5937 Feedback

Describe your experience with the latest version of FreeOrion to help us improve it.

Moderator: Oberlus

Forum rules
Always mention the exact version of FreeOrion you are testing.

When reporting an issue regarding the AI, if possible provide the relevant AI log file and a save game file that demonstrates the issue.
Message
Author
AndrewW
Juggernaut
Posts: 791
Joined: Mon Feb 04, 2013 10:15 pm

Re: 5937 Feedback

#16 Post by AndrewW »

Dilvish wrote:AndrewW, the invalid form happens when someone else makes a post while you've been typing yours, and so the post number or somesuch it had tentatively assigned to you is invalid. That can happen in any amount of time, even one minute. Yes, you do retain your post content in that situation. That is not what I'm talking about.
Oh, mine wasn't an invalid.

unjashfan
Creative Contributor
Posts: 175
Joined: Fri Dec 30, 2011 8:08 am

Re: 5937 Feedback

#17 Post by unjashfan »

Here's an example of what I think is bad defense. AI has two colonies at Eltanin and Canopus, and it should know that my empire is close by (minor skirmishes occurred earlier on) even though it can't detect my colonies yet. In those two systems, there's a grand total of just one dedicated warship, and for some reason there are two warships heading back to the capital when they're better off sitting at either of the two colonies in case I decide to attack it. There's even more of a reason to send warships for defense especially when you can't see the enemy (i.e: one can't assess enemy fleet strength, so it's better to have more than enough defenses). I think if the troop ships at Canopus were warships instead, it would really beef up the AI's defenses. I mean, what is it even invading?? If the AI is making them to invade me in advance, I think that's way too early for that.

I couldn't upload the AI log files because they were too big, even in compressed form. The screenshot is around turn 80.
Attachments
FighterMovements.png
FighterMovements.png (370.69 KiB) Viewed 1327 times

User avatar
Geoff the Medio
Programming, Design, Admin
Posts: 13587
Joined: Wed Oct 08, 2003 1:33 am
Location: Munich

Re: 5937 Feedback

#18 Post by Geoff the Medio »

unjashfan wrote:I couldn't upload the AI log files because they were too big, even in compressed form.
Even if you play just one turn? It should allow files up to 1 MiB. If it's still a problem, consider using http://pastebin.com/

User avatar
Dilvish
AI Lead and Programmer Emeritus
Posts: 4768
Joined: Sat Sep 22, 2012 6:25 pm

Re: 5937 Feedback

#19 Post by Dilvish »

unjashfan wrote:...for some reason there are two warships heading back to the capital when they're better off [doing something else]
It doesn't look to me like you can see all of the AI's systems, so especially without access to at least the AI log files it seems pretty pointless to even discuss whether the AI may or may not have had something potentially more important to do with some ships. Screenshots can certainly help clarify a situation in general, but in this type of situation there is a great deal of information they lack.
unjashfan wrote:and it should know that my empire is close by (minor skirmishes occurred earlier on) even though it can't detect my colonies yet.
Unj, I'll try to take it as flattering that you have high expectations of the AI. But it does get a bit tiring trying to explain to people over and over again what the limitations of the AI currently are. I have given the AI some ability to retain and use information about enemy fleets & planets previously seen (or infer information after battling without seeing them, when stealthing allowed that). However, the AI does not currently really have any significant concept of an enemy empire occupying a region of space. It therefore does not try to evaluate what that space might be. It also does not try to assess a general level of technology, or any particular technologies, for an opponent. Not that you mentioned the technology bit, I'm just throwing in that extra bit of info. These are just a few among the many things that the AI could usefully do, and which I hope to someday have it do at least somewhat. But I don't think they are the highest priority at the moment, and I don't expect to tackle either of those cited issues this week, or this month. There are so many things that the AI could usefully do, but corrently doesn't, that I will probably stop re-explaining this pont fairly soon.

