My journey of improving the AI
Moderators: Pandora Moderators, Slitherine Core
Re: My journey of improving the AI
I agree. I've watched some Let's Plays.
I've seen players who never moved their population to different tasks and who did not do any morale-management.
With the old 1.5.4 AI this still somehow works out for them since the AI will only use crap-units that are very easy to defend against and eventually falls behind in development since it didn't have morale-management either.
But when I imagine these players playing against my improved AI the same way, I'd say that Very Easy would give them a tough time. (I would actually love watching someone do it)
A problem here is that all people who I got to playing the beta or my enhanced-private-patches (since I cannot personally update on steam) are very experienced 4x-players and play hard and above. So I get no feedback about lower difficulty-levels and don't know if my claim is true.
From the experts perspective Easy is still Easy.
A bigger variety of difficulty-levels would clearly make sense.
5 levels feels too little to adjust. On both sides.
I'd add at least one Level on either side:
Very Easy -> 60% Pop Growth, Morale as in current Very Easy
Easy -> like the current Very Easy
Normal -> like the current Easy
Hard -> like the current Medium
Very Hard -> like the current hard
Extreme -> inbetween the current hard and Very-hard-level (up to 160% popgrowth, up to +12 morale)
Impossible -> like the current very hard
From a programming-perspective this probably is no big issue and I could do that myself. However, this would require new graphics that match the current art-style.
I've only ever had contact with Proxy-Studios programmers but not their artist. Maybe I can get them to do it, maybe not. I'll ask.
It would be a pity to have to just change the existing ones without adding new levels.
I've seen players who never moved their population to different tasks and who did not do any morale-management.
With the old 1.5.4 AI this still somehow works out for them since the AI will only use crap-units that are very easy to defend against and eventually falls behind in development since it didn't have morale-management either.
But when I imagine these players playing against my improved AI the same way, I'd say that Very Easy would give them a tough time. (I would actually love watching someone do it)
A problem here is that all people who I got to playing the beta or my enhanced-private-patches (since I cannot personally update on steam) are very experienced 4x-players and play hard and above. So I get no feedback about lower difficulty-levels and don't know if my claim is true.
From the experts perspective Easy is still Easy.
A bigger variety of difficulty-levels would clearly make sense.
5 levels feels too little to adjust. On both sides.
I'd add at least one Level on either side:
Very Easy -> 60% Pop Growth, Morale as in current Very Easy
Easy -> like the current Very Easy
Normal -> like the current Easy
Hard -> like the current Medium
Very Hard -> like the current hard
Extreme -> inbetween the current hard and Very-hard-level (up to 160% popgrowth, up to +12 morale)
Impossible -> like the current very hard
From a programming-perspective this probably is no big issue and I could do that myself. However, this would require new graphics that match the current art-style.
I've only ever had contact with Proxy-Studios programmers but not their artist. Maybe I can get them to do it, maybe not. I'll ask.
It would be a pity to have to just change the existing ones without adding new levels.
Re: My journey of improving the AI
If you're looking for someone dumb and new to the game, please add me to your list of betas
Re: My journey of improving the AI
Great.
So have you updated to the 1.5.5 beta yet either trough steam or the standalone?
If so, you can updated the update with the files in the archive downloadable from here:
http://www.civforum.de/attachment.php?a ... 1427822884
Simply overwrite the ones in the Binaries-directory of your Pandora-installation. But you have to upgrade to the beta first.
After that just play and report how it went for you here.
So have you updated to the 1.5.5 beta yet either trough steam or the standalone?
If so, you can updated the update with the files in the archive downloadable from here:
http://www.civforum.de/attachment.php?a ... 1427822884
Simply overwrite the ones in the Binaries-directory of your Pandora-installation. But you have to upgrade to the beta first.
After that just play and report how it went for you here.
Re: My journey of improving the AI
I don't have Nashira, which is a deal-breaker atm.
Told you I was dumb!
Told you I was dumb!
Re: My journey of improving the AI
Okay, I installed Nashira 1.5.5 first. The server shortcut got me a dosbox, which appeared to load your most recent files. Is there a way to confirm I have the latest files? I was going to use your "no agents" mod, but will not if you prefer it on for testing.
But if this sounds right, what settings are of best value to your research? Easy, # players, world size, etc.
Then I will give it a go.
But if this sounds right, what settings are of best value to your research? Easy, # players, world size, etc.
Then I will give it a go.
Re: My journey of improving the AI
Was this with the steam or the non-steam version?
