Genetic AI Project
- Thread starter The Great Apple
- Start date
The Great Apple
Big Cheese
Unfortunetly neither are really very helpful in this situation. The best idea is to upload all of the logs, except perhaps MPLog.txt which can get a bit big (the last 50 or so lines from this are useful however).Samuelson said:
Have you still got autosave enabled? I tested about 300 autosaves, but I can't test everything.
UglyBoy
Pseudo Impossible
Here's data set 84 done. Seems you fixed that bug neatly in v0.33 
I'll do some more when I've got time.
I'll do some more when I've got time.
The Great Apple
Big Cheese
It should be "CheatCode = chipotle".
PatrickDockens
Chieftain
I'm not sure if its because of your memory leak fix, but the .33 version seems to be moving along much faster than previous versions. Cool! 
The Great Apple
Big Cheese
Okies. Big issue - cheating.
The AI cheats. It can see units and cities that it shouldn't be able to. I've confirmed this with tests.
Two options:
1) Leave it as it is. While it may cheat it doesn't cheat much, or in really obvious ways.
2) Try and fix it. This would make the AI worse, but would give us a more pure, and probably fun, game. It would require quite a bit of work and a load of new code to deal with unknown factors.
Personally, while I'd like to say 2... it's alot of work, and it'd probably be me doing it, so I'd like to hear other people's opinion before I dive in. How crucial is a non-cheating AI? IMO Firaxis made a decision to have it cheating for the sake of competance, and perhaps for the sake of simplicity, however I believe it could be fixed without that much of a loss of competance.
The AI cheats. It can see units and cities that it shouldn't be able to. I've confirmed this with tests.
Two options:
1) Leave it as it is. While it may cheat it doesn't cheat much, or in really obvious ways.
2) Try and fix it. This would make the AI worse, but would give us a more pure, and probably fun, game. It would require quite a bit of work and a load of new code to deal with unknown factors.
Personally, while I'd like to say 2... it's alot of work, and it'd probably be me doing it, so I'd like to hear other people's opinion before I dive in. How crucial is a non-cheating AI? IMO Firaxis made a decision to have it cheating for the sake of competance, and perhaps for the sake of simplicity, however I believe it could be fixed without that much of a loss of competance.
Arlborn
Legendary Noob
- Joined
- Jul 29, 2006
- Messages
- 1,939
After to long palying, Im convinced, let AI cheat a long as it is not a major cheat like this one..We cheat also with reloading, wolrd build and such and besides that, AI intelligence is not that big...I believe that AI seeing some units that it didnt suppose to see(I already knw it from other threads) is ok cuz if not AI would make even more suicidal attacks..But well, just what I tihnk!
The Great Apple said:
I wouldn't waste your time with the AI cheat.
1. It needs it, as much as we'd like to think otherwise, there is just too much going on and some things that human players can readily identify (A strong foe, a well defended city, troop movements from a city you can see to a city you can't...) that the AI would have a hard time making informed guesses about.
2. It will take time away from the main focus of this project where I'm sure you can always find a dozen little things to be coding away on.
my 2 cents...
PatrickDockens
Chieftain
If its a major cheat, do you think the AI could learn to play well without it? If so than I say get rid of it! What is the ultimate goal of the project? To make the AI better at making decisions assumably without cheating, right? I think it should be fixed, but should not supercede more immediate goals, especially if it is a minor cheat.
The Great Apple
Big Cheese
http://forums.civfanatics.com/showpost.php?p=4545836&postcount=38DaveMcW said:
The easiest one to identify is the naval pillage cheat. The AI will know if a resource is guarded or not, and if it is, it'll ignore it for potential pillaging. The AI doesn't have to have LoS to the plot to make this decision. This is the easiest to spot as it looks at every plot on the map, rather then searching within a range around itself like other algorithms.
Other things involve knowing how well a city is guarded (based on best unit, number of units, and number of counter-attackers possible), how dangerous a plot is based on units it cannot see... basically pretty much all of the unit tasks assume that the AI is fully aware of the whole map based on it's tech level.
