Alpha Centauri 2

Sid Meier's Alpha Centauri & Alien Crossfire => Modding => Topic started by: Alpha Centauri Bear on November 03, 2019, 10:05:50 PM

Title: Technology values and chaining
Post by: Alpha Centauri Bear on November 03, 2019, 10:05:50 PM
Question: How important are technology values for mod?

I understand they help chose target technology in blind research when player select some research priorities. This is as far as they go. They are not used for targeted research. And even for blind research selection they are of limited use. They sure help you pick a technology among available ones. However, list of available technologies is not available to you. Some technologies ready to be discover may not be available for pickup due to weird formula obscuring 1/3 of them in a not clear fashion that is not humanly possible to follow deep into the game. Setting research priorities let you more likely to research corresponding value technologies but not guarantees it. It is not even give you a clear path though not yet researched tech tree levels either as any technology prerequisites may have completely different values blocking selection engine from reaching it. All in all it does change research sequence but this change is extremely difficult to visualize and not straightforward. One would need to keep a tech tree on a table along with obscuring formula handy and change priorities every turn to be able to fine direct their research. I guess this is not the way it was designed.

In light of the above do you think it makes sense in fine balancing technologies values at all? Also would it make sense to chain some similar technologies together to build some sort of ancestry for similarly valued groups? Like to make sure that at least one prerequisite is of the same of similar value? Of course, it doesn't solve the problem I outlined above completely but, at least it makes such grouped technologies more reachable. And even in this case, it does not make much sense to me either as most of the same color technologies could actually bring drastically different outcome and specific game value. For example both terraforming and native technologies are green. However, terraforming and native units are completely different game concepts you may want only one of them when you meant "green". So even if you are able to direct blind research more or less effectively, half of what you get will be not what you wanted.
Title: Re: Technology values and chaining
Post by: bvanevery on November 04, 2019, 01:59:09 PM
I spent 18 months balancing my tech weights.  Yes in my opinion it is worth doing, and yes the results are highly predictable from the game designer's standpoint.  Not the player's.

One thing you gotta watch out for, is that putting 0s in something creates a hard barrier to it being researched by that focus.  For instance, when I've had Information Networks as a pure Discover tech.  Only the AI factions with a Discover research focus would figure it out, the rest never got a clue.  My only Discover factions are the University and the Cyborgs.  If they're not in the game, it could get pretty stagnant on tech.

What am I doing lately with that?  Seems I do have D1 Information Networks as a pure Discover tech.  It's intended to be a hard barrier so that the Discover factions, particularly the University, can overcome it easily and get way ahead on research.  Meanwhile I've put Biology Lab in D1 Biogenetics, which is a mixed 3, 4, 0, 3 tech and very easy for most factions to research.  That means everyone can get some research improvement by building it, to get the +2 absolute improvement.  That matters quite a bit at the beginning of the game.

Putting fewer weights on things makes the tech harder to research.  Putting more weights on things make it easier to research.  You can construct entire flows through tech trees this way.  I have a fair number of pure Conquer techs for instance.  You're only likely to get them if you have a pure Conquer only focus!  Otherwise you will tend to discover the techs that have mixed benefits, and go up those parts of the tree.

From the player's point of view, the only way to communicate a sense of the weights besides trial and error, is to have a consistent metaphor for what each category means.  I'm pretty hard assed about Conquer techs.  It's where you're going to get your weapons and armor from, and you're not getting them any other way.  If it's a red tech, it's got a weapon or armor in it.  My Build is about getting minerals and energy benefits.  My Explore has evolved to be mostly about new chasses and mindworm stuff.  It's also about making people happy, i.e. population growth, but recently I conceded that bigger populations always produces more minerals and energy as well.  So lately I've been handling the happiness techs by saying they're primarily Explore with a weight of 4, but secondarily Build with a weight of 3 or 2, depending on what they give.

I generally try to keep research continuity between successive techs.  Going from "something needs Explore" to "something needs Conquer" can become a hard barrier in practice, a sort of cliff in the terrain of the tech tree.  Usually I don't want that, usually I want rolling hills that can be traversed.  However if I disapprove of some tech, and I want to soft hide / soft remove it from the game, I hide it behind 2 or 3 layers of changing research requirements.  The current exemplar of that is Genetic Warfare.  I did implement the prereqs sensibly, it's not willy nilly, but you're going to have to go through a convoluted Discover, Explore, Build, Conquer sequence to get it.  It's overpowered, and really easy to use if you get the Planetary Council to legalize atrocities.  Another way of thinking about it, is these transitional cliffs implement substantial delays.  I can make something that seems like it's earlier in the tech tree, in practice take way longer to get to.
Title: Re: Technology values and chaining
Post by: Alpha Centauri Bear on November 04, 2019, 02:50:58 PM
One thing you gotta watch out for, is that putting 0s in something creates a hard barrier to it being researched by that focus.