Just setting it to Nashira-beta will download files from almost two months ago.
The latest files have to be installed manually.
I keep posting the executables here:
http://www.civforum.de/showthread.php?9 ... ost6625615
Scroll down to find the latest version. It should always be the archive that I updated last. But sometimes it will be more experimental than other times.
You can have the No-Agents mod active to test since the AI of agent-usage has not been improved yet.
Other than that really any setting is fine. The AI is supposed to adapt to it. (e.G.: it shouldn't try to rush it's neighbour on a map so big that it takes them 20 turns to get there)
If you are not experienced, I hope easy will still provide a challenge for you, so feedback on that would be appreciated. If it's still too easy you can up it.
Btw. with the advanced settings you can also mix AIs of different difficulties. I do that a lot to simulate their adaptive behaviour on recognizing that someone is weaker.
Just setting it to Nashira-beta will download files from almost two months ago.
The latest files have to be installed manually.
I keep posting the executables here:
http://www.civforum.de/showthread.php?9 ... ost6625615
Scroll down to find the latest version. It should always be the archive that I updated last. But sometimes it will be more experimental than other times.
You can have the No-Agents mod active to test since the AI of agent-usage has not been improved yet.
Other than that really any setting is fine. The AI is supposed to adapt to it. (e.G.: it shouldn't try to rush it's neighbour on a map so big that it takes them 20 turns to get there)
If you are not experienced, I hope easy will still provide a challenge for you, so feedback on that would be appreciated. If it's still too easy you can up it.
Btw. with the advanced settings you can also mix AIs of different difficulties. I do that a lot to simulate their adaptive behaviour on recognizing that someone is weaker.
Re: My journey of improving the AI
I have a separate Matrix install, and yes, a bunch of core files were updated to 4/4/15.
So far, it's soooo much better than Civ BE.
So far, it's soooo much better than Civ BE.
Re: My journey of improving the AI
As far as I know the steam-executables also work with the standalone of the matrix-installation, so you can use these files anyways.
Re: My journey of improving the AI
Thanks. I'm keeping the Steam version at 1.4.3 for comparison's sake.
More people need to give this game a second look. It's nearly as addictive as Endless Legends and your work is appreciated.
More people need to give this game a second look. It's nearly as addictive as Endless Legends and your work is appreciated.
Re: My journey of improving the AI
Off topic, but odd. I did an Auslogic defrag at 198 gigs on Sunday. On Tuesday, I'm "missing" about 40 gig. Did the Avast thing, but can't figure this one out. If I count the space on my root directory folders, I should be 100 gigs richer.
I did notice an odd writing to disk sound and restarted the computer but didn't notice missing space until Tues. Pandora is safe, though
In the highly unlikely event someone else has this problem, an automatic Win7 and Avast update restored the disk space. Still a mystery to me.
I did notice an odd writing to disk sound and restarted the computer but didn't notice missing space until Tues. Pandora is safe, though
In the highly unlikely event someone else has this problem, an automatic Win7 and Avast update restored the disk space. Still a mystery to me.
Re: My journey of improving the AI
Sorry Xilmi I have no reports yet. I have given the project some publicity around the net. Also, trying to tear away from Brother vs. Brother.
Re: My journey of improving the AI
Just popping in to tell you your work is appreciated. I only played Pandora a short bit after its release, but got distracted by some other "shiney". Then there were so many negative posts on forums I frequent that I never came back. It crept back on to my radar, and from there discovered what you have been up to. Life is so hectic right now that I don't know when I can load this up, but regardless I am deeply moved by your passion. Thank you for sharing with us.
Jonathan
Jonathan
Re: My journey of improving the AI
But this is disturbing this forum, for it makes believe it's a released work, and not a beta.
Making believe this is even not good for this work, for it's removing consideration for the work.
Making believe this is even not good for this work, for it's removing consideration for the work.
Re: My journey of improving the AI
@Eriss:
The game is "ready" as in it is feature-complete.
That doesn't mean that it cannot be improved on. Especially the AI.
And that's what I'm doing.
The game is "ready" as in it is feature-complete.
That doesn't mean that it cannot be improved on. Especially the AI.
And that's what I'm doing.
Re: My journey of improving the AI
Yeah, but not officially released. Devs didn't give your patch to my released version game.Xilmi wrote:@Eriss:
The game is "ready" as in it is feature-complete.
Or, did I miss anything?
And at start your game was not so complete ready.
