FreeOrion

Forums for the FreeOrion project
It is currently Sat Dec 16, 2017 12:40 pm

All times are UTC


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.



Post new topic Reply to topic  [ 41 posts ]  Go to page Previous  1, 2, 3  Next
Author Message
PostPosted: Mon Oct 17, 2016 4:25 am 
Offline
Dyson Forest
User avatar

Joined: Fri Jul 29, 2016 10:12 am
Posts: 237
SkyCore wrote:
jadwin wrote:
You can't win without production and research.

You absolutely can win without having a single planet ever focus on research. Mass drivers and diamond armor are enough to conquer the entire galaxy. Simply getting population and industry levels up will give you all you need to out produce every other empire.
Aggressive expansion (especially early game) allows for geometric growth rather than linear growth.


Yah, I had my doubts too, but I tried it out. By turn 36 I had NAI, Robotic Hull, MD4 and Diamond armor.
The part I doubt is the part about not having any planets set focus on research. I had my Homeworld and a couple colonies set to research focus to get what I did that fast. However, if at that point I did kick everything over to production I could see myself steamrolling the AI's. You better win fast though, because you could find yourself in a hole Technologically.
So, yea, I'd say it's more than possible.


Top
 Profile  
 
PostPosted: Mon Oct 17, 2016 6:10 am 
Offline
Dyson Forest
User avatar

Joined: Fri Jul 29, 2016 10:12 am
Posts: 237
defaultuser wrote:
I've played games where I never set Research focus. Especially if I start with Egassem and only pick up natives like Mu Ursh. If you're using Exobots to hit all the asteroid belts, the Nascent AI points add up enough to take care of what you need. A recent game had towards the end over 1000 Industry and like 85 Research. I play small games so the likelihood of ending up with no good research species is better.


To be clear when you say you've played games where you never set a Research focus you mean on all planets except your Homeworld, right? Because, as Vezzra pointed out, you're looking at only producing about 10 or so RP's with NAI and several colonies if ALL your planets are set to Production focus. And in that case to get Robotic Hull, MD4 and Diamond armor would take you more like 80 turns to get or if playing a poor research race like Egassam more like over a hundred turns.


Top
 Profile  
 
PostPosted: Tue Oct 18, 2016 12:16 am 
Offline
Vacuum Dragon

Joined: Wed Aug 26, 2015 6:15 pm
Posts: 507
No, I mean never. Why would I set Research focus on an Egassem or George homeworld? I'm going to get a bit of research anyway from the cultural archives and lots of industry.

The way to boost research with an industry-forward species is (in my opinion) to use that extra industry to colonize more planets and pick up the Nascent AI flat-rate. And the Adaptive Automation industry flat-rate of course. I used to try setting the homeworld focus to Research with those species, but I wasn't satisfied with the results.

My current game is with George, and I didn't have any Research focus until I started taking over some Cray planets from the neighbors. As there were quite a few asteroid belts in my home territory, I researched straight to exobots (with Mass Driver 4 and robotic hulls of course just in case) while planting outposts. When exobots were ready, I started doing the "exobot shuffle" while I researched Growth until the belts could be colonized, then let all those colonies finish in a couple turns.

As I move towards end-game in this one, even after absorbing an entire Cray empire my Research is only 256 while my Industry is at 1.41k. I'm at turn 264 and still building robotic hull ships, and only up to Death Ray 2. It's been interesting in that the other AIs have been Eaxaw, Eaxaw, and Etty. So some battles were close, but ultimately went my way ;)

Even the captured Eaxaw worlds I've been putting on Industry, even though it's not all that necessary. That's their preferred focus, but they don't have a bonus/deficit either way.

I have recommended to you to play different species. They aren't just humans with funny ears, but real differences in how to play the game. It spices things for me. I played Chato last game, very different experience.


Top
 Profile  
 
PostPosted: Tue Nov 22, 2016 8:48 am 
Offline
Krill Swarm

Joined: Fri Nov 11, 2016 9:16 am
Posts: 14
I think one of the reasons the AI's are so weak compared to me (when I play) is that I pretty much avoid interacting with them and they fight each other. I know diplomacy is not worked out yet, but it would be nice if I could manually force computer alliances. I bet a human vs 6 allied computers would be difficult. Plus, one of my favorite ways to play 4X games is with another person coop against overwhelming AI opponents.

