Analyzing research rate

[Alpha Centauri Bear]

Hello everyone.
Everybody is probably aware of a skewed research pace in vanilla: slow at the beginning, blazingly fast at the end. Inductio came up with idea to justify research rate to match faction development level or something. Other people like me and dino like to contribute. However, first I'd like to collect some statistics on research rate to make good match and to not break things even worse.

Here is the call for you fellow players and AAR lovers! Please gather some statistics from your saves and post it here. Any contribution will be highly appreciated and lay a foundation of perfectly matched research pace fix! Thank you.

For each begin to end game please select some 3-4-5 save points throughout the game and list following metrics:

For game
1. Map size (either standard or in total number of squares).
2. Any parameters modifying normal research rate: mod alterations if any, tech stagnation, research rate changed in alphax.txt.

For save point
1. What part of the map is occupied by bases? Roughly.
2. Number of technologies already known to you.
3. RESEARCH SE rating.
4. Current cost of researching technology.

Alternatively, you can just drop saves on me.

[bvanevery]

If I noticed this before, I forgot.  Sorry.

Aren't I the only one who writes AARs?  I mean, in the past 2 years I think that statement is 95% true.

Nowadays I've only been doing AARs on Huge maps.  My mod does make those have more continentally land mass.

My tech tree progressions are nothing like vanilla.  Everything is rearranged, and everything is weighted to make the 4 research focus areas diverge.  This is only becoming more true over time.  I may not have much to develop in my mod anymore, but as I playtest over the long haul, I gradually notice little warts in the tech progressions that I squash down.  No sooner did I release version 1.39, when I found some minor tech dependency and weighting errors at the beginning of the game!  Sufficient to justify a 1.40 release in February 2020.

So the bottom line is, the only data you're going to get from sleuthing AARs, is how tech works in SMACX AI Growth mod.

The Discover part of my tech tree is deliberately shunted off from the rest of the tree.  If you want more and more advanced research facilities, you have to concentrate on that.  If you don't, you're going to run into barriers, such as D1 Information Networks being a pure Discover tech.  If you do concentrate on it, it's possible to get research done very fast, because I've made a lot of the advanced facilities available earlier in the game.  If the University is in the game, their tech can be scary.  And surprise surprise, they're quite capable of completing the Hunter-Seeker Algorithm before you do.

Actually I've adopted a "pure research barrier" for all of the techs now.  C1 High Energy Chemistry is pure Conquer.  B1 Doctrine: Loyalty is pure Build.  It gives the Extremist SE choice, which only gives ++SUPPORT --RESEARCH.  E1 Social Psych is pure Explore.  These are all Tier 1 techs though, and players do pop pods when wandering around, so some get discovered that way.  Anything missing, you have to trade or steal.

It's possible to never discover Information Networks, if none of the other factions were Discover focused, and you're a stick in the mud who refuses to change to Discover focus.  I am, because when I play I'm trying to test my default AI settings, to see what it's like when the AI runs the faction.  When nobody makes Network Nodes, the game has a tech stagnation vibe to it.

Biology Labs are easier to obtain as they're Tier 1 and cross-listed in 3 categories.  Since they provide an absolute constant +2 to research, they're a better deal for fledgling factions with little infrastructure.  I made these easy to obtain in order to alleviate some of the stagnation problems.  I actually haven't paid much attention to when the AI builds these though.  I figure human players need the help too though.  I usually build Biology Labs before I build Network Nodes, especially in the vicinity of my capitol.

[Alpha Centauri Bear]

I figured there won't be much feedback on this topic.


So I just went ahead and implemented my formula and now I sort of play test it. It looks good so far. By the turn 100 I acquire about 25-30% of technologies, which is on target. To the middle of the game it slows down mainly because of AI aggressiveness and unwillingness to trade technologies. I didn't get to the end of the game yet. However, with the pace I am picking it seams pretty feasible. Need to validate on different map sizes too.

[Mart]

Hi,


Does the application of formula need some change of exe file?
Or it is some tweak on alpha/x.txt rules?

[Nexii]

I'd also add ECON rating. For much of the game 2 ECON > 2 RESEARCH.

As well pacts/treaties tend to be huge as the game goes on. That's a harder one to describe but a peaceful world can easily double the rate with Global Trade Pact. Anyways, enough games will give an average either way.

