Technical details of the research choice

[DavidByron]

Not a great title.....
I think I mentioned this before in 1999... while researching a method of beating Zsozso's fast transcend time of AD2172 I investigated how the program determines what choice of techs to allow you to pick next. This allowed me to Transcend in AD2168 by skipping over 10 techs that aren't needed in SMACX (I think he transcended in SMAC so he had a few less techs for that reason).

This applies to the so-called non-blind research option. In fact you still can't chose any tech you want. Each tech is in one of three groups, lets call them 1, 2 and 3. These groupings have nothing to do with the colours of the techs. The rule is simple. Your Nth tech cannot belong to group (N mod 3). That is, on each turn you have about two thirds of the possible techs that are available in terms of prerequisites, actually available as a choice.

The "first" available tech is also always available to make sure you can't get into a situation where no further research is possible -- which you could just about have done otherwise.

Sometimes in some of the challenges racing to get to a key tech such as Industrial Automation (and here I'm thinking of the one city challenges where crawlers are indespensable if you are playing to transcendand/or on a huge map) it is useful to know how to do this minimally.

For example if the Peacekeepers had non-blind research they could get to IndAut in 5 techs. That is their 6th tech would be IndAut. IndAut is in group 1 so that's possible. The Usurpers on the other hand seem to get five starting techs (is that a patch thing? my rule book claims they only get three) so the minimum would be ten techs to get IndAut. Since 10 mod 3 is 1 this is not possible and Usurpers would have to get IndAut as their 11th tech.

----------------------------------------------------------------------------------

Now I normally play with blind-rearch off because I think being able to choose is an option that makes you think more.... hence a better game... however this morning I was investigating the effects of blind research and it seems to have no basis in common.

The various techs all seem to be able to be researched irrespective of what number of techs are already researched -- no mod 3 rule.

It seems very hard to get to just Industrial Automation with blind research. Typically between 12 and 25 techs are needed instead of the 5 that are listed pre-requisites.

I hypothesise that the blind research model has additional pre-requisites, or in some other way makes sure you cannot get to certain techs too quickly no matter how lucky and no matter what colour/s of tech you choose to research. In particular to get to IndAut it seems I pretty much had to have Advanced Physics, Centauri Empathy, Doctrine: Mobility, Social Psycology and Ethical Calculus. (For these tests I am assuming that using the edit scenario's "gain a tech" option functions the same as gaining a tech in a regular game --- it does with non-blind research)

Remarkably even after starting a race with only IndAut's two prerequisites and chosing "yellow" I could not get IndAut to come up until after numerous lower level techs had been researched.

There also seems to be a distinction between choosing all the colours and chosing none of them. Can't really say what.....

Even given this hypothetical additional pre-requisites, the colour choice doesn't always seem to be honoured (this is what really put me off about blind research -- if the thing worked properly - by picking from all the techs available of the chosen colours -- it would be fun -- as it is the colour choice seems virtually without effect).

---------------------------------------------------------------------------------

Does anyone regularly play with blind research "on"? Does anyone have any data on this?

[Ironwood]

I play with it on. Though non-blind may take more thought to play properly, I prefer to just "take what life throws at me" and use whatever it is to the best effect. I'm not a perfectionist, and don't generally like to play as such. In addition, it's more realistic to me to play blind. Though a leader might be able to predict new technologies to a certain extent in the immediate future, knowing exactly what to research to reach a goal a hundred years later is a bit much.

But I'll look at the "weighting" of the technologies that lead up to Industrial Automation. I think the way blind research works is that it just chooses the technology randomly from among all possible techs of the moment, though some are "weighted" more than others, depending on the research priorities you set.

I don't know exactly how it works, but were I to design the system, this is how it would work: I'd take the "weight numbers" given in the strategy guide, and add together the associated weights of the possible techs, take a number between 1 and the total, and whichever one it fell upon, would be researched.

So, for example, you're set to build, and there are three possible technologies that have build weighted into them. One has a build weight of one, the second of two, and the third of three. Total is six. If the random number from 1 to 6 is a 1, then it's the first (less build oriented). 2 or 3, and it's the second, and 4, 5, or 6, the heavily build oriented. I'm not sure how clear that is, but hopefully it'll demonstrate itself as I develop this.

So, to trace it back:
Xplr = Explore, Disc = Discover, Bld = Build, Conq = Conquer
The number after the code is it's "weight"

Industrial Automation: Xplr 3, Disc 1, Bld 4

Requires:
Industrial Economics: Xplr 2, Build 5
Planetary Networks: Xplr 1, Disc 4, Bld 3

Indust Econ Requires:
Industrial Base: Disc 1, Build 3, Conq 2

Planetary Networks requires:
Information Networks: Xplr 1, Disc 3, Bld 2

So there's your four required technologies. Not all of them are as heavily "build" related as the others (Planetary networks and Info Networks is more "Discover" related). So let's suppose for a moment that you are Morgan, and that my theory regarding how likely one is to be researched rather than another is correct. You start with Industrial Base.