Clarification: by that focus solely when it is not selected as part of multi priories. In other words, If sum of technology values in all selected research priorities is zero, that technology will be last in line. It is still can be researched when all other choices are exhausted. I am sure you know this, just clarifying.

I agree with everything you said but it does only strengthen my point that it is obscure for player and almost impossible to project priority to result. Like one picking yellow would result in faction that is military weak, behind in research, behind in terraforming technologies, behind in native units, but quite ahead in mineral/economy production and having a lot of facilities in their bases. Nope. You cannot have them separated that cleanly in the game. For example, you cannot flourish in minerals without terraforming techniques and large number of economy multiplying facilities also affects your labs indirectly, etc.

In other words, these priorities should be redefined to make more sense as actual faction development direction and not just technology color group.
Title: Re: Technology values and chaining
Post by: bvanevery on November 05, 2019, 02:06:48 AM
> It is still can be researched when all other choices are exhausted.

Which in a tech tree with research continuity such as I've implemented, will never happen.  There's always a "bread crumb" more tantalizing than the zero.

There are only 2 ways out of that pickle: 1) it's a Tier 1 tech, so you can pop a Supply Pod to gain it.  This may happen eventually.  You can get very far into the game with it never happening.  2) you trade or steal it from someone else.  In some random opponent games, none of the AIs ever learn it to begin with, so the "barrier tech" is never gained that way.

> but it does only strengthen my point that it is obscure for player and almost impossible to project priority to result.

That is false.  The player can read the README about how the mod was designed.  I've told players how things are categorized.  I've also taken great pains to try to make the lore of the game, fit with the actual tech given by the game.  There have been revisions of my mod where that got shaky, but recently, I feel I've won the battle of "intuitive relevance" as to what awards what.  This is a question of authorship, and I have not had enough feedback from players, to verify my own high opinions of my work / choices / results.  Nobody yet has said to me, "Brandon, seriously, WTF?"  :-)  Give it time, give it time.

Anyways, it's false that you can't explain things to the player.  You have to put a lot of effort into making it consistent though.  It is not easy to reshuffle the whole damn tech tree and have it all actually end up making sense.  You typically end up with a tech or two somewhere that doesn't quite fit the lore.  It's like working some kind of giant mosaic puzzle, trying to find the right piece of tile to put here or there.  Over time, I've become intimately familiar with the tiles available.  I have tried to fit the tiles in similar arrangements, multiple times, and have gone back and forth between styles of fittings.  That's what happens over 18 months of iteration on the problems.  I've never allowed lore issues to block a release; gameplay is king.  But on successive releases, as other concerned happened, I've usually found ways to improve the lore fits.

I've also got high falutin' ideas about how I'd never do the lore <--> tech linkage the way they did in this game.  It is too cumbersome to iterate on.  But that's a subject for another day, with a brand new game.

> Like one picking yellow would result in faction that is military weak,

That's not an abstractly provable assumption.  What is provable, is running my Morganites and my Pirates against my other AI factions, and seeing how they do against them.  Both are Pacifist pure Build factions nowadays.  One on land, one on sea.  How well the AI factions do, depends on how all the other AI factions go about things, the size of the map, etc.  I've been iterating on these issues for a long time.  There is no theory that will govern the results, you have to playtest.  I have my AI lab results, but I am awaiting the feedback of more playtesters as more people (hopefully) try out my mod over time.

> behind in research

Getting too far ahead in research can be a disadvantage, at least for the AI.  You have to actually produce all those units and facilities, and not just keep jacking up the cost of everything.

> Nope. You cannot have them separated that cleanly in the game.

You can / I did.  Until you put your own mod in front of enough people to give you feedback, you are merely theorizing about what can or can't be done.  I have put these separation concepts to the test.

The degree of separation has also changed over time.  For instance, the population and growth aspects of Explore vs. Build, are not nearly as cleanly separated as they used to be.  The rate at which factions also get through early parts of the tree, has also changed over time.  It is governed by how fat or thin the early tree is, like a funnel squirting water.  I've seen improvements or regressions in AI behavior simply by changes in that fatness or thinness.  I say again: you cannot theorize this stuff, because it is a complex system.  You have to playtest.  I'm still not sure if there are key early techs that end up changing stock AI behavior, once they are acquired.  I suspect there are, and I'm not sure how the pacing of these things in my mod has changed over time.  Only that in recent iterations, like say the past 10 releases, they've changed for the better.

