AU mod: How to help the AI with research choices

[Rellin]

Thanks a ton, I figured out the dummy resource thing, had to make 22 different ones and just got done and read your post. I had already made the naval transport change to the Curragh.

Any reason that the commercial trait was neglected being assigned a flavour?

What about a Civ like France that is Commercial and Industrial? They just get Generic?

What changes did you make to Ironclad tech, it isn't listed on your spreadsheet that I can see.

Gunna go start a standard size map, emporer level game with 7 random civs, pangea with 60% water, will let you know how it turns out late late tonight or tomarrow.

I look forward to anything you do in regards to this with AI build preferences as well.

[Theseus]

Dear god, when is Firaxis going to hire alexman?

:applause:

[alexman]

Thanks Theseus.

Originally posted by Rellin
Any reason that the commercial trait was neglected being assigned a flavour?

Yes, it's because the Commercial trait doesn't get any cheap buildings, so there is no real Commercial-specific research strategy.

What about a Civ like France that is Commercial and Industrial? They just get Generic?

Yup, Generic.

What changes did you make to Ironclad tech, it isn't listed on your spreadsheet that I can see.

Good catch, thanks! It's Seafaring and allows alliances.

[Rellin]

Gah, they just posted the 1.15 beta patch. If I want the new corruption model to be used, do I have to make an entire new biq file, or can I just open the new conquests.biq and import all the rules from my already created mod? It has been awhile since I played Civ3 before the past few weeks, and I forget how patches affect mods. Will be such a pain to go through and remake all the changes, especially after just getting done with flavors a few hours ago.

[alexman]

It depends on whether they changed the conquests.biq file in the 1.15 patch. I don't think they did.

Anyway, here is the scenario file (created under 1.15 just in case) that incorporates the AI build preferences listed above. Please try it out. It includes the Naval transport flag for the Curragh, optional Philosophy, the removal of Electronics from Motorized Transportation, and the move of Longevity to Sanitation, as all of these things affect AI research.

[Rellin]

Well, I had already redone it last night after I posted and assumed you were asleep, and I didn't want to wait for a response this morning, but I dloaded yers to look over and make sure I did things the same way.

One thing though, I kinda don't agree with a couple of changes you guys have made with techs, and I have read all the threads on them and reasoning behind your changes, but I don't want to make Philosphy optional, I just make it not give a bonus tech, because regardless the free tech is still worth it for the player to go after every single time, and I can't stand that.

The same goes for the Thoery of Evolution, that it is worth it every single time to get that wonder for the two free techs, yes there is a cost in choosing it with the Electronics requirement change to Motorized Transportation, but it doesn't matter, your still going to get the ToE for Hoovers and free Hydro Plants. So, I removed ToE wonder completely (I know huge stock change you guys probably shudder, was also considering making ToE do something like doubling science in the city it was built in) and added Sea Trade and Bridges to Scientific Method. Besides those changes though, everything else in your download my personal mod incorporates.

[Rellin]

*Applause*

The research is excellent. They all research things, it works very well and makes the AI much more fun to play against and more challenging as well.

Now if you can have results this good with improvement building it would be amazing, but probably much more difficult to get working well.

I just wish there was some way to affect the build order right at the start of the game, cause them sitting there for 10 turns waiting for thier city to grow to size 3 to pop out one settler is so dumb, plus then thier city is back to size one and they start right away on another settler.

But anyway, your tech design worked great from what I can tell just by playing, real good work. I don't know why this type of stuff isn't built into the game in the first place, especially this being the thirrd expansion. *Boggle*

[swat-spas2]

While we are on the subject, since this present (if somewhat convoluted) proposal appears to be showing progress on changing AI research.. why not start to test/proceed to the building changes? Flavors could undoubtedly be needed there as well. Being able to focus the build-often list more on units instead of the necessary markets and factories and cathedrals should be a nice boost for the AI.

Addressing the earlier concerns on making the AIs predictable, I'd say they are pretty well fixed in their ways in the stock rules based on the charts alexman has posted and my own play experiences. While you'd have some fixed positions on different civs, you would at least be provided some variety on this and in which civs are in the game as well. Say for example your start location puts you with one of the more 'flavourful' opponents. You could perhaps adjust your own research in this situation, with some but not complete certainty of techs that this AI would have for you upon completion. Plus you'd have to know who was around first anyway. The 'fixed' AI research paths by no means appear to be guarantees, with some new elements of very slight 'random' behavior and some more apparent 50/50 (or at least higher %) choices being made. I'd say this is an aid to the AI as it removes the research exploits of the player since you won't always know what you're up against. It may be a question if this is what is required, to make the hole even deeper on higher difficulty levels, but it sounds to this player like a start in the right direction.

I don't think we are going to be able completely reprogram the AI's research paths, but making it actually make semi-intelligent decisions now and then is a nice touch. Good work so far.