I think that the computers should also should take their opponents technology in to account (from what they see on the enemy ships and from what they see on planets). If the AI sees that you put all of their research in to getting Xenological Hybridization and Exobots at turn 50, the AI should be extremely aggressive and punish you. I don't know how practical that is, but I think that if my computer neighbors weren't so cautious about me when I have no military tech, they would overwhelm me before I could research or build what I need.

Quote:
But I believe to really 'fix' this problem would be to build in a lot of cheating for the AI and I believe a lot of people are philosophically opposed to that.

I think the AI cheating would be a good setting. I just think it should be clear, like have a setting AI Research Bonus in the setup.


Top
 Profile  
 
PostPosted: Wed Nov 23, 2016 1:14 am 
Offline
Vacuum Dragon

Joined: Wed Aug 26, 2015 6:15 pm
Posts: 507
I am not in favor of cheats for the AIs.


Top
 Profile  
 
PostPosted: Wed Nov 23, 2016 3:29 am 
Offline
Space Squid

Joined: Thu Dec 11, 2014 3:22 am
Posts: 54
Ditto.

A non-cheaty AI feels very different from the other kind.

One thing I notice is that the AI does a better job early on; it knows how to scout and then send a troop ship or even a small task force to a vulnerable system. It knows how to colonize nearby systems. It likes to research Mass Drivers 4 early, to put on some pressure. It knows how to produce stuff.

But it does a poor job later on. It knows how to concentrate a fleet but does not realize that it is usually best either to just leave the fleet where it is or to take the whole fleet and smack something, rather than peel off a ship or two and send them to their doom, which has the added bonus of weakening the main fleet. It does not know how to balance research and production past the early game, fielding large fleets of junk that do little damage versus shields. It has little to no sense of history, unable to remember that the system it can no longer see is probably still well-defended, that the fleet it can no longer see is probably just one hop away, unable to build up for a long-term objective.

It does also like to leave a mass of troop ships undefended.

In my most recent game, I found myself as Gysache in the middle of the galaxy, under a lot of pressure for the first 80 turns, after which I increasingly sliced through the surrounding AIs like a hot knife through butter. (I stopped playing around Turn 155 because the game had slowed down and there really was no effective AI opposition, and I'm not sure it's worth all the clicking I will have to do to slog through the rest of the galaxy so that I can finally face the Experimentors.)

I liked the AI pressure a lot, but the tipping point came fast and sudden. Lots of 4x games are like that though, including Moo2 which changes dramatically once Zortium Armor becomes available.

Anyway,

Ken
Anyway,

Ken


Top
 Profile  
 
PostPosted: Wed Nov 23, 2016 8:32 pm 
Offline
Space Squid

Joined: Thu Dec 11, 2014 3:22 am
Posts: 54
Hi,

Even Zortrium Armor isn't quite required for a while.

I tend to prefer research in 4x games, but the other night I went all in with Egassem, building a very cruddy fleet of 15 basic Destroyers (2*MD, yucky armor, large hull), which did a fine job slicing through non-guardian monsters, AIs, etc.

Stopped playing around Turn 70 because the AIs were all far weaker, and even my research was on par (with the sucky AI research, granted, but...) Having conquered the Mu-ursh for even better cheap fleets was just icing.

I do think that it is best to either go extreme research or go extreme production early, rather than compromise. Extreme production helps you go non-linear by increasing the number of colonies planted, but extreme research helps you snag Automation much earlier, and then Autobots plus the growth needed to colonize asteroids, allowing good tech *and* good production. The better approach depends on map and species, I suppose.

(*ponders* I think that natives are cool. But even though the design intent is for playable species to be balanced based on the existence of natives, I think the opposite is true. Nearby early natives utterly blunts the disadvantages of species with powerful advantages, such as Gysache and Egassem. Environmental preferences and penalties mean less when finding another species is easy. Etc. Xenophobia becomes a much worse and crippling disadvantage out of proportion to the other species benefits.)

Anyway,

Ken


Top
 Profile  
 
PostPosted: Wed Nov 23, 2016 11:05 pm 
Offline
Creative Contributor
User avatar

Joined: Fri Jun 28, 2013 11:45 pm
Posts: 3295
ovarwa wrote:
it is best to either go extreme research or go extreme production early, rather than compromise. Extreme production helps you go non-linear by increasing the number of colonies planted, but extreme research helps you snag Automation much earlier, and then Autobots plus the growth needed to colonize asteroids, allowing good tech *and* good production. The better approach depends on map and species, I suppose.

