Q about how game chooses Blind Research?

[David Byron]

Ok, you *can* get Doctrine: Loyalty without Applied Physics and Ethical Calculus without any green techs at all. What I observed was that the techs that were missing seemed to rotate with every third new tech choice.

The suggestion is that all techs are in one of three groups and each time you choose you choose only from two of the three groups based it seems on nothing more than how many total techs you currently have.

In addition the 'first' tech in the list of choices will always be present no matter what. This one tech is always present, and the others all have a 2/3 chance (which means if correct they will never be absent from the choice of options twice in a row).

If this is true it would be helpful to know what the three groups are, whether the limit applies to blind tech and aliens too, or not.

[zsozso]

Even that much randomness (2/3 chance) seems too hard to accept with my experience of the Threshold of Transcendence (over 30 games always missing Threshold unless the E13 and E14 done).

Perhaphs the rules are more complex and depend on tech level too...
Just number of techs doesn't explain it I'm sure. If you have everything else except the 5 techs I listed for transcendence, then it doesn't help to research any 2-4 of them if you are missing E14, you still don't get the Threshold choice. But if you only research E13 and E14, you will get it and can transcend with the other 3 missing.

Zsozso

[Chowlett]

Ok, you people are misunderstanding my statement:

quote: And if there's no obvious formula to what's missing, just some arbitrary law then you can't work out what will be missing unless you've been there before. In that case, the choice is effectively random.

I am not saying it's random. I accept that in the light of other evidence the techs missing at any particular point are rigindly defined. Fine. But you can't tell how they are defined until they get there. Let me give you an example of a simple system which follows this. It's called Langton's Ant - if you're feeling bored you can even program it into your favourite language. It works like this.

Imagine an ant on an infinite grid of squares - initially all painted white. At each instant of time, it moves one square, and changes the colour of the square it lands on, then turns according to the original colour - right for white, left for black. Simple. You know that the behaviour is rigidly defined. You even know the rule. Great. So of course it's obvious that after 10,000 or so steps of chaotic ramblings it eventually produces a straight "road". Sorry, no. You can only tell by running through the first 10,000 steps.
This is my point, and it applies to SMAC. There are situation where, without having been there, you don't know what the rules will produce - even if you know the rules. And in SMAC you don't even have that advantage. So while the tech selection is fixed, at any point it is unpredictable without working through to that point.

Now, any questions?

[David Byron]

That's just not necessarily the case at all, no. If the system is a simple as three groups then you just figure out the three groups, mark them off on your tech chart and keep an eye on which is going to be off limits each time. Simple.

Zsozso: you seem to be suggesting that you can have the prerequisite techs for transcendence but it doesn't show up as an option not just one time, but two times in a row. That would be out of step with my guess. Or did you mean that on the way to getting the prerequisites you'd be forced to get E13 and then when you had the prerequisites you'd be forced to get something else?

My only other guess is that the tech map is wrong and E14 is an actual prerequisite but unlisted.

[Misotu]

Well, I don't want to put a spanner in the works or anything, but I'm replaying the Ultimate scenario to try a different strategy and this time round, blind research has worked a real treat.

I have been choosing mostly explore early on and mostly build later. Set it to discover and conquer only once each, I think. I've just researched my *fourth* build tech in a row - and it even gave me tree farms instead of hybrid forests first.

In tandem with the research, I've been stealing techs pretty successfully. Almost all the techs I stole in the early game were either discover or conquer techs.

In fact, it was behaving so well I started to wonder why. I kept checking the % researched bars and basically, in this game it seems very happy to give me the type of tech I want if my research across the four fields is reasonably balanced. For example, it gave me the fourth build tech in a row when the research stood at something like 31% explore, 30% discover, 30% build and 28% conquer.

I'm not saying this is the whole of it, but in this game it really feels like this is an element.

Then again :-)

- Mis

[Misotu]

PS Having said that, my very first choice was explore and it obligingly gave me a discover tech ...

Now that I really don't understand.

- Mis

[zsozso]

I just did some testing in the scenario editor: setup my pet-example tech state and tried to choose tech. Then I set to have what I've choosen and tried the choices again...

The bottom line is: I was wrong. You can get Treshold of Transcendence without E13 and E14. I still do not see any obvious rules, but it is certain, that if I play around wiith the tech states and return to the same tech set from time to time, I keep getting the exact same choices. So it is definitiely not random, completely deterministic on the set of techs you currently have.

Zsozso

[warpage]

Does the order that you got the previous techs matter, or is it merely the state that you are in at that moment (i.e., just the techs you currently have)?

This is a very interesting topic. I hope to see what you all have found out.

[David Byron]

Further messing around tends to confirm the theory that each tech is in one of three groups. Take the number of techs you have researched and reduce modulo 3 -- this number is the number of the group which will be blocked from your choices by non-blind research that turn. The lowest level green tech is always availible (or if no green then blue, then yellow then red).

Off the top of my head the groups include,

Group #1: Social Psych, the other level one yellow, Ind Aut...

Group #2: Centauri Empathy, Flexibility, Progenitor Psych, Advanced Physics

Group #3: Network Nodes, Planetary (Nodes?)...

Sorry can't remember-- the colour coding didn't seem to last beyond the level 1 techs. Doctrine: Mobility is the first "always availible" level 1 green tech so it doesn't have a group. Alien research works the same way, but blind research *seems* to ignore these groups.