> For example, you cannot flourish in minerals without terraforming techniques

That's not much of an example.  A Former is a Tier 1 tech.  You can certainly plant Forests and win things with those minerals.  "Forest and forget" is a known, viable approach to the game.  You can boost minerals with a big mine on a minerals special square, there is no restriction on such output.  My opinion is, you can win the game with Impact Squads if you're determined to do so.

> In other words, these priorities should be redefined to make more sense as actual faction development direction and not just technology color group.

I believe I did that, and I do not see how someone would propose to do differently than what I did.
Title: Re: Technology values and chaining
Post by: dino on December 07, 2019, 10:45:31 AM
I think blind research is stupid feature for a player to use, there is too many absolutely essential techs and too many useless ones and bogging down on useless stuff early can even cut your development speed in half. What is the point of using blind research, if you feel the need to mod it in order to direct it to your liking anyway. The tech tree in this game is not designed to provide different balanced paths, supporting different playstyles and that would be necessary for a blind research to be fun to use by an advanced player.

Finetuning weights can improve AIs performance great deal though and for this reason it's worth playing with. For AIs you want to make it differentiate between factions less and only provide some slight flavour in a form different priorities and early short detours, while making sure that all factions will get essential techs not too far apart from each other.
Title: Re: Technology values and chaining
Post by: bvanevery on December 07, 2019, 02:50:59 PM
there is too many absolutely essential techs and too many useless ones and bogging down on useless stuff early can even cut your development speed in half.

I don't agree with your vehemence about the problem.  I've played games of stock SMAC with blind research for almost 20 years, and I consider turning that off to be cheating.  The system can't be completely broken or I wouldn't have kept playing the game.

But it can be problem, i.e. a PITA.  In my mod I completely redesigned the tech tree, so that it's not a problem.  I also re-weighted the Explore, Discover, Build, Conquer weights according to what I consider to be a rational scheme.

Quote
What is the point of using blind research, if you feel the need to mod it in order to direct it to your liking anyway.

If by "mod" it you mean to change your research focus a lot while playing the game... well I never do that anymore.  Every game I play, is implicitly a test run for the default research focus of the faction I'm playing.  So for instance in my current massive chemical gassing During Action Report, I'm playing the Morganites as a pure Build faction.  This is how I feel out whether the AIs would have a hard or impossible time with their research sequences.  I've iterated through this enough times to be very confident that none of my modded factions are experiencing research barriers I don't want them to have.

Quote
Finetuning weights can improve AIs performance great deal though and for this reason it's worth playing with.

Yes I 100% agree.
Title: Re: Technology values and chaining
Post by: Alpha Centauri Bear on December 07, 2019, 04:06:53 PM
I think blind research is stupid feature for a player to use

Absolutely agree. This is a brilliant idea to simulate not directed research - sort of natural way the science progresses when you channel scientists efforts somewhere but cannot predict what exactly will be discovered.
However, as many other brilliant ideas in SMACX, it was poorly implemented. The blind research priorities engine depends on other features: 1) research values assigned to technologies, 2) actual technology dependency tree graph, and 3) the weird mechanism of quasi-random hiding about 1/3 of already available technologies. Even the slightest variation in these parameters above throw blind research off path pretty chaotically. As a result, selecting research priorities do not give you what you want. Yes, if you select say "yellow" research priority you next technology will be yellow in about 30-40% of the time. Which is somewhat above average 25% but not too much to notice it in game.

What it actually does good is to deny you unmatched technologies. In the example above selecting "yellow" research priority will postpone all non-yellow technologies (those with 0 in yellow value) as long as possible. However, exact technology value distribution is absolutely not visible to user and is not even reflected in datalink. One should browse alphax.txt research tree to understand each technology value combination and build a graph of dependencies to be able to trace research progress and tweak it correspondingly. Which is, of course, not user friendly.

, there is too many absolutely essential techs and too many useless ones and bogging down on useless stuff early can even cut your development speed in half. What is the point of using blind research, if you feel the need to mod it in order to direct it to your liking anyway. The tech tree in this game is not designed to provide different balanced paths, supporting different playstyles and that would be necessary for a blind research to be fun to use by an advanced player.

You are saying my words, man. See comments above about blind research mechanics. I agree that blind research does not support different play styles.
Due to that I didn't bother with fine tuning technology research values and proper tree dependencies in my mod. Instead I tried to distribute features across technologies more or less evenly so you don't have an absolute empty technologies (like Optical Computers) anymore. This at least soothes player frustration a little.

