MrCynical
Deity
Soren Johnson works at Firaxis. He probably knows something about how the calculations work. 
Does the AI have preknowledge of the map?
- Thread starter Goombaz
- Start date
Soren Johnson said:
I truly appreciate you posting, testing the map and even spending any time on this Soren; thank you
Nice to know this is going to be tightened up a little. Please bear in mind that the test also works for resources that are out of sight, but 'known' ... in other words, I can get the same effect by using DYE that hasn't been explored ... the AI will still head for it. Same leakage?
Psyringe
Scout
Soren Johnson said:
Hi Soren, thanks for stopping by.
This explanation is consistent with our findings. It also explains why, in Condor's first test, one additional grass tile at one arm was enough to draw the AI's attention from the oil tile. The two food that the grass tile provides are valued higher than the one hammer that leaked through the yield calculation for the oil. I think we can close this case then.
There's another issue with the view distance of the AI. It seems that the AI is able to spot the best settling place even in (for the AI) unexplored territory. Can you tell us whether this is how it should be, or a bug, or whether we made some mistake in our tests?
Efexeye
Silly Donkey
CivScientist said:
Uh, he's the lead designer on the game?
CivScientist
Warlord
- Joined
- Nov 7, 2005
- Messages
- 125
Ok. I didn't know and there were no introductions.
Efexeye
Silly Donkey
Heh, Kolyana.
Scientist, it's all good- the Firaxis avatar and title "Firaxian" mean that the poster is someone who works for Firaxis.
Scientist, it's all good- the Firaxis avatar and title "Firaxian" mean that the poster is someone who works for Firaxis.
Exavier
Destroyer of Worlds
Kolyana said:
Lmao... i thought thats how all game developers communicated with the public
Meffy
humanoid skunk
<EdSullivan> "We got a really big show for you tonight, a really big show... ladies and gentlemen -- THE CREATOR!" </EdSullivan>
o/~ She loves you, yeah, yeah, yeah... o/~
o/~ She loves you, yeah, yeah, yeah... o/~
The Condor
On an excellent adventure
Hi, I'm back. Cool. Soren Johnson stopped by. Wish I would have been here earlier
. Anything else I could do for someone? Also, if oil is >800 what is the most heavily desired thing for an AI? Do you also have a chart or something that ranks how important anything is to an AI? That would be nice and thanks if ever come Soren.
. Wish I would have been here earlier. Anything else I could do for someone? Also, if oil is >800 what is the most heavily desired thing for an AI? Do you also have a chart or something that ranks how important anything is to an AI? That would be nice and thanks if ever come Soren.@Condor (or anyone who likes testing):
Would you like some more tests to run? Try the following:
First Test
Purpose - to test whether or not the AI can tell what kind of resource is in a tile even if it can not "see" the resource due to a lack of the needed tech.
Create a small island -9X9, all grassland, all with rivers for fresh water access- with the AI starting in the center of the island.
- Starting units: 1 settler, 12 workers, 1 archer (The 12 workers is intentional. It is to make doubly sure that the AI can develope the land *before* researching the techs needed to see the resources...and to speed the test.)
- Resouces within 1 square of the setler (ie. when the setler founds a city these resources should be within the cities borders): 1 coal, 1 oil, 1 alumanim.
-Starting techs: Mining and Agriculture ONLY.
- Difficulty level: Nobel
Does the AI build mines on the coal and aluminum (even though farms should have been a better choice given what the AI is supposed to know) and NOT on the oil?
Second Test
Purpose - to test whether the AI sees the "unrevealed" terrain leading up to a resource, or just knows the direction that the resource is in.
Create a small, 9X9 island, all grassland, with an AI city in the middle. Create a 1-tile-wide path that heads north for 2 tiles, turns west for 8 tiles, heads north for 2 tiles, and then heads east for 8 tiles until it is right above the starting city. At the end of this path place an oil resource. Now, create a 1-tile-wide path leading off of the islands east coast and heading as directly towards the resource as possible.
- Starting units: 1 settler, 2 archers
- Resources: as mentioned
- Starting Techs: Nothing at all!
