The AI likes Apples...

Arkatakor

King
Joined
Mar 11, 2006
Messages
619
Location
Stockholm, Sweden
This was what the AI was willing to give me for some apples (see attachments); I play with tech diffusion off, so all the techs he was offering were totally unresearched by me. He offered so much I had to take 2 screenshots to scroll down and display the stuff he was offering!
 

Attachments

  • apple1.jpg
    apple1.jpg
    374.6 KB · Views: 506
  • apple2.jpg
    apple2.jpg
    374.4 KB · Views: 390
He likes his apples I suppose!

Just curious, how did you get your resolution and diplomatic options like so? I can't get the game to stretch the UI for my laptop, only the entire thing, which is wrong. And for me diplomacy is full screen.
 
I've noticed this myself in a recent game - in particular Pleased or Friendly AI players are quite happy to give you almost everything they have for certain resources, and not necessarily strategic resources such as oil or uranium. It feels a bit broken to be honest as it's too easy to make huge leaps in score whilst simultaneously damaging their economy.
 
He likes his apples I suppose!

Just curious, how did you get your resolution and diplomatic options like so? I can't get the game to stretch the UI for my laptop, only the entire thing, which is wrong. And for me diplomacy is full screen.
I have not changed any settings; however my resolution is 1080 * 1920 which is probably why it fits the way it does...
 
I have not changed any settings; however my resolution is 1080 * 1920 which is probably why it fits the way it does...

I used to play at that resolution but I got bad graphics problems latter in the game. Unit polygons would extend into space somewhere. have you had that problem? If not I may go back to it as it is easier on the eyes.
 
Tech Diffusion


Tech Diffusion causes civilizations that fall behind technologically to slowly recieve free research towards technologies that other civilizations already posess. The more civilizations posess a technology, the faster other civilizations will recieve free research towards it.


how exactly does that make you loose old tech.. i dont get it.. as to that, now food resources are very expensive and you can get quite a lot early for them

as to what hes offering when looking at your tech, its 2-3 tops turns for you to research and useless research for you.. so hes basically offering nothing of value ..
 
I've noticed this myself in a recent game - in particular Pleased or Friendly AI players are quite happy to give you almost everything they have for certain resources, and not necessarily strategic resources such as oil or uranium. It feels a bit broken to be honest as it's too easy to make huge leaps in score whilst simultaneously damaging their economy.

Try going to an AI and asking for a resource like Dye. They will ask for the moon just as they are now giving the moon.

JosEPh
 
Try going to an AI and asking for a resource like Dye. They will ask for the moon just as they are now giving the moon.

JosEPh

yep.. having 3 dye resources allow me to buy half world every 10 turns
but i have to sell same world to get marble
same with stone.. having couple of it makes you billionaire.. makes me feel bit bad about making those deals
 
Try going to an AI and asking for a resource like Dye. They will ask for the moon just as they are now giving the moon.

JosEPh

I agree that's true. However a human player will not agree to give the AI the moon, whilst the AI will. So on balance, the AI loses, and loses a lot.
 
Try going to an AI and asking for a resource like Dye. They will ask for the moon just as they are now giving the moon.

JosEPh

Indeed. I think this is really broken. You're asking for a single resource and then they demand a dozen of yours, hundreds of gpt, and a few techs.
 
The problem is AI as it is in Civ4 engine and further refined by Afforess has yet to be a long term thinker.
 
how exactly does that make you loose old tech.. i dont get it.. as to that, now food resources are very expensive and you can get quite a lot early for them
He said nothing about "losing" old techs, just that he never bothered to research most of those.

as to what hes offering when looking at your tech, its 2-3 tops turns for you to research and useless research for you.. so hes basically offering nothing of value ..
Researching them would probably be somewhat of a waste of time, but getting them for free also saves him the turns it would take to research those, and also they might be prereqs for some techs that he DOES want, so I think you underestimate their usefulness.
 
Not sure if Afforess has looked at this yet, but I recently had a really bad example in a C2C game. Essentially it allowed me to go from a difficult (but interesting) situation to an easy game very fast. I was the power chart leader of the 4 civs I knew, but one of them was significantly ahead technologically (like at least 100 turns of research worth), and from 'Most Advanced Civ' pronouncements I knew he was nowhere close to being the tech leader, so things looked like they were going to be an up-hill struggle longer term. Anyway, he suddenly decided he was prepared to trade most of his techs (I'll go into why later, but I don't think there is [much of] an issue there as it turns out), and I was able to offer him a resource he lacked, for which he gave me 4-5 civs. I was able to end the agreement every ten turns and extract a similar batch of techs each time until I was pretty much caught up after about 30 turns. This is obviously rather exploitative!

Anyway, I have drilled into what's going on and here are my findings and the changes I am making:

1) Willingness to trade a tech at all (apart from special reasons like he's building a wonder dependent on it, or he just hates you) depends in very large part on how many of the players the owning civ knows about already have the tech. So in my game, only one player I knew had all these techs, but HE knew about 8 others that all had them, and consequently he devalued them considerably because 'most of the world knew them'. This aspect I think is basically ok. I'm wondering if maybe I should tweak it a bit so that the AI is also less inclined to trade to someone 'up the power chart' from them, but I don't feel strongly. opinions welcome on this one....