Finetuning weights can improve AIs performance great deal though and for this reason it's worth playing with. For AIs you want to make it differentiate between factions less and only provide some slight flavour in a form different priorities and early short detours, while making sure that all factions will get essential techs not too far apart from each other.

It sounds contradictory to what you and I said above about unpredictability of blind research. True, AI gets some technology set skew based on set priorities. However, the exact skew is 1) very unpredictable and unreliable, 2) very-very-very difficult to fine tune with research weights and tree dependencies, 3) dilutes very easily with technology trading. One probably can play thousand games and discover that certain priority set leads to certain technology set on average. However, slightest change in weights or tree build changes this and one would need to play another thousand games to understand how exactly.

In summary, it's not worth optimizing.
Title: Re: Technology values and chaining
Post by: Alpha Centauri Bear on December 07, 2019, 04:19:16 PM
I say even more. It does not seem like it was designed to support different play styles. It is more of a "spice" of the game.
Indeed, you wouldn't expect certain play style to give you Singularity engine early in the game while others won't have it until the end! Repeating same thing once again - the technology trading equalizes this anyway. Sooner or later AI would trade Singularity Mechanics and now others would have it. Technological advantage in this game is the advantage within currently actively researched layer of technologies. Anything before that is already known by everybody and anything beyond it is not known by anybody.
Title: Re: Technology values and chaining
Post by: bvanevery on December 07, 2019, 09:14:06 PM
However, exact technology value distribution is absolutely not visible to user and is not even reflected in datalink. One should browse alphax.txt research tree to understand each technology value combination and build a graph of dependencies to be able to trace research progress and tweak it correspondingly. Which is, of course, not user friendly.

A coherent expectation can be built into the player as to what will happen, if the weights are chosen to have a predictable continuity, and the player plays many games to get the feel for it, and the player tends to choose static research strategies, so that the same research phenomena are repeated over and over again.  In short, if conditions are like when I test my mod.   :D  YMMV if jumping to all kinds of different research categories because I don't do that.  And I will note: the difference between picking two research categories, and one category, is big.  It significantly reshapes your journey through the tree.

I have no idea if the stock game resembles the stable conditions of my mod.  It's been a long time since I've played the stock game in earnest.  So long, that I don't have a muscle memory for it anymore.

One thing I've noticed, is I highly suspect that cashing an Artifact for tech, typically gives you a tech outside of your usual research focus.  That could be deliberate, or it could be it simply doesn't use weights to decide what's learned, so you have pretty good odds of hitting something you're not looking for.
Title: Re: Technology values and chaining
Post by: bvanevery on December 07, 2019, 09:28:24 PM
Instead I tried to distribute features across technologies more or less evenly

My personal aesthetics are, that's exactly the kind of design mistake in the original game that I've ruthlessly suppressed.  I don't believe in "doling out useful items" to everybody.  I regimented everything in terms of choice.  I had to do a lot of work to make it actually work in the real world.  It works AFAICT.  Haven't yet received playtester complaints to the contrary.

Quote
so you don't have an absolute empty technologies (like Optical Computers) anymore.

That's not a compelling reason to "dole out".  There's enough stuff in the tech tree to fill the holes, it's not basically impoverished.  For instance, Industrial Base doesn't have to give both Synthmetal Armor and Recycling Tanks, it can just give 1 benefit instead of 2.  Make that kind of decision enough times, and you have the surplus needed to fill up Optical Computers with something.

I figure the emptiness of Optical Computers is some accident of the original game's development history.  It's clearly not some ideal planned thing.  They probably had something in there, found it wasn't working somehow, and ran out of time to polish it up any better.  Could have been an errant tech, could have been a bit of dialogue that wasn't working, could have been a bug.

Considering how much time I've spent jousting at the tech tree, for 18 months, I can totally see why the original developers cut their losses and ran away from such problems, declaring a law of diminishing returns.  It's not like anyone's paid me to perfect this.
Title: Re: Technology values and chaining
Post by: Alpha Centauri Bear on December 07, 2019, 09:31:44 PM
One thing I've noticed, is I highly suspect that cashing an Artifact for tech, typically gives you a tech outside of your usual research focus.  That could be deliberate, or it could be it simply doesn't use weights to decide what's learned, so you have pretty good odds of hitting something you're not looking for.