[pvzh]

Ok. I do not want to ruin anybody work, but my opinion that flavours itself is ill-minded idea. They break things more than fix.

They create situation for human to exploit AI's. I will give you an extreme example from Age of Discovery. Protestantism is given protestan flavour and protestant to catholic set to 0%. Now England (protestant) research protestantism and do not trade it to other AI (Netherlands is another protestant and does not have enough to offer even adjusted by AI to AI factor). I, Portugal (catholic), by Protestantism for 10 or 11 gold (AI thinks that protestantism is woth 0 to me) and sell it to Nethrlands for 10 or 15 gpt (yes, gold-per-turn).

In AU mod you did not create such drastic differences, but still the fact that certain AI's will pay +40% to certain tech's will create situations were human player can act as middle man in tech trades and get tech and most of his money back. AI does not do this kind of trades.

[alexman]

Don't forget that the human civ also has a flavor. So the AI will sell you improtant (to you) techs for 40% more than they would under stock rules.

The flavor system provides opportunity for making money in a trade, but at a cost. Do I make a profit from selling Education to the Greeks, or do I keep it to my self so they don't get Libraries? Isn't it natural for them to pay more for Education? Wouldn't scientific humans pay more for Education? Well, now they have to.

[swat-spas2]

I wasn't aware this affected the human players as well, but since it appears to make sense, I don't see a huge problem with it. Could this be exploited, yes. But in point of fact, much (not most, I hope) of what human players do to win the game is 'exploit' the AI. For example; the current tech situation as it stands now basically throws all the AIs down the same basic tech paths, irrespective of their str/weaknesses as a civ. This also means you can count on things like a civ researching Communism or Map Making some other useful tech that you don't want to waste time on (for that game at least). Since they aren't necessarily playing to their strengths as a human might, the AI is weakened. If you had to pay extra in trade for certain techs, which would in general enhance your civ, just as the AI does, then this might provide a balance against a high degree of exploitation, since now you'd have to go thru the process of research instead of ordinarily 'reasonably' trading for something useful.

Also, this seems to provide a greater degree of strategy in research. Not only would you have more things to research (since it appears the AI would be spreading things around more), but you could choose (here's the 'exploit' part) to research things you know the AI will pay thru the nose for and doesn't yet have to get more bang for your buck, even if its not inherently useful for you.

Secondly, as it is currently implemented, there are no 0 to 100 ratios (as in the scenario). alexman posted a test chart on this based on trade values varing with flavor value; indeed it showed if something were worth nothing to you, you could get it for almost nothing and then trade it away for a good deal more to a civ you know wants it bad. However, the present flavours as implemented tries keep the value of all techs at cost, but will tend to help encourage the AIs try get things that will directly help them (by making them appear more expensive), thus making these specific techs more expensive to acquire in trade. A successful human player will understand the strengths of his civ and try to adapt them to when it will suit his advantage. You 'might' go for education with a sci civ earlier than with others for example. Getting the AI to do that seems to be what is attempted here. Could this be exploited, perhaps, but it sounds like the AI gets to 'exploit' itself sometimes too (it won't trade those juicy techs to someone who wants them without a +40% payoff either, including you).

I'd say this is an enhanced strategy for research, giving you options instead of guarantees, giving you a potential payoff point in exchange for having to do a few things yourself sometimes. It could perhaps be tweaked further, seeing if other weights and measures can bring the AI into line research-wise without giving a 40% jump in cost.. but since this would make the AIs act somewhat distinctly (and towards their strengths) instead of all the same, I currently support this method.

[ducki]

Does this affect the discussion on Philosophy and the suggested cost-increase?

Also, it seems backwards that Scientific civs will have to pay more(whether in beakers or gold) for a tech that is more "important" to them - I would think a Scientific civ would be more suited to scientific endeavors, and therefore those techs that are flavored scientific would be _easier_ for them to research(edit: and therefore cheaper to buy), not harder. A military tribe should have lower costs for the techs suited to their aptitude and should definitely not pay more for them, but less, as it should be easier for, say, the Mongols, to learn how to ride horses and shoot arrows, than to read books.

Of course, it's possible I've completely misunderstood this whole flavor thing and gotten it all turned around backwards in my head, and if so, my apologies.

[alexman]

Well, as it turns out with flavors, Scientific civs would of course have to spend the same amount of beakers as a non-Scientific civ to get a Scientific tech. They would be more inclined to research that tech, and they would pay more to obtain it through trade.

[ducki]

It's the "pay more to obtain it through trade" thing that strikes me as backwards. Especially if you take a wealthy civ that can afford to pay a Flavor premium and monopoly premium.

I'm not trying to rain on any parade - I just wonder how this will affect the (temporarily) well-to-do civs that get gouged for a flavored tech. It may not be an issue, let's try it and see.