If you are impatient for the AI to be stronger and wish to contribute by coding up some additional capabilities for it, you would be very welcome to do so.
If I provided any code, scripts or other content here, it's released under GPL 2.0 and CC-BY-SA 3.0

unjashfan
Creative Contributor
Posts: 175
Joined: Fri Dec 30, 2011 8:08 am

Re: 5937 Feedback

#20 Post by unjashfan »

But it does get a bit tiring trying to explain to people over and over again what the limitations of the AI currently are...If you are impatient for the AI to be stronger
Just want to let you know that I really appreciate all the work on the AI, Dilvish. I'm sorry if I sound impatient :oops: , because I'm definitely not impatient, and I have no right to be. I must have went overboard trying to emphasize what I thought would help the AI. Thanks for clarifying the AI's capabilities - I really did overestimate it (in the good way :D ).

User avatar
Dilvish
AI Lead and Programmer Emeritus
Posts: 4768
Joined: Sat Sep 22, 2012 6:25 pm

Re: 5937 Feedback

#21 Post by Dilvish »

unjashfan wrote:...I still think that the AI makes too many troop ships...
I finally saw this problem appear in a game & was able to investigate it. I've adjusted the AI code to put some more constraints on troop ship production; hopefully it's balanced better now (though this initial adjustment was just that, initial, I expect I'll make some further adjustments to it after a bit).
If I provided any code, scripts or other content here, it's released under GPL 2.0 and CC-BY-SA 3.0

MiniMe
Space Squid
Posts: 75
Joined: Wed Jan 30, 2013 2:49 pm

Re: 5937 Feedback

#22 Post by MiniMe »

MiniMe wrote:Here is an example of a dead colony, presumably colonized due to the growth special (system Nusakan, AI #2).
Granted, i do not know if this was their 1st/2nd colony, but it is relatively early in the game and that AI only has 3 colonies set up in total (including that dead colony).

Attached are the AI_logs.
Same in 5983 :?
I just found an AI that put its first colony on a growth special. It died out and is now an outpost.
Cannot attach the AI_logs, they are too large...but in the quoted post there are logs from that game.

AndrewW
Juggernaut
Posts: 791
Joined: Mon Feb 04, 2013 10:15 pm

Re: 5937 Feedback

#23 Post by AndrewW »

MiniMe wrote:Same in 5983 :?
I just found an AI that put its first colony on a growth special. It died out and is now an outpost.
Cannot attach the AI_logs, they are too large...but in the quoted post there are logs from that game.
Usually the logs will be small enough if you compress them first. Alternately if you have another place you could put the file up and just provide a link to it.

User avatar
Dilvish
AI Lead and Programmer Emeritus
Posts: 4768
Joined: Sat Sep 22, 2012 6:25 pm

Re: 5937 Feedback

#24 Post by Dilvish »

MiniMe wrote:
MiniMe wrote:Here is an example of a dead colony, presumably colonized due to the growth special (system Nusakan, AI #2).
Granted, i do not know if this was their 1st/2nd colony, but it is relatively early in the game and that AI only has 3 colonies set up in total (including that dead colony).
Attached are the AI_logs.
Same in 5983 :?
I just found an AI that put its first colony on a growth special. It died out and is now an outpost.
Cannot attach the AI_logs, they are too large...but in the quoted post there are logs from that game.
I think I was travelling when you posted those logs before, I had tried keeping up on the forums but had some trouble, sorry I missed that. Since there are only a couple turns in those logs, I can't get the full picture, but my observations don't line up with yours. Perhaps you got logs from the wrong game?. What I find is:

* The logs cover turns 89 and 90, which doesn't seem 'early' in the game to me
* None of the AIs owned any planets in the Nusakan system, but Nusakan 2 is reported by AI 2 as having nonzero defenses and a zero target population, so sounds like an outpost owned by yourself.
* AI 1 had discovered the existence of the Nusakan system, had last observed its contents way back on turn 4, and had never actually visited the system
* AI 2 had last observed the Nusakan system on turn 85
* AI 2 saw value in the system largely because it was a blue or white star and AI 2 did not own any planets in a system with a blue or white star, so high value for a Solar Orbital Generator
* AI 2 knew it would have negative population on Nusakan I and couldn't settle Nusakan III since it was a GG:

Code: Select all

AI_2.log:7089:2013-04-04 11:54:56,033 DEBUG AI : ['Nusakan I : ', 'PRO_SOL_ORB_GEN  400.0', 'GGG  from Nusakan III  400.0', 'maxPop -2.0', 'indVal -19.2', "Non-positive population projection for species 'SP_GEORGE',  so no colonization value"]
AI_2.log:7090:2013-04-04 11:54:56,033 DEBUG AI : ['Nusakan III : ', 'PRO_SOL_ORB_GEN  400.0', "Can't Settle GG"]
* AI 2 was targeting Nusakan II for invasion. I don't know what the invasion analysis was since that info just gets logged while the invasion is being considered, not after troops actually get assigned, which was already the case when these logs start at turn 89

So, from these logs, everything looks fine. If you had earlier captured the planet from AI 2, it's very possible you found it as an outpost simply because the AI wanted it for a solar generator & so established an outpost, not because the AI mistakenly thought it could survive there. Alternatively, perhaps AI 2 *did* have a positive population expectation while the planet was supply-connected to other growth specials it owned, but then you or someone blockaded it and so it died. Or else monsters killed it with biological attacks.

If you want a more thorough analysis, such as of your current situation, I'd need more logs & preferable a saved game. But please be sure you've really thought through it & it appears there is a high likelihood of a problem, because it does take a while to go through all this.

For the saved game please specify not just your version but also your source (prepackaged Windows, MacOS, etc, to help me be able to load the game since we have some incompatibility problems even between builds of the same version but built on different systems). For the logs, for situations like this more than 2 turns will generally be needed. If the logs are too long, even compressed, then you could try attaching them to separate posts, or only posting the ones you believed mattered (here AI's 1 and 2 would have been all that mattered), or posting them on some third party hosting site.
If I provided any code, scripts or other content here, it's released under GPL 2.0 and CC-BY-SA 3.0

MiniMe
Space Squid
Posts: 75
Joined: Wed Jan 30, 2013 2:49 pm

Re: 5937 Feedback

#25 Post by MiniMe »

Hi Dilvish,

thanks for the detailed reply!

I do not know for certain if/that it's a bug.
It just seems too much of a coincident that i was finding dead colonies from AI's that only had settled 1-2 colonies in total. And they had in common, that a growth special was present which matched their race/species.

Concerning the old AI_logs: i cannot remember that game anymore :(

I looked at my current game again (5983) and by looking at the situation it could be possible that the AI didnt have much other option than to colonize this hostile planet.
The other surrounding planets were "poor" for that AI.
I also didnt know that AI has a priority for blue-star systems. I'll keep that in mind when i come across this again.

If i find a better or more certain situation i will post all needed files :)

User avatar
Dilvish
AI Lead and Programmer Emeritus
Posts: 4768
Joined: Sat Sep 22, 2012 6:25 pm

Re: 5937 Feedback

#26 Post by Dilvish »

MiniMe wrote:It just seems too much of a coincident that i was finding dead colonies from AI's that only had settled 1-2 colonies in total. And they had in common, that a growth special was present which matched their race/species.
It used to be that an AI might colonize a planet with a zero or negative population expectation because that would put an Industry Focus on the planet that would last even if the population died and which would give GGG (or asteroid) production bonuses. It could be your feeling of encountering this a lot is primarily left over from that; it didn't change so long ago. That is no longer the case, and the AI will no longer knowingly settle a planet that doesn't have a positive population expectation. That said, the AI currently does its own calcs for that expectation, and it's possible there is an error in my code that I haven't discovered yet. So I am interested in getting documentation on cases that appear to show an AI colony being planted where it shouldn't.
I looked at my current game again (5983) and by looking at the situation it could be possible that the AI didnt have much other option than to colonize this hostile planet. The other surrounding planets were "poor" for that AI.
The AI will not (any more) use up a colony ship where it really just expects it to wind up an outpost, and it won't plunk down a colony on a planet just to use up the colony ship because it had no better options -- it must expect a positive population and for it to generate enough resources to be worthwhile (a tiny planet with expected population of 1 (and no other specials or key star type to influence the decision) would NOT get colonized by the AI, even if it currently had no other options -- it would wait for better.
If I provided any code, scripts or other content here, it's released under GPL 2.0 and CC-BY-SA 3.0

