I'm disappointed that Brad hasn't found this to also be true...

Well, I can understand Brad's reaction (or lack thereof) perfectly well.
You see, with GalCiv1, the same thing was going on over and over again - people accusing the GalCiv-AI of cheating and providing 'evidence'.
Now, in 95% of the time, those people just didn't understand the game mechanics (for example, all the 'How can the AI build ships so fast! Cheat!' - people : Well, it simply buys them...), or they didn't have much clue about programing and/or made wild assumptions. (NOTE: I'm not saying everyone was/is like that. But well, 95%, yeah)
The other 5% were bugs, which were usually fixed by Stardock.

Now it's starting again. See, there is nothing wrong with detecting unusual behaviour of the AI and report it.
But there is a definite difference between saying:
"Look, Brad, something is strange here and there, could you look into it?"
and saying:
"Your AI sucks you cheat!!!!!!"

This attitude is the problem, guys. Just assuming that Brad is lying all the time about is AI not cheating. You see, Brad&Co. aren't SOE or EA. They hang out at the IRC, they do an incredible amount of updates, they hang out in the forums, they even post code snippets. And usually, the listen to there fans. But if they are approached with a hostile attitude and half-baked claims, why the hack should they waste their time over and over again defending themselfes?
For that I learned from similiar GalCiv1-discussions: Everytime Brad proved peoples accusations wrong, they would just go on trying to find the next 'definite proof' that Brad's a liar. It wasn't about improving the game, it was all about 'Brad is lying'. Not for all of them, but for most. And yeah, I can understand that Brad prefers to ignore this. I would to.


I hope I didn't come of to fanboyish. To the current discussion, I haven't seen anything so far that I'd accept as proof for the AI knowing where the good planets are when it shouldn't, but some stuff that strongly suggests that there might indeed be something wrong. But I guess, if this should indeed be the case, I bet it's a bug and not Brad trying to fool us. And in this case, devs and fans should work together to hunt it down and fix it. Not against each other.

Oops. That was a rather long posting. Got to go back to work now and fix my own bugs.

Is it possible that it may also share knowledge gleaned from scout vessels? In a large multi-race map perhaps enemy scout vessels not only discover habitable worlds for their race, but also for the other AI? Just a thought.

I don't think anyone has tried to say brad was lying. People have posted claims of weird behaviour, others have confirmed, and I have given a nice objectional view on a possibility where the error might be.

Originally I said that I don't believe there is a bug, but now I think there is because of what I have seen, and been able to reproduce in my example.

Now, to make sure we all get the idea... here are some points:

1) Whenever someone says "The AI Cheats", they really mean "There is a bug in the AI code", they just emphasise it wrongly.
2) It has already been stated several times that the rules are different for the AI as it is for the human, and so if it is considered normal behaviour for the AI to do what it does in my example, then it is a feature, and not a bug.
3) FrogBoy has said himself that the AI will not send a colony straigh to a habitable planet without prior knowledge of its existence from a scout first. My tests dis-proves this (and if you don't believe that, then try it out yourself).

So, I personally don't care if this is changed or not. all I wanted to was was find out what made the original poster and others believe the AI was "sending colony ships when they shouldn't", and this is most certainly how it does it.... It may not be the only way it does it, but it is definately _A_ way.

So, what should be done about it? well, whatever the programmers want to do about it, but now the people experiencing this oddity knows why it is most likely happening...

anyway, this is a forum... everyone argues on a forum

Is it possible that it may also share knowledge gleaned from scout vessels? In a large multi-race map perhaps enemy scout vessels not only discover habitable worlds for their race, but also for the other AI? Just a thought.

If this were true, it would be very disturbing to say the least. Interesting theory though.

So, what should be done about it? well, whatever the programmers want to do about it, but now the people experiencing this oddity knows why it is most likely happening...

Well, I'd personally like to play against a fair opponent, so it should be fixed somehow, no?

Okay this topic is obviously a very important and hot issue.... so my question is
" where are the developers? "

A quick 5_min post from the developers would place this hot important issue to sleep... an example:
DEV_A : The AI does not cheat based on programming... ya da ya da yaa da.
or
DEV_B: Currently the AI does cheat when reaching the good planets... we are working to improve the AI... ya da yada.

I've seen in other game forums where the developers will actually comment on the hot topics.

In the future, if someone wants a serious response, they should consider better titles. Having some guy come on and declare that the AI isn't smart AND that it cheats is a bit off putting and erroneous.

Completely agree...was very bad taste, but it sparked a decent conversation at least.