For both artifact and pod discovered technology you get the cheapest one combined value one.
Title: Re: Technology values and chaining
Post by: bvanevery on December 07, 2019, 10:04:12 PM
2) very-very-very difficult to fine tune with research weights and tree dependencies,

False.  It's actually rather straightforward.  I've fully implemented an example of how to do it in my tree.  The rules I've followed are:

Now, you're welcome to try my mod, declare it capricious :D, and tell me exactly how it's capricious in your play-through.  But I claim my mod is an example of something that works and isn't that difficult to implement.  If you think otherwise, you should at least calibrate and formalize your expectations vs. the observable reality.  Maybe you want lots more control than I do, and think things are irritating where I don't?

Quote
3) dilutes very easily with technology trading.

You stop technology trading by putting Secret Projects in the way.

Quote
In summary, it's not worth optimizing.

That's a personal value judgment, not an absolute.  I thought it was worth optimizing and I did so.  It was piles of work.  That's one of the reasons my mod is under an Attribution license.  I did all this work, and if anyone wants to benefit from it, they can jolly well acknowledge that I did all that work, not them.  Not that they can't do their own pile of work on top of it, but, building on the shoulders of giants and all of that.  I poured blood on the table to make it actually work.  Many many design iterations, which is all that a game designer can do in the face of such problems, IMO.

It clearly wasn't worth it to the original game developers, to push it any farther than they did.  They had to ship the product and make money.  That said, we don't know how much they pushed it just to get to the point they got to.

Wrestling with this problem is definitely not low hanging fruit.
Title: Re: Technology values and chaining
Post by: bvanevery on December 07, 2019, 10:08:05 PM
For both artifact and pod discovered technology you get the cheapest one combined value one.

Pods only yield Tier 1 techs.  I think it's impossible to substantiate any statistical spread in that case.  You pick up techs too fast to have any idea.  Only way you could prove it, is by correctly disassembling the game's code.  Have you done so?

For an Artifact, your claim is possible.  Cheapest weighted techs, i.e. only 1 category weighted "4" in my mod, are actually not easy to discover.  Those are what I use for "hard research barriers".  The easiest techs to discover have the most weights in the most categories.
Title: Re: Technology values and chaining
Post by: Alpha Centauri Bear on December 07, 2019, 10:19:38 PM
2) very-very-very difficult to fine tune with research weights and tree dependencies,

False.  It's actually rather straightforward.  I've fully implemented an example of how to do it in my tree.  The rules I've followed are:

You are doing it for too long, man. I didn't say it is impossible. The fact that you did it doesn't make it easier for general public.
:D

And yes, I express personal opinion here and don't pretend it is absolute. This is just my observation. There are thousands mods out there and only you are focused on encoding certain evolution path into the tech tree. Nobody else bothers. This is a prize to you as a dedicated researcher. I don't want to go this path, though.
Title: Re: Technology values and chaining
Post by: bvanevery on December 07, 2019, 10:23:03 PM
It does not seem like it was designed to support different play styles.

You are conflating different play styles with different combat styles.  They are not the same.  Nobody said you have to play this game by focusing on combat.  If you choose to be a Spartanesque war machine, yes your options are going to be narrowly channeled.  Bigger weapons are better, thicker armor is better.  You do have some combined arms variance.  You do have some different naval, air, hovertank, and mindworm invasion options.  The variety of what you can do to prevail in combat is limited, but it's not zero.

I think it is difficult to establish what is a valid or invalid play style, when the AI is not competent enough to pursue many of them.
Title: Re: Technology values and chaining
Post by: Alpha Centauri Bear on December 07, 2019, 10:28:38 PM
It does not seem like it was designed to support different play styles.

You are conflating different play styles with different combat styles.  They are not the same.  Nobody said you have to play this game by focusing on combat.  If you choose to be a Spartanesque war machine, yes your options are going to be narrowly channeled.  Bigger weapons are better, thicker armor is better.  You do have some combined arms variance.  You do have some different naval, air, hovertank, and mindworm invasion options.  The variety of what you can do to prevail in combat is limited, but it's not zero.

I think it is difficult to establish what is a valid or invalid play style, when the AI is not competent enough to pursue many of them.

I meant exactly play style, not combat style.
Title: Re: Technology values and chaining
Post by: bvanevery on December 07, 2019, 10:31:38 PM
You are doing it for too long, man. I didn't say it is impossible. The fact that you did it doesn't make it easier for general public.
:D

What's stopping you or anyone else for following my rules about weights?  You don't even have to license that, you just have to apply the same sensibilities.  Primary categories are 4s.  Secondary are 3s.  Blockers are 0s.  There should almost never be 1s or 2s.  Rarely, make something a 5.  Don't do what the original game did, with an 8 somewhere.  This isn't hard stuff.  You just have to have faith that the regime actually works.  I've proven that it can work.