I agree with this, basically, I've played a couple games recently where I speed blitzed with Egassem, and am currently doing similar with George: I normally go the other way and pile onto research early, whereas when I've tried to play 'balanced' I tend to get stuck and "behind" where I like my numbers to be.

We have in the planning stages a new mechanic called Influence, which will probably have a focus setting and be required for various things including colony and fleet upkeep, I'm hoping that we can balance things out a bit more with this new mechanic and make a balanced approached more viable compared to heavy concentration.

I'm also fairly convinced that we need to rethink the flat rate bonuses, especially for things like Adaptive Automation which is almost certainly very overpowered.
Quote:
(*ponders* I think that natives are cool. But even though the design intent is for playable species to be balanced based on the existence of natives, I think the opposite is true. Nearby early natives utterly blunts the disadvantages of species with powerful advantages, such as Gysache and Egassem. Environmental preferences and penalties mean less when finding another species is easy. Etc. Xenophobia becomes a much worse and crippling disadvantage out of proportion to the other species benefits.)

I think I agree here: the intention with colonising natives is that they're meant to be broadly not as good as imperial species but that isn't always the case: while, for example, Tae Ghirus have one lower stat than Gysache they're also telepathic and otherwise just as good, especially in the later game where you're on production. Mu Ursh are definitely overpowered as a native, etc.

But it's not something I/we have had time to put a huge amount of work into, I want to but it always gets put back in favour of other things, we've been tweaking various things but not gone trhough and redone the natives to make them more balanced: most of my work has been to get the playable imperial species closer to ideal.

Hopefully, once Fighters is ready to merge in I/we can also put some time into this but it's always going to be a lower priority: making natives less willing/able to colonise would be a start I suspect, having them is cool, being able to colonise with them is cool, but imperial species should be better. How to go about that is a bigger discussion.

_________________
Mat Bowles

Any code or patches in anything posted here is released under the CC and GPL licences in use for the FO project.


Top
 Profile  
 
PostPosted: Thu Nov 24, 2016 2:56 am 
Offline
Space Squid

Joined: Thu Dec 11, 2014 3:22 am
Posts: 54
Hi,

MatGB wrote:
I'm also fairly convinced that we need to rethink the flat rate bonuses, especially for things like Adaptive Automation which is almost certainly very overpowered.
Quote:

As is, flat rate bonuses are necessary because of the way focus is all or nothing.

Getting rid of the flat rate bonuses is a fine idea, but whether you do that or not, I think that focus needs to be changed from an all or nothing affair. This would probably make life simpler for the AI too.

So instead of what we currently have:

Industry Focus: 75% Industry, 25% Research
Research Focus: 25% Industry, 75% Research
Any Other Focus: 25% Industry, 25% Research, other Focus benefit

This ensures that the AI cannot accidentally starve itself of either research or industry, even when it feels a need for Protective Focus. Industry and Research would be calculated with all the bonuses for choosing that focus regardless of whether that focus was chosen, and then multiplied by 0.75 or 0.25 as appropriate.

This also mitigates some of the lopsided species bonuses while bringing the species penalties into play: That 75% Research or Industry penalty *always* has some effect.
Quote:
I think I agree here: the intention with colonising natives is that they're meant to be broadly not as good as imperial species but that isn't always the case: while, for example, Tae Ghirus have one lower stat than Gysache they're also telepathic and otherwise just as good, especially in the later game where you're on production. Mu Ursh are definitely overpowered as a native, etc.

The Kobukoru cloud things or whatever they are called are also fantastic. 200% Research? Yes please.

But even the non-overpowered natives are powerful. Ordinary pilots? Great if you're Gysache. Ordinary research or industry? Your species penalty is now meaningless. Narrow habitation range? There's a native species for that.

Taming some of the more extreme native bonuses is fine, but I don't see that as the biggest thing.
Quote:

But it's not something I/we have had time to put a huge amount of work into, I want to but it always gets put back in favour of other things, we've been tweaking various things but not gone trhough and redone the natives to make them more balanced: most of my work has been to get the playable imperial species closer to ideal.

Hopefully, once Fighters is ready to merge in I/we can also put some time into this but it's always going to be a lower priority: making natives less willing/able to colonise would be a start I suspect, having them is cool, being able to colonise with them is cool, but imperial species should be better. How to go about that is a bigger discussion.


I think there's a place for modifiers that affect an entire empire, rather than just the starting species.