Wasn't the vanilla tech formula the same as SMAX? That is it's a mostly linear cost increase for each tech

Also you can get more data points by switching to the AI factions and seeing what their tech per turn is too

[Alpha Centauri Bear]

Does the application of formula need some change of exe file?
Or it is some tweak on alpha/x.txt rules?

This is exe change.

[Alpha Centauri Bear]

I'd also add ECON rating. For much of the game 2 ECON > 2 RESEARCH.

As well pacts/treaties tend to be huge as the game goes on. That's a harder one to describe but a peaceful world can easily double the rate with Global Trade Pact. Anyways, enough games will give an average either way.

Wasn't the vanilla tech formula the same as SMAX? That is it's a mostly linear cost increase for each tech

Also you can get more data points by switching to the AI factions and seeing what their tech per turn is too

Vanilla == SMAX? Or I missed something?

Yes, I thought about it. However, due to intense trading and acquisition, we need like 1000 games to just collect relevant statistics.

[Mart]

I think Vanilla, we would think it is SMAC, without expansion.


SMAX is the expanded game with Alien Crossfire addition.


The formula probably is the same in both game versions, they most likely did not change it when making expansion.

[Alpha Centauri Bear]

I was using term "vanilla" as "original game" as opposite of the mod, for example. So SMAC and SMAX are both vanilla and if someone want's to distinguish them they will name them specifically.
Yes. Research formula is the same. Actually, I believe all game mechanics is the same.

[Mart]

Problem of research rate, I tried to solve using alphax.txt modification. This can work for players, who install the game and then using "out of the box" game version (vanilla), they could play scenario with such fixes.

What I tried to do in MGM mod, is that I slowed research rate to 10% speed, but then I needed to make it faster in early game. It is difficult to accomplish, but I had some success with just few changed settings.

And it worked, so I had late game with sufficient number of turns, and it was not very short and meaningless.

What I wanted to propose for exe modding, would be that simple factor multiplying cost of research even linearly with research progression could work sufficiently. So we would multiply cost for level 1 techs, x1 (no multiplication). For the last level we would multiply by ten: x10.
I do not have the formula here at the moment, but it might be simple to implement. I am not sure though.


And, this x10 could be further tried, if needed to be change to x8 or x12, etc.

[Alpha Centauri Bear]

Problem of research rate, I tried to solve using alphax.txt modification. This can work for players, who install the game and then using "out of the box" game version (vanilla), they could play scenario with such fixes.

What I tried to do in MGM mod, is that I slowed research rate to 10% speed, but then I needed to make it faster in early game. It is difficult to accomplish, but I had some success with just few changed settings.

And it worked, so I had late game with sufficient number of turns, and it was not very short and meaningless.

What I wanted to propose for exe modding, would be that simple factor multiplying cost of research even linearly with research progression could work sufficiently. So we would multiply cost for level 1 techs, x1 (no multiplication). For the last level we would multiply by ten: x10.
I do not have the formula here at the moment, but it might be simple to implement. I am not sure though.


And, this x10 could be further tried, if needed to be change to x8 or x12, etc.

I doubt you can do it with alphax.txt modding. The only parameter there is to make it proportionally faster/slower. Vanilla formula is very convoluted. The modded formula is complex too. I don't want to expose tuning parameters as most people don't need such complex math. Feel free to let me know if research pace need to be adjusted at any stage of the game.

[Nexii]

I think it might be possible to slow down late game tech enough with later advanced terraforming, more expensive satellites, and nerfed down specialists. I just redid my entire tech tree. Will report how it works out for research rate. i look at techs/turn and also the total number of turns for the AI to typically transcend.

[Nexii]

Problem of research rate, I tried to solve using alphax.txt modification. This can work for players, who install the game and then using "out of the box" game version (vanilla), they could play scenario with such fixes.

What I tried to do in MGM mod, is that I slowed research rate to 10% speed, but then I needed to make it faster in early game. It is difficult to accomplish, but I had some success with just few changed settings.

And it worked, so I had late game with sufficient number of turns, and it was not very short and meaningless.

What I wanted to propose for exe modding, would be that simple factor multiplying cost of research even linearly with research progression could work sufficiently. So we would multiply cost for level 1 techs, x1 (no multiplication). For the last level we would multiply by ten: x10.
I do not have the formula here at the moment, but it might be simple to implement. I am not sure though.


And, this x10 could be further tried, if needed to be change to x8 or x12, etc.

Yea when I was playing around with the curve what I did was give +1-2 E to the base square so the early game wouldn't crawl.