If you want to make a tech difficult to get, while still keeping it in the same place in the tree, disrupt the continuity of research category like 3 times in a row.  My textbook on this is Retroviral Engineering.  Genetic plagues are a game buster and I didn't want anyone to have them easily.

Quote
Nobody else bothers.

How many nobodies are really under discussion though?  I mean "nobody" mods SMAC in any huge way, if you get right down to it.  I presume, based on public comments, that a lot of people change a few settings here and there, as they like.
Title: Re: Technology values and chaining
Post by: bvanevery on December 07, 2019, 10:49:29 PM
I meant exactly play style, not combat style.

Why should I agree with that then?  You can colonize endlessly, you can extract every Artifact from Planet, you can grow big cities and vote yourself the Best Thing Ever, you can overrun your neighbors, you can dig Thermal Boreholes, you can pit factions against each other with covert missions.  There's lots of variety for how you want to reach a victory condition.
Title: Re: Technology values and chaining
Post by: Alpha Centauri Bear on December 08, 2019, 12:29:04 AM
Quote
Nobody else bothers.

How many nobodies are really under discussion though?  I mean "nobody" mods SMAC in any huge way, if you get right down to it.  I presume, based on public comments, that a lot of people change a few settings here and there, as they like.

I didn't see other modders bothering with encoding certain evolution path into the tech tree.
Title: Re: Technology values and chaining
Post by: Alpha Centauri Bear on December 08, 2019, 12:33:25 AM
I meant exactly play style, not combat style.

Why should I agree with that then?  You can colonize endlessly, you can extract every Artifact from Planet, you can grow big cities and vote yourself the Best Thing Ever, you can overrun your neighbors, you can dig Thermal Boreholes, you can pit factions against each other with covert missions.  There's lots of variety for how you want to reach a victory condition.

You are taking things out of context and arguing with who said which word. Don't rush it. We have all the time in the world.

The blind research mechanics does not support play styles. I didn't say there are no play styles. There are. However, there is no direct correspondence between play style and the research priority that suits it best. At least not in vanilla.
Title: Re: Technology values and chaining
Post by: bvanevery on December 08, 2019, 01:18:16 AM
In my mod if you want combat, you do Conquer.  If you want money and minerals, you do Build.  If you want mindworms, you do Explore.  If you want population growth, you primarily do Explore, but recently I decided you also secondarily do Build.  That's sort of a prevarication back to the original design of the game, recognizing that working more squares does make you more money and minerals.  If you want faster research, you do Discover.

My main beef is it feels to me like Build and Explore are used for too much stuff.  Feels like there should be a 5th category, but that wouldn't have fit their original marketing agenda.
Title: Re: Technology values and chaining
Post by: dino on December 08, 2019, 11:34:20 PM

Finetuning weights can improve AIs performance great deal though and for this reason it's worth playing with. For AIs you want to make it differentiate between factions less and only provide some slight flavour in a form different priorities and early short detours, while making sure that all factions will get essential techs not too far apart from each other.

It sounds contradictory to what you and I said above about unpredictability of blind research. True, AI gets some technology set skew based on set priorities. However, the exact skew is 1) very unpredictable and unreliable, 2) very-very-very difficult to fine tune with research weights and tree dependencies, 3) dilutes very easily with technology trading. One probably can play thousand games and discover that certain priority set leads to certain technology set on average. However, slightest change in weights or tree build changes this and one would need to play another thousand games to understand how exactly.

In summary, it's not worth optimizing.

You can't make it satysfying replacement for direct research, but you can prevent AIs from shooting themselves in the foot in the early game.

I've found that assuring that all AI factions get Centauri Ecology, Industrial Automation and resource lifting techs relatively early is essential, especially with thinker mod.
Getting at least 4/2 units early is essential for survival while having aggressive neighbour.
And for flavor you can tune it, so aggressive faction go for impact weapon first, economy techs second and the rest other way around.
Beyond described tweaks for early game, I've found further tweaking tech weights to have little impact on AIs and not worth bothering indeed.

The way thinker mod changes the weights is not effective, since many factions skip Centauri Ecology with its tech_ai feature on and also remain defenseless for too long.
I've suggested Inducti0 additive, instead of multiplicative bonus to make it better, but don't remember if he implemented it. Just fine tuned weights in alphax.txt work even better anyway.
Title: Re: Technology values and chaining
Post by: bvanevery on December 09, 2019, 02:20:18 AM
Industrial Automation and resource lifting techs relatively early is essential, especially with thinker mod.