- Difficulty level: Nobel
Does the AI see that the resource is to the north -but not how to get there- and head along that path, or does it head away from the resource to the east and along the more direct route?
Third Test
Purpose - to see how far the AI will travel to get a special resource and to see if this changes based on map size.
Create a Duel-sized map with *no* water. Place your starting city and the AIs starting city on oposite ends of the map so they don't conflict.
- Starting units: Settler, 2 archer.
- Resources: one special resource 5 tiles away from the AI city.
- Starting Techs: None!
- Difficulty level: Nobel
Does the AI head strait for the resource? If so, repeat this test with the resource farther out. Keep repeating this test, while increase the distance to the resource, to see how far the AI is willing to travel to get a special resource.
Repeat this test on different size maps to see if this distance is different based on the size of the map.
Would you like some more tests to run? Try the following:
First Test
Purpose - to test whether or not the AI can tell what kind of resource is in a tile even if it can not "see" the resource due to a lack of the needed tech.
Create a small island -9X9, all grassland, all with rivers for fresh water access- with the AI starting in the center of the island.
- Starting units: 1 settler, 12 workers, 1 archer (The 12 workers is intentional. It is to make doubly sure that the AI can develope the land *before* researching the techs needed to see the resources...and to speed the test.)
- Resouces within 1 square of the setler (ie. when the setler founds a city these resources should be within the cities borders): 1 coal, 1 oil, 1 alumanim.
-Starting techs: Mining and Agriculture ONLY.
- Difficulty level: Nobel
Does the AI build mines on the coal and aluminum (even though farms should have been a better choice given what the AI is supposed to know) and NOT on the oil?
Second Test
Purpose - to test whether the AI sees the "unrevealed" terrain leading up to a resource, or just knows the direction that the resource is in.
Create a small, 9X9 island, all grassland, with an AI city in the middle. Create a 1-tile-wide path that heads north for 2 tiles, turns west for 8 tiles, heads north for 2 tiles, and then heads east for 8 tiles until it is right above the starting city. At the end of this path place an oil resource. Now, create a 1-tile-wide path leading off of the islands east coast and heading as directly towards the resource as possible.
- Starting units: 1 settler, 2 archers
- Resources: as mentioned
- Starting Techs: Nothing at all!
- Difficulty level: Nobel
Does the AI see that the resource is to the north -but not how to get there- and head along that path, or does it head away from the resource to the east and along the more direct route?
Third Test
Purpose - to see how far the AI will travel to get a special resource and to see if this changes based on map size.
Create a Duel-sized map with *no* water. Place your starting city and the AIs starting city on oposite ends of the map so they don't conflict.
- Starting units: Settler, 2 archer.
- Resources: one special resource 5 tiles away from the AI city.
- Starting Techs: None!
- Difficulty level: Nobel
Does the AI head strait for the resource? If so, repeat this test with the resource farther out. Keep repeating this test, while increase the distance to the resource, to see how far the AI is willing to travel to get a special resource.
Repeat this test on different size maps to see if this distance is different based on the size of the map.
The Condor
On an excellent adventure
O.k. I would love to do it right now but I am at a different computer (it does not have civ on it) so I will do it in an hour or two. Is that o.k.?
The Condor
On an excellent adventure
I did the first test and the AI (I did the Chinese guy with the poofy hat and blue shirt since he starts with agriculture and mining) did not work the terrain with the "hidden resources" Instead he sent three of the twelve workers out to work land that did not have resources.
Here is a pic of the isle:
I also got the save in case anyone is interested. Is this what you were talking about?
Here is a pic of the isle:
I also got the save in case anyone is interested. Is this what you were talking about?
The Condor
On an excellent adventure
Hi, again. I did the second test and believe it or not (I got the save however) the second settler plopped a city not near the oil resource but within reach of the capital and also I think it settled their and not by the resource because a barbarian lion appeared next to the resource in worldbuilder and also because it would have the best amount of workable tiles there
.