[Misotu]

Hmm. Well, I don't know. I'm finding it hard to follow this discussion (inherent incompetence in mathematical and science subjects) and I don't know what modulo 3 is anyway. But here's something else, for what it's worth.

I just tracked the last 40 techs I received on a blind research game. I was comparing what colour I was given with its rating in terms of the already-researched-in-this-colour percentage.

Here are the results:

I got the colour I asked for in 53% of cases overall (which immediately shows that blind research is *not* completely random, even though it feels that way sometimes!)

When the colour I requested had the *lowest* %-researched rating of the four colours, I was given that colour in 79% of cases. This is higher than you would expect but ...

... *much* more interesting is the fact that I actually got the colour I requested in *** 100% *** of cases in the early to mid-game, provided that it had the lowest %-researched rating. So unless I just got lucky (over and over again) this might suggest that %-researched is a highly significant factor in blind research? The system only went haywire when I was researching the final seven technologies to complete the tree. Up to that point, it was a sure thing. This would suggest that the AI behaves differently as you approach transcendence and tries to delay you for as long as it can.

By comparison, where the colour I requested had the largest score in terms of research already completed, I was given the correct colour in only 25% of cases.

In 58% of cases where I did not get the colour of my choice, the colour I did get was the one with the lowest research-completed percentage. And this again went haywire in the last seven techs. Prior to that, the likelihood was 80%. So again, the AI clearly favoured the colour with the lowest percentage-researched score.

I could do lots more analysis, but it's only 40 techs so it's not really valid. On the other hand, it's 40 techs where I was trying for real goals and following real paths relevant to a real game that was going on. Bear in mind that the %-research figures were severely distorted early-mid game because I was stealing techs, some of which were way ahead of my own research levels. So you can't just attribute the results to straight colour availability.

I didn't track how likely I was to steal a colour based on its %-researched value. But my impression was that I was stealing techs in colours I had deliberately chosen not to research so heavily ...

I only played blind research for my first few games of SMAC, and have always felt intuitively that %-researched-to-date is a big factor in how likely you are to get what you want. And not just because, if you've not researched many green techs there are more green techs available, and it's therefore more likely to come up.

The figures seem to bear this out.

Course, this doesn't tell you *what tech* you'll get ... but it does suggest that if you need several techs of different colours go for the one with the lowest-already researched rating, because you're more likely to get it than the others. And once you have it, your % ratings will have changed and you can make a new, highly-likely choice.

- Mis

[Aredhran]

If you look at the alpha.txt, in the technology section, you'll notice that each tech is weighted in all 4 categories (B,C,D,E).

I'm not sure if and/or how SMAC does this, but it seems reasonable to assume that these factors are weighted in when selecting the new technology.

Maybe one of you guys with more time on your hands than I do, could monitor the progress closely and see if the discovered tech is predictable based on those numbers.

Aredhran

[warpage]

Any new news on this front?

[MichaeltheGreat]

With blind research on, you get the same tech choices as with blind off, but they are not visible - only the color. In other words, if your non-blind choice at a particular state would be Doc: Initiative or Ethical Calculus only, your blind choices would be the same.

When you select color bands for blind tech, you are "preferring" the techs (from what comes available) with the highest attribute under that color, as shown in the Alpha.txt file.

If you pick green (explore) only, you will next get the tech with the highest explore value. If there are two with the same explore value, you'll get the one with the highest total of the non-preferred (in this case, build, conquer and discover) values. This means you will sometimes get the higher level tech, of those available (but hidden)

There are at least two factors involved in what's available to choose from:

There is a randomization factor picked in advance which will selectively prevent you from choosing every tech for which you have the prerequisites. This happens in both blind and non-blind. If you reload from the same game, or scenario, you can repeat the same exact tech sequence. However, other times, you can't, and if you make a different choice than in a previous version of the same game or scenario, you can scramble the selection of available techs, and not be able to duplicate the tech tree from the previous version.

There is a rule that you can't get a tech at a particular level, if ALL the techs three levels below haven't been researched. For example, as UoP, I commonly do SotHB, pick Cent. Ecology as bonus, then research Cent. Empathy, a level 3, even though I haven't researched all the level 1's. That's allowable, but regardless of prerequisites, you can't research any level 4, without completing all level 1's, and so on up each level.

The combinations of these rules restrict you from going in a perfect straight line along the most efficient tech route, by limiting your choices to less than the number of available techs. The only compensation for this is from linking artifacts:

If you link a pod from an artifact, there is a randomization factor that will allow you, a certain percentage of the time, to get a tech for which you only have one prerequisite, as long as the second prerquisite is available to you. For instance, I've frequently gotten Adv. Spaceflight from an artifact link, when I had Spaceflight, but not Organic Superlube, although I had Fusion and SFF, the prerqequisites for Superlube.

For this reason, I try to link artifacts as late as possible, when there are some juicy tech options, rather than early in a game. If you're in a lethal tech race, you might not have that choice, but if you can get away with it, it's the way to go.
[This message has been edited by MichaeltheGreat (edited February 28, 2000).]

[Chowlett]

^bump^

due to recent query.

I notice I was a little cocky in this thread. I now humbly apologise.

[Googlie]

LOL Chowlett, it happens to all of us

G.