MiniMe
Space Squid
Posts: 75
Joined: Wed Jan 30, 2013 2:49 pm

Re: 5937 Feedback

#27 Post by MiniMe »

Dilvish wrote:The AI will not (any more) use up a colony ship where it really just expects it to wind up an outpost, and it won't plunk down a colony on a planet just to use up the colony ship because it had no better options -- it must expect a positive population and for it to generate enough resources to be worthwhile (a tiny planet with expected population of 1 (and no other specials or key star type to influence the decision) would NOT get colonized by the AI, even if it currently had no other options -- it would wait for better.
In this case there might be something wrong afterall.

Attached is the save-game (5983, Windows, Full Installer) and the AI_logs (i hope they show the info, i have continued the game for many rounds in the meantime and their file size got smaller).

The dead colony is in the system Fluttershy. This is the AI's only colony (their capital was Girtab):

Image
Attachments
AI_logs.zip
(217.29 KiB) Downloaded 84 times
bug_specials2.zip
(397.98 KiB) Downloaded 59 times

User avatar
Dilvish
AI Lead and Programmer Emeritus
Posts: 4768
Joined: Sat Sep 22, 2012 6:25 pm

Re: 5937 Feedback

#28 Post by Dilvish »

MiniMe wrote:In this case there might be something wrong afterall. Attached is the save-game (5983, Windows, Full Installer) and the AI_logs (i hope they show the info, i have continued the game for many rounds in the meantime and their file size got smaller). The dead colony is in the system Fluttershy. This is the AI's only colony (their capital was Girtab)
When you load a game from a savedgame file, it kills the old AIs and restarts new ones, and restarts the log files, that's why the file size got smaller. These logs only cover turn 89, at which point everything's all over, the AI just has a couple scouts and that one planet as an outpost, and there's no information about whether it was originally colonized or outposted, and why that was done. You call this a dead colony, but don't say that you ever actual observed it as a dying colony -- did you? Do you have any other info to provide? I explored the savegame a little, but in this case I don't see anything else pertinent I can glean from it.

Geoff sometimes asks for a single turn's worth of logfiles, but that's mostly for debugging crashes and whatnot. In general, if the issue you're thinking needs looking into is an AI action (like colonization), the log files will need to cover a time period of at least several turns leading up to the colonization; you can know whether the logs will cover that based on what turn you last loaded a savegame (if you have an idea of when the planet was colonized/outposted).
If I provided any code, scripts or other content here, it's released under GPL 2.0 and CC-BY-SA 3.0

MiniMe
Space Squid
Posts: 75
Joined: Wed Jan 30, 2013 2:49 pm

Re: 5937 Feedback

#29 Post by MiniMe »

Ah, thanks for explaining the log files. I will keep this in mind for the future.

I did not see the colony dying out. Can AI place an outpost on habitable planets? That would explain then.

User avatar
Geoff the Medio
Programming, Design, Admin
Posts: 13587
Joined: Wed Oct 08, 2003 1:33 am
Location: Munich

Re: 5937 Feedback

#30 Post by Geoff the Medio »

MiniMe wrote:Can AI place an outpost on habitable planets?
If this implies you, the human player, can't, then note that you can as well. Just select an owned outpost ship in the same system in the fleets window first.

Post Reply