Well, I have tried the method put by SmackleFunky :

- Launch GC 2 in cheat mode
- start a small universe with 1 opponent
- at the first turn:
- clear the FOW (CTRL + U)
- clone enough colony ship (CTRL + C)
- teleport them near colonizable planets (CTRL + T)
- Colonize all planets except one, which need to be reachable by the AI colony ship but outside its sensor range
- Create a battle ship (CTRL + near the ennemy flagship. Upgrade it (CTRL + A) and kill the flag ship
- hit turn and kill any produced souts

Now you will see that the ennemy colony ship will move toward the only available planet, even if the AI hasn't visited it yet.

I hope that Frogboy will be able to devote some of his precious time to enlight us about what happen in this particular case.

Have you changed it when creating your race? Or was it correct the first time and get screwed woith CTRL+N?

Hmm I got that from ctrl+n I believe. Not sure when the first time it happened. Now when I start a new game in the main menu, it still name the star/system custom star. Home planet has the right name, however.

I also get all kinds of weird home star/system. From 1 planet that can be colonized (home planet) to 4 with one of them being a PQ 23.

Maybe it's some kind of bug.. I'll see what happens after the next update.

But like I said, I loaded up the game in the debugger, played for quite awhile (I was up until 6:30am incidentally) and looked into this. I saw no evidence of cheating and in fact, I saw some flaws in the AI colonization which you can see in there.

Mmm. I was up to 6:30am, too, but that's just because my sleeping schedule likes to run on 26 hours. Dang artificial lighting.

At this point, I'm more inclined to agree with you that I must've missed a scout somewhere. I ran a couple more tests - this time horribly abusing cheat codes. Not to remove fog of war, but to clone and teleport sensor ships all over the place to accomplish roughly the same thing, since people have been concerned that the AI might function differently with that particular code turned on.

I did see some very peculiar behavior on the part of the AI, but not with colony ships. It really does seem that if they don't get a planet within sensors, they won't send colony ships out, not even blindly.

I also saw some other places the AI could be improved. I'm going to move my observations to a new thread. Oh, and if people want the save game files or notes from those tests, give a yell. I haven't bothered uploading them yet. There's one more test I want to run involved with reloading saved games tonight. It seems that just saving a game doesn't do anything, but I've seen some odd things reloading that I want to check out.

crickel

Edit: Well that took all of about 15 minutes to check out. I loaded up a couple saved games and let them run to see what would happen. Didn't see anything odd with any colony ships. So much for my other hypothesis that reloads might be affecting it!

It really does seem that if they don't get a planet within sensors, they won't send colony ships out, not even blindly.

Well that much can be said about the code. AI has demonstrated its inefficiencies as many dare not send out a colony ship until a planet is 'known,' even if all the scouts they build get blown up in the process.

1. Build scouts
2. Secnd scouts.
3. Scouts find planet (planet is 'known')
4. Immediately purchase colony ship and send the planet that way.

From that we can conclude the code is not in err. It is clear that the code is written so the AI would not send out a colony ship until it is 'known.' If there is a bug in the code, it's not in the know planet/colonize planet part. It's how the AI acquire the information on the planet.

Now the question is.. how does the AI know the planet even though the scout got nowhere close to it?

The part of the code that needs to be examined is probably the part that relate to how AI come to know about the planet.

Like SmackleFunky mentioned: "What does the colony ship do if it has no where to go?"
In game, we often see the colony ship just staying in the same spot.. but that's because all the colonizable planet in range have been taken. What if there is a planet in range that is not discovered? Would the colony ship somehow head toward it.

I would suspect the code here might be different because the ship is already in place. Earlier, the ship was purchased and autopiloted to the planet in response to discovering a planet.

What the colony ship would do if it was already created isn't really tested except by SmackleFunky.

There were cases of player reporting that the Ai would send the ship to another planet if the destined planet is colonized.. so maybe that part of the code need examination.
1) head toward destined planet
2) if destined planet taken, automatically head toward next planet... or next 'known' planet?

To further that note;

The code posted by FrogBoy, I believe, is to evaluate whether or not to create more colony ships. And so the error (if there is one) is not necesserily there, as a colony ship may already exist. Secondly, its not how does the AI gain knowledge of a planet, because it has been shown that they never magically gain knowledge of a planet

Rather, how does a colony ship move through space? I believe that somewhere in that code, the AI is going through the list of known, free planets - picking one and heading for it.

But something is going wrong when no known, free planets are find. Possibly not checking to see if that list is 0, which results in the colony ship going to the next planet on the list.

This is my suggestion to the cause of the bug. If anyone else has another suggestion that isn't just a random guess, but something with a procedure to back it up and show examples, then I would like to hear it.

Well that has definitley cleared up my findings - I can no longer do what I mentioned before. So from me.

I guess at some point, under the conditions I gave it, it was not getting initialised properly... odd.. but all is good. I'll try the more complex scenario by crickel just to be sure.

On second thoughts, using Crickel's old save game would be inconclusive, since its a save before the change and therefore unreliable... Oh well, we'll just have to wait and see if it pops up again in a new game.