Or there's a general penalty for all non-starting species in an empire, to research, industry, troops and piloting.

But the biggest thing is that development and improvement is ongoing.

---

BTW, having played more I think that Asteroid and Organic hulls do need adjustment. Unless fighters are utterly unbalancing, organic hulls have a lot less to offer due to a poverty of external slots. The robotic path is very easy, requiring little research and zero special buildings to start. Getting an organic hull requires more investment and isn't as good as the base Robotic Hull. The Asteroid path is even worse. Lots of research, lots of requisites, then a special building, and finally a hull that's less interesting than a Robotic Hull. I do still have to try stealth though.

Anyway,

Ken


Top
 Profile  
 
PostPosted: Thu Dec 01, 2016 12:03 am 
Offline
Dyson Forest
User avatar

Joined: Wed Jul 16, 2014 6:14 pm
Posts: 212
Location: Black Forest
I agree you can win the map without research-focus, you just need to get Nascent ASAP and colonize as many worlds as possible even if they're very low on target-pop. Still your total research output will be very low compared to a mixed slider strategy so you need to carefully evaluate your tech-path, only focusing on the important stuff.

However, that doesn't mean such a stategy may be a good idea for the AI...

IMO, the AIs should mostly go for the biggest bang of the buck, meaning if a world bring bonuses so that research will be above average levels it should pick this one, likewise with production. Such a methodology would be very easy for an AI having a computerized calculating ability at hand. It may not always lead to the best choice locally but it will result in an overall max, and with multiple AI opponents on a map will ultimately mean that the last surviving AI will rise to become a tough challenge for the player.

That's the thing to realize, all AIs together only function is to beat the player, and so if one AI defeats another AI and takes his worlds this goal is nearer than if all AIs act as if in a global stalemate.


Last edited by Kassiopeija on Thu Dec 01, 2016 12:14 am, edited 1 time in total.

Top
 Profile  
 
PostPosted: Thu Dec 01, 2016 12:10 am 
Offline
Dyson Forest
User avatar

Joined: Wed Jul 16, 2014 6:14 pm
Posts: 212
Location: Black Forest
defaultuser wrote:
I am not in favor of cheats for the AIs.


I would call it handicap, not cheats. It's something that the player sets in advance of the map generation in full knowledge of what it is supposed to do.
And it may indeed only act to counter-balance the fact that, currently the human player can use so much more ingame stuff than the AIs... isn't that "cheating" as well?


Top
 Profile  
 
PostPosted: Thu Dec 01, 2016 1:01 am 
Offline
Vacuum Dragon

Joined: Wed Aug 26, 2015 6:15 pm
Posts: 507
Kassiopeija wrote:
And it may indeed only act to counter-balance the fact that, currently the human player can use so much more ingame stuff than the AIs... isn't that "cheating" as well?

There's nothing that I'm aware of that they can't use, so you're describing areas where the AI developers need to improve.

That's something I've mentioned in some of the discussions regarding fundamental changes to gameplay. The AIs are already behind, and any changes will need recoding just to get back to what they already are handling. Human players will adapt quickly to rule changes, where the AIs will need programmers to implement upgrades.


Top
 Profile  
 
PostPosted: Thu Dec 01, 2016 2:17 am 
Offline
Dyson Forest
User avatar

Joined: Wed Jul 16, 2014 6:14 pm
Posts: 212
Location: Black Forest
defaultuser wrote:
There's nothing that I'm aware of that they can't use, so you're describing areas where the AI developers need to improve.

The information I have on this subject is very limited and may also be outdated, but out the top of my head there are some areas:
- Logistics Focus
- Many items involving "knowledge of the map" like Planetary Starlane Drive, Stargates
- Special stuff like Bio-Terminators etc
- I've never seen the AI creating an artifical Black Hole, errecting a Gateway to the Void
- You can actually derive from the gamefiles which technologies (and their order) will be enqueued by the AI - all other remaining techs won't be used and I expect the AI isn't coded to use them even if they get them once they run out of queuable techs (ie. towards Transcendence)
Maybe around 30% of the technology tree that is... mostly fancy stuff but still the point stands. Thing is the trend is actually to add new stuff to the tree which only increases the gap of the player having an advantage over the AI, not reducing it.
Realisticially, you either do the AI coding by yourself or the thing you're asking for is never going to happen.
It may not even be a good thing for the AI in the first place - even in many commercial games AIs normally can't use the game's stuff fully - because this involves increasing the complexity of the code, which (a) requires time (b) harbours more bugs (c) incites the AI to do silly stuff in unusual situations.

