However, no AI, can build up a planet as well as a human.  It is literally the traveling salesmen challenge in a game.

We knew that when we decided to support adjacencies that are implemented in an ordered way but people like being able to cleverly specialize their worlds. My view is that once a player gets that good, you just give the AI bonuses to make up for it at that point since it's literally impossible to make the AI design up a planet as well as the best humans.

Reply #7 Top

The adjacency system needs to be nerfed.  Keep the planetary production wheel, but nerf adjacencies.  

The problem is that with your adjacency system, you've developed an economic model that grows EXPONENTIALLY.  

Giving the AI bonuses isn't going to work in an exponentially growing system.  Too much bonus and the human never catches up.  Too little and it stops mattering once the human gets to the accelerating part of the exponential curve.

Obviously you programmers aren't very good at math.  

Reply #8 Top

It's not exponential.

Reply #9 Top

But it is.  2 factories with adjacency bonuses provide +60% production (+25% from each one plus +10% from adjacency).

Add a third with adjacency bonuses and it becomes +120% production (+25% from each one for +75%.  3 adjacency bonuses means an extra +15% on each factory.  75+15*3 = +120%).  

Arrange 6 factories in a hexagonal pattern around 1 factory and you get:

7 factories * 25% = 175% (base factory production ignoring adjacency)

Central factory gets 6 adjacency bonuses for an extra +30%.  

6 of the outer ring factories gets 3 adjacency bonuses for +15% each.

You're looking at 175+30+6*15 = 295% production.

 

Now add in raw production boosts from starbases:

Surround your planet with 4 economic starbases, and that's a +40% boost to RAW PRODUCTION.  Add in the interstellar government line for another +60% boost in raw production, and we have a +100% in RAW PRODUCTION.

The factories MULTIPLY the raw production.

So imagine the pop of the planet was 6, for 6 raw production.

(6 *2 (raw production bonuses))* 2.95 = 36 production.

In other words, 7 factories and 4 economic starbases with the interstellar government line will multiply production on a planet by a MULTIPLE OF 6.

Please do the math.  

 

Reply #10 Top

MULTIPLE OF SIX doesn't mean 'exponential'. If you had 15 population on the same planet, you'd get (15*2)*2.95=88.5, which isn't an exponential of 15 (well, it is, but it's just over 15^1.65, which is not remotely how the number has been arrived at).

Your empire grows exponentially if every time you land a ship you immediately begin building colony ships there - you go 1, 2, 4, 8, 16, 32, 64 etc. You're increase by a power of 2, because you're doubling every time. Hence, exponential growth.

 

Your colonies don't increase exponentially. The distinct bonuses get multiplied by each other, but nothing ever applies a power to something else. The adjacency bonus is exponential, but only up to 3 - because you can't place 4 factories which all adjacent to each other. 

 

You're just making yourself look foolish with this one, Marigold.

Reply #11 Top

Ok, fine technically it's POLYNOMIAL TO THE POWER OF THREE growth.  But that doesn't sound nearly as awesome as IT'S EXPONENTIAL GROWTH.  

According to Taylor series, it's hard to tell apart between the two in short time frames.  EXPONENTIAL GROWTH CAN BE APPROXIMATED (up to a certain point) USING POLYNOMIAL FUNCTIONS.  BASIC CALCULUS.  

Regardless of what it's called, the basic problem is Gal Civ III economies in the hands of a competent human grow MUCH FASTER than Gal Civ II economies.  THIS LEADS TO MASSIVE BALANCE ISSUES WHERE THE AI, BEING ECONOMICALLY INCOMPETENT SIMPLY CANNOT KEEP UP.  THE COMBINATION OF ADJACENCY BONUSES WITH STARBASE MODULES WITH LINEARLY GROWING POPULATION MEANS POLYNOMIAL TO THE POWER OF THREE GROWTH (AKA "EXPONENTIAL GROWTH").  