Or, maybe, sorry, I may not talk about you, but who's making the project of continuing to patch the game.
You see, as he talk like it's an already officially released work, we can no longer in this forum swiftly well see what's really officially happening new or work-in-progress/to-be-officially-released-soon.
Re: My journey of improving the AI
You are right, when you are talking about the 1.6.0 patch. It is not officially released yet because I cannot update steam myself. The devs have to do it.
However, you can download it here: http://www.civforum.de/showthread.php?9 ... ost6683203 and copy the files over the 1.5.5-Beta-Files.
If you are going to wait til I'm completely "ready", you are in for a long wait. I do not intend on stoping to do this anytime soon.
I make patches every few days but you have to download and install them by yourself.
However, you can download it here: http://www.civforum.de/showthread.php?9 ... ost6683203 and copy the files over the 1.5.5-Beta-Files.
If you are going to wait til I'm completely "ready", you are in for a long wait. I do not intend on stoping to do this anytime soon.
I make patches every few days but you have to download and install them by yourself.
Re: My journey of improving the AI
1.6.0 is now available on the steam beta-branch. No need to go to an unofficial site for that anymore.
Re: My journey of improving the AI
Here's some issues I'm still seeing and which I hope I can resolve before the patch becomes official:
AI-Artilleries can not "kite". Usually there's little point in doing so but when fighting stuff like Messari-Motherships that would be really helpful. I'd like to teach them to avoid being directly adjacent to whatever they want to shell and rather step back first (if they have the movement to do so).
AI-Artilleries sometimes get ahead of other units and thus are killed off before the units meant to "tank" for them arrive. I'm not really sure yet about how to go about it as possible solutions might lead to other issues. "Artilleries don't do stuff on their own!" So I need some kind of judgement wether it is a good idea to attack a target alone or to wait for support.
AI shall specialize cities according to their boni coming from territories instead of the combined boni including buildings as this might lead to a loss of overall empire-efficiency.
AI shall coordinate formers better. When in the lategame 15 formers go to a newly captured city all at once, they might notice: "We have potential for fungi, let's grow fungi"... when the fungi are done they realize: "Too much pollution, we need purifiers!" Once the fungi are razed they realize: "Hey, we can build fungi here!". Basically a former within the range of a city should check what other formers in the range of the same city are doing and take that into consideration. Sounds pretty tough to program though.
AI-Artilleries can not "kite". Usually there's little point in doing so but when fighting stuff like Messari-Motherships that would be really helpful. I'd like to teach them to avoid being directly adjacent to whatever they want to shell and rather step back first (if they have the movement to do so).
AI-Artilleries sometimes get ahead of other units and thus are killed off before the units meant to "tank" for them arrive. I'm not really sure yet about how to go about it as possible solutions might lead to other issues. "Artilleries don't do stuff on their own!" So I need some kind of judgement wether it is a good idea to attack a target alone or to wait for support.
AI shall specialize cities according to their boni coming from territories instead of the combined boni including buildings as this might lead to a loss of overall empire-efficiency.
AI shall coordinate formers better. When in the lategame 15 formers go to a newly captured city all at once, they might notice: "We have potential for fungi, let's grow fungi"... when the fungi are done they realize: "Too much pollution, we need purifiers!" Once the fungi are razed they realize: "Hey, we can build fungi here!". Basically a former within the range of a city should check what other formers in the range of the same city are doing and take that into consideration. Sounds pretty tough to program though.
Re: My journey of improving the AI
If you need a quick solution: Consider to limit the amount of fungus that is constructed per city per turn, so the AI has a chance to look at the effects, before deciding to construct more.
Re: My journey of improving the AI
I'll do the specialization-thing first.
It's the easiest and at the same time the one that I guess has the greatest overall benefit.
The most important about it is that food and minerals are preferably harvested in cities without any terrain-modifier-bonus on either science or production so that no potentially more efficient scientists have to be working on fields and in mines while other, less suitable scientists do the science.
Basically something like that: If (cityhasterritorysciencebonus) then farmerpriority*=0.9 and minerpriority*=0.9.
It's the easiest and at the same time the one that I guess has the greatest overall benefit.
The most important about it is that food and minerals are preferably harvested in cities without any terrain-modifier-bonus on either science or production so that no potentially more efficient scientists have to be working on fields and in mines while other, less suitable scientists do the science.
Basically something like that: If (cityhasterritorysciencebonus) then farmerpriority*=0.9 and minerpriority*=0.9.