.Wow...awesome! Thanks man. I would do these myself but I'm just sooo lacking in time to do much testing. Yeah that first test was just what I was wondering. It seems that even if the formulae tells the AI that there's a special resource there the AI doesn't actually know what it is. That's good news
Interesting result with the second test. Did you try repeting it? Did the AI always head in one direction before founding the second city? What I'm wondering is if the computer went to go towards the resource, and not knowing how long it would take to get there just decided to settle a city before getting too far away from the capital (that would increase costs). If the AI setteled a second city every time in the same direction that might indicate this...but it would also depend on how the map was set up. I'll know more when I get a chance to look at it.
Interesting result with the second test. Did you try repeting it? Did the AI always head in one direction before founding the second city? What I'm wondering is if the computer went to go towards the resource, and not knowing how long it would take to get there just decided to settle a city before getting too far away from the capital (that would increase costs). If the AI setteled a second city every time in the same direction that might indicate this...but it would also depend on how the map was set up. I'll know more when I get a chance to look at it.
The Condor
On an excellent adventure
I did the third one and the AI went five spaces away at the max for the resource (it moved six total[went down one space so it would not settle the oil]).
ZippyRiver
Prince
- Joined
- Nov 27, 2005
- Messages
- 454
Until the resource is available from tech discovery, I don't believe you can mine a grass tile. The resource appearing is what enables the worker to build a mine. And with all the other grassland tiles, there would be no reason to farm the resource only to have to replace it later. Perhaps if the resource tiles were grassland, and the other tiles jungle/plains. But even then, temporary farming is not a bad idea.
onedreamer
Dragon
MrCynical said:
Of course I would. The purpouse of beta testing is not just to play the game and enjoy it but to find out bugs that the developers didn't think of, because if they would have thought about them, obviously they would have corrected them. Hence during beta testing a tester should try to do everything that comes to the mind, even the most ridiculous things. This issue that is being tested is not very hidden, I understood that the AI has some knowledge of hidden resources quite simply through the blue circles when I try to settle. Now the problem is that even if I have this advice, I don't know where exactly and which are the resources, while the AI does.
As a bottom line, if a bunch of players whom are not beta testers came up with this thread and are cooperating to test something, I don't see why this couldn't have come to the minds of the "real" testers.
onedreamer
Dragon
So in short, correct me if I'm wrong, Soren said that the AI does not know of a resource being there if it doesn't have the technology to see it (wether it also can't know of a resource if it didn't previously explore that portion of map he didn't specify); though it knows of the effects the resource will have in the total yeld of a plot (without improvements I guess). This is not working as intended and should be fixed with the next patch... well I think we can consider this thread a success then
MrCynical
Deity
For the first test the coal and aluminium are on grassland tiles, so I don't think the AI can actually mine them at this point. I haven't seen them break the rule that you can't mine a flatland tile with a resource until the resource has actually been discovered. If you made all the city's tiles hills do they go for the coal and aluminium tiles to mine first?
@Onedreamer: I am still of the opinion that these tests only seem obvious after a substantial amount of time playing the conventional game (It was nearly two months for this thread to appear) and when you know more or less what you're looking for. One again I state that I doubt that the Beta testers would create these test maps for each and every rule. From Soren Johnson's information the effect is fairly small, and therefore only going to be clear on normal maps very occasionally. The number people on this forum and the amount of time played must by now exceed the amount in the Beta test by several orders of magnitude. It is only as a result of this that it is evident these individual tests are worth doing. You state a beta tester should do whatever comes to mind, and I agree. I'm just not convinced this would have come to mind.
@Onedreamer: I am still of the opinion that these tests only seem obvious after a substantial amount of time playing the conventional game (It was nearly two months for this thread to appear) and when you know more or less what you're looking for. One again I state that I doubt that the Beta testers would create these test maps for each and every rule. From Soren Johnson's information the effect is fairly small, and therefore only going to be clear on normal maps very occasionally. The number people on this forum and the amount of time played must by now exceed the amount in the Beta test by several orders of magnitude. It is only as a result of this that it is evident these individual tests are worth doing. You state a beta tester should do whatever comes to mind, and I agree. I'm just not convinced this would have come to mind.