I just delay the hell out of them.  No Thermal Boreholes for you!  A long delay does level the playing field.  You seriously have to watch out for whoever builds the Weather Paradigm in Thinker Mod though, because it's such a game breaker.  I can't delay the WP too much, I already put it as a Tier 3 project, same as all the other Secret Projects.  Half the point of the WP is to give terraforming abilities earlier.

Quote
Getting at least 4/2 units early is essential for survival while having aggressive neighbour.

I kneecapped everyone's offensive capability, in favor of early defense.  I did recently reintroduce Lasers though.
Title: Re: Technology values and chaining
Post by: Alpha Centauri Bear on December 09, 2019, 05:30:35 AM
You can't make it satysfying replacement for direct research, but you can prevent AIs from shooting themselves in the foot in the early game.

There are plenty of direct and predictable methods to balance factions if you like. Giving them starting techs for one.

The way thinker mod changes the weights is not effective, since many factions skip Centauri Ecology with its tech_ai feature on and also remain defenseless for too long.
I've suggested Inducti0 additive, instead of multiplicative bonus to make it better, but don't remember if he implemented it. Just fine tuned weights in alphax.txt work even better anyway.

I'm sorry. I wasn't aware Thinker changes any game mechanics besides AI choices. If it does, what are additive and multiplicative bonuses you are talking about?

Have you tried my mod? It is built on top of Thinker. You get challenging game from AI plus my changes highlighting different game features and strategies. The most noticeable one is combat units: smooth weapon/armor value progression, sane unit pricing, solved reactor problem, distributed timing, etc. See the readme for complete view.
Title: Re: Technology values and chaining
Post by: dino on December 09, 2019, 08:13:16 AM
I haven't tried it yet, but I sure will try it, when I go on another AC playing spree.
At least I've downloaded it and had a read, I like what you are doing.

tech_balance=1 from thinker.ini
This feature multiplies ai weights from alphax.txt of selected few techs by 4.
Problem was, if original weight is 0, it did nothing to direct AI towards this and the other extremely high weights would make some AIs skip some essential techs like Ecology.
I suggested to make it additive +8 bonus instead, he agreed, but you'd have to find it in source, or patch notes, if he actually changed it.

I personally had it turned off and used manually edited in alphax.txt weights instead.
With slightly less heavy weights, so AI is more likely to deviate from it a bit for variety and few more techs, including some techs leading to the essential ones:

Industrial Base,            Indust,  4, 3, 5, 3, None,    None,    000000000
Information Networks,       InfNet,  3, 5, 3, 3, None,    None,    000000000
Applied Physics,            Physic,  5, 3, 2, 3, None,    None,    000000000
Social Psych,               Psych,   2, 3, 5, 4, None,    None,    000000000
Centauri Ecology,           Ecology, 4, 4, 5, 6, None,    None,    100000000
Nonlinear Mathematics,      Chaos,   6, 4, 3, 4, Physic,  InfNet,  000000000
Planetary Networks,         PlaNets, 3, 5, 4, 3, InfNet,  None,    000000000
Industrial Economics,       IndEcon, 3, 3, 5, 4, Indust,  None,    000000100
Industrial Automation,      IndAuto, 4, 4, 6, 5, IndEcon, PlaNets, 000000100
Gene Splicing,              Gene,    3, 4, 6, 5, Biogen,  EthCalc, 000100000
Environmental Economics,    EnvEcon, 3, 3, 6, 5, IndEcon, EcoEng,  000000100
Ecological Engineering,     EcoEng,  4, 4, 6, 5, Ecology, Gene,    000000000

I still consider increasing weights even more, but this is what I played with last time.
Title: Re: Technology values and chaining
Post by: dino on December 09, 2019, 09:23:05 AM
If you use the latest thinker mod, it has a new tech cost formula that base cost on tech level, not on the number of known tech.
Since you've reduced the number of tech levels in your mod, you should either edit the source code of the new formula, or disable this feature in thinker.ini.
Title: Re: Technology values and chaining
Post by: Alpha Centauri Bear on December 09, 2019, 03:08:42 PM
tech_balance=1 from thinker.ini
This feature multiplies ai weights from alphax.txt of selected few techs by 4.
Problem was, if original weight is 0, it did nothing to direct AI towards this and the other extremely high weights would make some AIs skip some essential techs like Ecology.
I suggested to make it additive +8 bonus instead, he agreed, but you'd have to find it in source, or patch notes, if he actually changed it.