In Gal Civ II, the ONLY way to increase production was through population.  And pop grows very slowly.  The factories and research labs and starbase modules DID NOTHING to increase base production.  Their job was to CONVERT raw credits from population to other forms like starships and what-not.  

Reply #12 Top

Arguing about the symptoms does not solve the problem. Since there are multiple threads discussing the economic model/system in GCIII, my hope is that it will get some attention from SD or at least modders - like Naselus or others - with the skill and heart to tackle it.

Reply #13 Top

We need to diagnose the problem before finding the right solutions.  Stardock programmers are apparently not very good at math.  For some silly reason they think the planetary production wheel is the root cause of the evil.  IT IS NOT.  Consequently it's necessary to beat them over the heads with multiple threads on this issue before they'll start to understand.  

Reply #14 Top

As you increase your production, the rate at which you can increase your production increases.

This is exponential growth.

Consider: first constructor takes 8 turns. 

Now that your planet is surrounded by an economic starbase, it can now produce the next constructor in 7 turns.  

Now that you've got 2 economic starbases, you can now build the NEXT economic starbase in 6 turns.  

By the time you have 2 economic starbases and TWO factories, you can build the next starbase in HALF THE TIME than originally.  Which in turn will increase your rate of production EVEN MORE so now you can build the NEXT economic productivity improvement EVEN FASTER.

SEE MY POINT????????

While Stardock programmers might be reasonably competent at programming, they appear to be not-that-competent at math.  Otherwise they would not have designed the economy of this game like this.  

Reply #15 Top

[quote who="marigoldran" reply="13" id="3586673"]
We need to diagnose the problem before finding the right solutions.  

 

The problem is not unknown. I should have written that an argument about the "definition" of the symptoms would not solve the problem. However, arguing about the symptoms of an illness has never cured anyone. Only action matters. A German saying is roughly, "don't fight with your arm, but with your head".  

Reply #16 Top

Frogboy - I don't mean to sound rude or anything but if you seek feedback you have to listen to it and respond instead of dissing it. I am no expert but it seems to me that a lot of people in the forum are making the point that specialisation is the way to go given the design of the game. But to say that you find the wheel too gamey and therefore is a good enough reason to nerf it sounds a little strange. Its like saying this game is mine to do as I will. Sure. But all your customers have bought into it and I am sure you don't mean to imply that you have no obligation to any of them. I think all that players want is a level of control which they feel is being taken away. So it may be nice if alternatives are presented such as totally customisable governors or whatever. I have been having lots of fun with GC3 and we all want it to be even better.

 

Edited I just saw the "Running a galactic civilization"n post by Frogboy which elaborates on this point. Ok. I appreciate that most customers don't get beyond normal level and there is an obligation to meet their requirements. Still its the hardcore user who will generate the ideas and designs that will make things better for the rest. Like Formula One cars :). So keep that window open.

Reply #17 Top

Quoting Kreissig, reply 15

[quote who="marigoldran" reply="13" id="3586673"]
We need to diagnose the problem before finding the right solutions.  

 

The problem is not unknown. I should have written that an argument about the "definition" of the symptoms would not solve the problem. However, arguing about the symptoms of an illness has never cured anyone. Only action matters. A German saying is roughly, "don't fight with your arm, but with your head".  

I'm not a programmer so there's not much I can do on that front.  The only thing I can do is to lobby for changes by pointing out the mathematical design flaws in the game.  

WHICH I MOST CERTAINLY HAVE BEEN DOING.  

Reply #18 Top

The major difference between Gal Civ II and III is that in II factories did NOT increase the productivity of a planet.  What factories did instead was they CONVERTED economic productivity from population into other forms like starships.

In contrast, in Gal Civ III factories and starbases directly INCREASE the productivity of a planet, leading to exponential feedback loops.  Building a factory on a planet increases the economic productivity of a planet, thereby increasing the rate at which you can build another factory.  In Gal Civ II you couldn't do this because you'd go bankrupt if your population wasn't large enough to support the production from the factories.