Edit: Human Scientific civs would be extra-suited to this. Would it kill the whole flavors thing to remove flavors from the first rank techs of all Ages except Ancient? Is it even a big enough issue to worry about?

[pvzh]

If I understand correctly how flavours works (averaging several and adjusting the trade cost of tech by that value). Your resulting "flavour" will always be less than. Thus, tech in any trade will cost lesser than in stock C3C, it means that at almost all levels it is cheaper to buy tech and save lots of gold. The only point of research would be to gamble on SGL because your tech lead will not bring much gold either... hmm...

Commercial and industrious is strengthened (no flavour attached besides low, generic high).

Does not look good to me at all.

Edit: re-worded to make it clearer.

[alexman]

No, the trade cost is not always less than 1. It's about 1 at 99% flavor, and about 1.4 at 100% flavor. See my tests above.

[donZappo]

Good job Alexman! Looking over all of the results I'm surprised by how well it works out with only a very small change in the flavor relationship. I guess my previous estimations about how flavors worked were in fact wrong. *sigh*

I like the values that you've assigned to the techs throughout all of the trees and how they actually play out upon testing. It's a real great bonus that desired techs are more expensive for the civ to purchase as well! Thanks for running all of these tests since I have first hand knowledge how long these tests take and know how much work was put into it.

I don't know where to really take the tests from here, though. You seemed to have hit it right on the head and nothing appears to need tweaking. I'll download that flavor mod of yours and see if anything jumps out at me, but I'd really like to see this proposal go to a vote like a lot of the other AU proposals.

-donZappo

[alexman]

Thanks donZappo!

The next AU game will have three versions: Stock, AU mod, and AU mod + flavors. We will vote on the changes in this thread after we have had some in-game experience, just to make sure I have not missed something.

In the meantime, it looks like flavors have no effect on build preferences, so there's not much else we can do with flavors, unless I'm wrong.

[donZappo]

Originally posted by alexman
In the meantime, it looks like flavors have no effect on build preferences, so there's not much else we can do with flavors, unless I'm wrong.

Really? I know that buildings can have flavor preferences so I always assumed that you could affect build orders as well. Does this not work out in practice? My guess is that by using the flavors as they stand right now, 99% flavor relationship isn't enough to nudge building preferences one way or another. Is that right, or do they just not work at all?

-donZappo

[alexman]

I set up a scenario where the AI could build only two improvements, and it built each of them 50% of the time with no flavors.

Then I gave a 100% flavor to one of the buildings, and nothing changed. I reduced the flavor to 50%, and then even 0%, and the AI still built both improvements equally often.

I hope someone else can find a case where flavors do affect builds, but I couldn't.

[donZappo]

Now that's quite strange. Maybe Breakaway originally intended that feature to be in there but at the last second took it out and left that part of the interface in there? Quite the mystery...

[sabrewolf]

i've noticed a small problem with flavours in AU501 (i havn't posted my DAR yet). because the AI have different research priorities, it's now even easier to get a tech lead if you trade well.

it's probably no advantage on pangea maps, but if you're exploring a lot more that the AI (normal on archipelagio maps), chance is, that you're going to dominate techs for at least the ancient age.

i'm interested to see, what the really good players (i consider myself average) think of this possible exploitation. do these rulechanges help the human more than the AI?

i like most, seeing AI writing coming early. before, you could nearly always win a 50-turn writing gambit, now only if you maxxing out your research.

but anyway, alexman, the work you did with the flavours is great and AIs don't compete for the same tech each tech well done!

[alexman]

OK, I think we can start voting for this proposal, as implemented in AU 501.

So, AU mod panel, you have a week to decide!

My vote (surprise): Yes.

[Fosse]

Originally posted by sabrewolf
i've noticed a small problem with flavours in AU501 (i havn't posted my DAR yet). because the AI have different research priorities, it's now even easier to get a tech lead if you trade well.

I've been playing with the research changes for a while now, and have found this to be more of a frustration than anything. It's true that I have more contacts than anyone - and always will be so long as curraghs are in the game and galleys in ocean are possible - but it's also true that anything I research is likely going to be known by at least one cluster of AIs.

I consider myself average, have little taste for MM except in the very early game, and recently promoted myself to Demigod.

I like the changes a great deal because it adds... well... flavor. I hope the panel votes for them.

[lockstep]

YES.

[ZargonX]

[Stuie]

Yes.

[Dominae]

Quite frankly, I've not looked at alexman's research closely enough to vote in good conscience. Please forgive me if I do not vote this time...


Dominae

[alexman]

All this work, and the panel doesn't even care!

[Dominae]

I care enough not to vote...

(Sorry alexman, maybe when I've got more time on my hands. It's not like I'm voting against it until then!)