Thank you for reminding me Thinker has this feature. I need to reread it. It was long ago.
I should probably align my changes with this.
Title: Re: Technology values and chaining
Post by: Alpha Centauri Bear on December 09, 2019, 03:09:43 PM
If you use the latest thinker mod, it has a new tech cost formula that base cost on tech level, not on the number of known tech.
Since you've reduced the number of tech levels in your mod, you should either edit the source code of the new formula, or disable this feature in thinker.ini.

That is too. I should pay attention to it.
I didn't reduced number of levels and I never planned to. It is still 15 or something. Similar to vanilla.
Title: Re: Technology values and chaining
Post by: dino on December 09, 2019, 06:47:02 PM
You use the latest release version 0.9, while I'm up to date with his develop builds, it was the source of my confusion regarding missing variables in the thinker.ini, the new tech cost formula is not implemented in 0.9

Thank you for reminding me Thinker has this feature. I need to reread it. It was long ago. I should probably align my changes with this.

I think it's better kept disabled, you can achieve the same, or better just by editing weights in alphax.txt.
The only advantage ( or not ) of the thinker bonus, is that it affects AIs, but not blind research for the player.

In the 0.9 version weights of selected techs are multiplied, in the latest develop build there is additive bonus probably, you can always check the code to be sure.
Title: Re: Technology values and chaining
Post by: bvanevery on December 09, 2019, 08:21:27 PM
I think it's better kept disabled, you can achieve the same, or better just by editing weights in alphax.txt.

Yes actually "why won't you respect the weighting system that other modders implement, that is there precisely for this purpose" was the original source of friction between us.  I think that issue abated, not entirely sure.  More recently, we can't possibly agree on 'fair' eco-damage rules for the AI, or ignoring AI faction social engineering choice requirements.  So from my standpoint, it's good to have some other modder doing something with the Thinker Mod tech.  I have a much better chance convincing Tim what The Right Thing To Do is.
Title: Re: Technology values and chaining
Post by: Alpha Centauri Bear on December 09, 2019, 08:38:02 PM
You use the latest release version 0.9, while I'm up to date with his develop builds, it was the source of my confusion regarding missing variables in the thinker.ini, the new tech cost formula is not implemented in 0.9

Thank you for reminding me Thinker has this feature. I need to reread it. It was long ago. I should probably align my changes with this.

I think it's better kept disabled, you can achieve the same, or better just by editing weights in alphax.txt.
The only advantage ( or not ) of the thinker bonus, is that it affects AIs, but not blind research for the player.

In the 0.9 version weights of selected techs are multiplied, in the latest develop build there is additive bonus probably, you can always check the code to be sure.

I agree with you that this is wrong use of exe patching to try to add some weight to certain techs when same can be done directly in alphax.txt. If this is such an incredible technology player blind research should benefit from this too. It would be only fair.
Title: Re: Technology values and chaining
Post by: Alpha Centauri Bear on December 09, 2019, 08:40:11 PM
That is actually creates a dilemma for me. Should I use the latest Thinker development build or release? I'd opt for release as it is easier to track changes. Otherwise, it'll be a mess.
Title: Re: Technology values and chaining
Post by: bvanevery on December 10, 2019, 12:23:49 AM
You are faced with the dilemma of all the "downstream" Open Source developers.  What you decide, depends on you, and what is happening "upstream".  For instance, if the upstream developer is not in the habit of making official stable releases, but rather is in a permanent state of development release for many years at a time, then you have no choice but to use their development release.  Even in that case though, sometimes a developer will mark a particular version in their repository.  If they don't... then pretty much it falls to you, to provide such structure in your own work.  If it helps you.

For it to be "your" mod, you pretty much have to keep an eye on what changes are occurring upstream.  Lest they derail something you personally are trying to accomplish.  This is thankfully something I don't have to deal with, when using the official binary as my target platform.  Whatever its deficiencies, it's stone cold stable!

Some people have asked about things that would be "downstream" of my mod, like making a chart of the techs, or adapting it to Yitzi's patch.  They have been waiting for my own work to stabilize, so that they aren't just getting their own work ruined by my subsequent tech changes.  My mod might actually be at that point of stability now.  The caveat is I haven't heard back from enough playtesters to be sure.  But, I myself don't have anything additional to do to it anymore.  My current DAR is a pretty darned long game, and all the tech has gone the way I expect and want it to.
Templates: 1: Printpage (default).
Sub templates: 4: init, print_above, main, print_below.
Language files: 4: index+Modifications.english (default), TopicRating/.english (default), PortaMx/PortaMx.english (default), OharaYTEmbed.english (default).
Style sheets: 0: .
Files included: 31 - 840KB. (show)
Queries used: 14.

[Show Queries]