Possible techs with a build emphasis:
Applied Physics: Disc 2, Bld 1, Conq 4
Biogenetics: Xplr 2, Disc 3, Bld 2
Information Networks: Xplr 1, Disc 3, Bld 2
Social Psych: Xplr 2, Disc 1, Build 3
Centauri Ecology: Xplr 3, Disc 1, Bld 2
Industrial Economics: Xplr 2, Bld 5

Obviously, Industrial Econ has the plurality... though the possibility is slight, 1 in 3 possibility. Add in the possibility of getting Information Networks, and there's now a 7 in 15 (slightly less than half) chance of getting the next step toward Industrial Automation. Let's look at what Industrial Econ and Information Networks open up:

Planetary Networks: Xplr 1, Disc 4, Bld 3.

Assuming you got really luck and *only* got Info Nets and Industrial Econ (after getting Industrial Econ, there's only a 1 in 5 chance of getting Info Nets, which ultimately plays out to just less than a 1 in 10 possibility of getting the two pre-reqs in sequence), there is now a 1 in 3 chance of getting the next pre-req. The chances of getting that far without any pitstops (particularly Social Psych) is now just less than 1 in 30. (It gets much worse if another tech with a build weight get's introduced in the mix with a "pitstop" to another tech).

Planetary networks opens up:
Industrial Automation: Xplr 3, Disc 1, Bld 4
Assuming you got this far without taking any other techs (you got *really* lucky) you have an individual chance of 4 in 13 (just less than 1 in 3, once again). multiply it by the previously caluclated odds, you now have a possibility of 1 in 90 of getting here without opening up another arm of the tree! Let's suppose, just before getting Ind Auto, you get Social Psych. That would open up:

Ethical Calculus: Xplr 3, Disc 1, Bld 3

Because this has the same Build rating as Social Psych, your chances don't change. It's still 1 in 3 for this incident. But this is more time spent researching something that doesn't lead to Industrial Econ. There are other numerous possibilities to research before Industrial Economics becomes more than half likely to research, though I don't want to calculate that.

In short, the "Blind Research" option is weighted *against* straight-shotting toward some key technology. You basically meander about, sometimes getting lucky and moving on to a really good one, other times just meandering about the low end of the tree.

One really potent thing to keep in mind is the high level *Conquer* technolgies that are tied to Discover techs... but aren't tied to earlier Conquer techs. I've found that if I make sure to grab High Energy Chemistry early in the game (fairly easy if you leave it on Conquer early on), you can pretty much persue Discover techs the rest of the game, until someone else picks a fight with you. Go straight to a Discover/Conquer combo, and you'll likely end up with Shard weapons before you get things like missiles, etc!

[DavidByron]

I'm wondering if the numbers are just for the AI -- to help it pick (cheat) sensibly? The highest number does indeed indicate the colour (I edited my Centauri Ecology stats and it became red!) but I still get Applied Physics in preference to Centauri Ecology....

[Blake]

Some time ago I recall a thread about blind research, and the finding that the weights matter more than the color...
So for arguments sake we take

Industrial Automation, IndAuto, 0, 1, 4, 3

Ya'll note that IndAuto has weights 'BUILD =4', 'EXPLORE =3', 'DISCOVER =1' , so the optimum combination of research fields is actually 'BUILD + EXPLORE' - that gets IndAuto quicker than pure 'BUILD'.

Take another stubborn tech:
Environmental Economics, EnvEcon, 0, 0, 4, 3
Again, the best combination is actually 'BUILD + EXPLORE'.

And for arguments sake:

Adaptive Economics, AdapEco, 0, 1, 5, 2
This time pure 'BUILD' is the best, so if you have pre-req for both Ada.Eco and Env.Eco, and set priority to 'BUILD' only, most likely you will get Ada.Eco first (and from experience this is the case).

Anyway, this isn't actually my research (for memory it was documented about the time of Ultimate Builder Challenge, and the observation by many participants that Env.Eco and Ind.Auto were REALLY stubborn about coming up in the tech lotto)

[DavidByron]

Editing the alpha.txt doesn't produce results along those lines as yet. I changed the "build" number to 8 for IndBase, IndEcon and IndAut. No difference. Still takes ages to get there. Still getting other yellow and other non-yellow techs in preference to IndAut. Still seemed to have to get those five additional techs I mentioned, and a bunch of others......

I also tried lowering Secrets of Human Brain to 0,0,0,0 to see if that would discourage AI from picking it. It didn't seem to have any noticeable effect either.

[DavidByron]

Tell me if you can get to IndAut (using blind research) without having all of these:

Advanced Physics, Centauri Empathy, Doctrine: Mobility, Social Psycology and Ethical Calculus

[yavoon]

I do know this. If ur Lal and u go social psych,then SOB. I'VE NEVER been able to take centauri ecology as my free tech. maybe this is too simple for how advanced u guys are. but if it helps, thats cool.