2) The real problem is that BTS originally never allowed per-turn trade items to be traded for permanent items (e.g. - resources-for-peace-treaty-time for techs), and when this was introduced the scaling of the valuations between the two categories was a bit arbitrary. In particular the code is loosely intended to normalize everything to a gold value, so for gold-per-turn for example it just multiples by the peace treaty length. However the valuation of resources uses the same routine that the city AI uses for assessing resource importance which has no such normalization concept (as its just used to compare against similar values really). In particular it scales in number of buildings and units that have some sort of dependency on the resource, regardless of their OTHER pre-reqs, so as buildings and unit types are added (relative to BTS AND adds a lot, and C2C a lot more still) the resulting values tend to inflate.

I have made the following changes (all to resource valuation):

i) When assessing the value for trade purposes, buildings or units that cannot yet be constructed are disregarded (on the basis that the trade is a short term thing, so taking account of future techs/civics/religion choices makes little sense)

ii) Similarly buildings/units that are dependent on religions are now weighted by the CURRENT presence of that religion in the player's cities at the time of the trade (i.e. - don't count buildings that require religions we don't have). Same for corporations (Sid's Sushi distorts coastal resources a lot for example even when its nowhere near being founded, and nobody could possibly care about it)

iii) Building values are divided by 3 in general to reflect the fact that if a building DOES have a resource requirement it's going to go non-functional when the trade ends anyway, even if you manage to build it. Similarly units, though not reduced by as much (reflects the fact that you won't have time to build many)

iv) When assessing a resource value for trade purposes an overall divisor is also now present. This is just an approximate normalization factor based on what I was seeing to account for the fact that really resources and other trade items have values on unrelated scales (i.e. - it's a fudge factor)

I'll be pushing these changes to SVN (both C2C and AND) later today hopefully. Just doing some final testing first. Nothing will be affected apart from trades that involve resources (and obviously the relative values of resource for resource won't change as category, though individual resource types will move up/down due to removal of some distortions like the Sid's Sushi effect I mention above).

Edit - changes now pushed to SVN (including DLL update)
 
Not sure if Afforess has looked at this yet, but I recently had a really bad example in a C2C game. Essentially it allowed me to go from a difficult (but interesting) situation to an easy game very fast. I was the power chart leader of the 4 civs I knew, but one of them was significantly ahead technologically (like at least 100 turns of research worth), and from 'Most Advanced Civ' pronouncements I knew he was nowhere close to being the tech leader, so things looked like they were going to be an up-hill struggle longer term. Anyway, he suddenly decided he was prepared to trade most of his techs (I'll go into why later, but I don't think there is [much of] an issue there as it turns out), and I was able to offer him a resource he lacked, for which he gave me 4-5 civs. I was able to end the agreement every ten turns and extract a similar batch of techs each time until I was pretty much caught up after about 30 turns. This is obviously rather exploitative

2) The real problem is that BTS originally never allowed per-turn trade items to be traded for permanent items (e.g. - resources-for-peace-treaty-time for techs), and when this was introduced the scaling of the valuations between the two categories was a bit arbitrary. In particular the code is loosely intended to normalize everything to a gold value, so for gold-per-turn for example it just multiples by the peace treaty length. However the valuation of resources uses the same routine that the city AI uses for assessing resource importance which has no such normalization concept (as its just used to compare against similar values really). In particular it scales in number of buildings and units that have some sort of dependency on the resource, regardless of their OTHER pre-reqs, so as buildings and unit types are added (relative to BTS AND adds a lot, and C2C a lot more still) the resulting values tend to inflate.

This wasn't addressed sooner, wow. This is really an AI bad behavioral matter and like you said very exploitative.

2. So more or less what your saying is that the "programmer" for BtS, just wanted to get something that "fairly" worked and just more or less put something in not knowing really what he was doing. A fast work around.:(

I am so glad SOMEONE understand what the heck is going on, thx.
 
Koshling...if what you're saying is correct & your changes work...I would imagine this will have a huge effect on gameplay.

To think this behavior wasn't researched sooner is surprising to say the least. Great job on delving into this and perhaps making a fairly large AI improvement : )
 
Is it normal that tech is valued half as gold? I would have thought it should be valued more than gold, since gold is easier to get...

Well...I noticed this when I went through the code. It's actually 66% of gold (at least before modifiers), and there is an explicit comment from Afforess in the code saying 'Techs are not work more than gold' next to which he modified the BTS original value of techs being worth 150% of gold to the current AND 66%. I suspect Afforess did this because there are mechanisms to turn gold into beakers (indirectly, and to some limitted extent) but not really beakers into gold.

I must admit I personally also value tech higher than gold and generally regard the AI as selling techs for gold too cheaply, but it is somewhat situational I think. Since this is deliberate change Afforess made in AND previously I don't plan to touch it, but Aff may like to comment...

If I WERE to change this I would make the conversion rate variable between about 50 and 200% of gold, depending on whether the AI is behind in research or in available gold (demand) and similarly for the proposed trading partner (supply). However, I don't think it's really a major issue, and I certainly wouldn't want to fiddle with it until the other changes I made have had time to be assessed by people play testing.
 
Top Bottom