edit:
And the greatest disadvantage they have is that they can't play the game in such an intelligent and reactive style like a human can. The ability to look at the map, monitor what's going on, and adjust the research of technologies in accordance to that is much more worth than a handful of flat-out bonuses increases IMO


Top
 Profile  
 
PostPosted: Thu Dec 01, 2016 3:19 am 
Offline
Creative Contributor
User avatar

Joined: Fri Jun 28, 2013 11:45 pm
Posts: 3295
Kassiopeija wrote:
defaultuser wrote:
There's nothing that I'm aware of that they can't use, so you're describing areas where the AI developers need to improve.

The information I have on this subject is very limited and may also be outdated, but out the top of my head there are some areas:
- Logistics Focus
- Many items involving "knowledge of the map" like Planetary Starlane Drive, Stargates
- Special stuff like Bio-Terminators etc
- I've never seen the AI creating an artifical Black Hole, errecting a Gateway to the Void
- You can actually derive from the gamefiles which technologies (and their order) will be enqueued by the AI - all other remaining techs won't be used and I expect the AI isn't coded to use them even if they get them once they run out of queuable techs (ie. towards Transcendence)

To my knowledge, most of this is correct—the AI does use Artificial Black Hole in the late game if it hasn't found a normal black hole, I've captured a half complete one once, the rest I've never seen: doesn't even use terraform or gaia transform yet.

It does/will research everything not given a priority if/when it's run out of other things to research, so in the very late game it starts having hidden planets, but it's not (yet) set to use the esoteric buildings and similar.
Quote:
Maybe around 30% of the technology tree that is... mostly fancy stuff but still the point stands. Thing is the trend is actually to add new stuff to the tree which only increases the gap of the player having an advantage over the AI, not reducing it.
Realisticially, you either do the AI coding by yourself or the thing you're asking for is never going to happen.
It may not even be a good thing for the AI in the first place - even in many commercial games AIs normally can't use the game's stuff fully - because this involves increasing the complexity of the code, which (a) requires time (b) harbours more bugs (c) incites the AI to do silly stuff in unusual situations.

I agree with this mostly, what I think I would like to see is if individual AI empires are set to use 'different' special things so you do sometimes have one using, say, starlane bores (this'll be hard, they don't get topography yet) and a different one using spatial distortion or gateway to the void.

There is some ongoing work (that I am barely able to follow) to give the AIs different traits and/or character but overall, yes, the backend devs and scripters have managed to race ahead of the AI team in the last year or so so there are new features that aren't accounted for and, frankly, we need to give them time to catch up before doing more weird stuff that requires adapting: fortunately some of the work they have done is to set things up so that it's easier for me/non AI coders to adapt the AIs for basic changes, which has been essential in the work I've been doing on the Fighters branch, for example—when I made up a basic tech tree, I could tell the AIs how to use the stuff by changing a single, easy to understand file and not go into the guts of the code. This speeds things up and allows me to make other cost/balance tweaks without waiting on the AI devs to be able to test things.
Quote:
edit:
And the greatest disadvantage they have is that they can't play the game in such an intelligent and reactive style like a human can. The ability to look at the map, monitor what's going on, and adjust the research of technologies in accordance to that is much more worth than a handful of flat-out bonuses increases IMO

Yup, we can plan (and setup distractions and misdirections), the AI can basically only react or set targets. It's getting better at that, but it still sometimes does weird stuff.

(some of this is why, despite wanting to do other things, my main work this cycle is going to be on Fighters and tweaking the Species balance, because the AI is already setup to know different species stuff and doesn't require more work)

_________________
Mat Bowles

Any code or patches in anything posted here is released under the CC and GPL licences in use for the FO project.


Top
 Profile  
 
PostPosted: Thu Dec 01, 2016 5:29 am 
Offline
Vacuum Dragon

Joined: Wed Aug 26, 2015 6:15 pm
Posts: 507
Right, and it's what I've been kind of getting at. There's a pretty substantial gap between even moderately experienced players and the AIs. Add more tech and features just widens the gap because the humans will pick up and incorporate them into game strategy (or decide not to waste time on them) because they can learn.


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 41 posts ]  Go to page Previous  1, 2, 3  Next

All times are UTC


Who is online

Users browsing this forum: Exabot [Bot] and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB® Forum Software © phpBB Group