Arlborn
Legendary Noob
- Joined
- Jul 29, 2006
- Messages
- 1,939
Man, its already proved as well but I have no idea where to find the thread back..DO this(or something like), put a strong stack fortified in a part of the map(sorry have no idea of the word in english so i try to explain) that is the easiest and fastest way to somebody pass from 1 part of the world to other(or from 1 CIV to other without pass by enemy territory)..Then declare war with the AI(I would prefer the AI declare war on you and the AI would be better be strong) that is in the other side of this narrow passage that you are fortified in..They CANNOT SEE IN ANY WAY YOUR STACK and still they know its there and they just go by the worse-slowest way to a city of you. Take off the stack and watch the AI changeing its way to go by that narrow trerritoy..Put stack back and see they going again by the slowest way..Do it again and again and have fun and laugh on the AI untill you can ask for peace!
Its really in some thread, I guess in warlrods part, but cant find it..
Its really in some thread, I guess in warlrods part, but cant find it..
Iustus
King
Arlborn said:Man, its already proved as well but I have no idea where to find the thread back..DO this(or something like), put a strong stack fortified in a part of the map(sorry have no idea of the word in english so i try to explain) that is the easiest and fastest way to somebody pass from 1 part of the world to other(or from 1 CIV to other without pass by enemy territory)..Then declare war with the AI(I would prefer the AI declare war on you and the AI would be better be strong) that is in the other side of this narrow passage that you are fortified in..They CANNOT SEE IN ANY WAY YOUR STACK and still they know its there and they just go by the worse-slowest way to a city of you. Take off the stack and watch the AI changeing its way to go by that narrow trerritoy..Put stack back and see they going again by the slowest way..Do it again and again and have fun and laugh on the AI untill you can ask for peace!
As I pointed out in the other thread, this works two ways. Use the worldbuilder to an enemy stack in this spot, and tell your troops to "Goto" the other side, and they will automatically path the long way, somehow knowing the stack you cannot see is there as well.
This is just how pathfinding works, and frankly I think it would be a major overhaul to change it. You would almost certainly have to add some way for the computer to remember where it saw units in the past, and assign them a probability of locations, based on how recently they were seen, what their predicted actions might be, etc. Not to mention perhaps splitting a stack, to assign a unit to watch the gap, to see if the units are moved, while moving the rest the long way, or deciding to attack the stack, to clear the way. This is way beyond anything that is currently being done.
While I would love for the pathfinding to ignore units, and automatically break for a decision when there is a blocking unit (having the AI actually decide whether to go around or attempt to destroy the blocking stack), even this would probably be a large undertaking to avoid making the AI incredibly easy to trap into fighting over losing bottlenecks while avoiding undefended terrain.
Definitely option (1). I do not believe you would be able to do (2) without a near rewrite, as there are some fundamental things that would have to be completely redone in order to achieve the same level of competence. Probabalistic guesses as to unit locations perhaps not even being the most difficult of them.
-Iustus
Aussie_Lurker
Deity
My problem with the AI 'pathfinding' is that in the wrong hands-so to speak-it can make the AI easy to manipulate. I have heard of too many instances where a player has had a stack of enemy units moving back-and-forth, aimlessly, by simply shifting his units in and out of a city. To me, this classes current AI pathfinding as both a Cheat and a player Exploit. As such, it should be replaced with something better.
(Oh, btw, many people said that removing 'resource omniscience' in Civ3 would make things too hard for the AI-yet it WAS removed, for Civ4, with the overall effect of greatly improving the gameplay experience!)
Aussie_Lurker.
(Oh, btw, many people said that removing 'resource omniscience' in Civ3 would make things too hard for the AI-yet it WAS removed, for Civ4, with the overall effect of greatly improving the gameplay experience!)
Aussie_Lurker.
Iustus
King
Aussie_Lurker said:
My point is that without doing a major rewrite, changing this would only make such things more exploitable. Instead of being slightly difficult to do, it would become quite trivial to trap the AI this way.
Again this makes my point. I am not saying it is impossible, but rather it is as large an undertaking as going from Civ3 to Civ4. It is entirely possible to redo the pathfinding. But you cannot just simply remove the ability to see into the fog. That would cripple the AI. You are going to have to do something a lot trickier.
-Iustus
strategyonly
C2C Supreme Commander
deleted wrong MOD area, sorry.
EDIT: Sorry about that was playing your mod, and thinking of another.
EDIT: Sorry about that was playing your mod, and thinking of another.
Similar threads
