AI enhancement (test games)

[kyrub]

Kylearan,
Moo is made for Dos, in C+. I have all sorts of problems with RE and I am able to ask a lot of questions per day (as some people here already know). - You have been warned. I will PM.

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

I think changing around tech tree ... definitely goes beyond just fixing bugs and improving AI behaviors.

Wise words. I tend to forget my own decision to avoid any change to the game rules, the way game is played by a human, in the patch. So, Cyneheard, let's put our ideas aside for the time being.

I also like the "prefer top-tier" idea - there's value to every tech choice in Orion...
...It pretty much obviates b) (top tier instead of top tech means no one tech dominates its tier/field combination)

Don't forget the AI is unable to get the best of "every tech", it often works with the whole tech branch, so, for instance, when it makes a design of a ship, it randomly degrades from the best possible component to the worst - so that reducing the number of Battle comp techs could result in seeing a lot of ships with BC I (only a blind example, did not check this specifically in the files).
Top tier = no dominating tech, that's true, but it also means that AIs conjecture will largely depend on RNG... it's not exactly good design, is it? But as long as it means a lot more AI variety, I agree to give it a try.

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

Here's a sketch of a simple algorithm, where X is the number of top-tier techs, Y is the total of available projects, Z is the project to be researched

a. RNG --> 85% b. / 15% c.
(top tier or lower?)

b. Z = Y - RNG (X)
(select top tier)

c. Z = Y - X - 1
(choose the tech just bellow the top tier)

All the techs in top tier get the same % of occurence, top techs may be still marginally preferred due to the way the researchable techs are set up at the game start. The c. case, the tech "just bellow" ensures the AI does not research uselless junk from much lower tiers. It also gives a small overall % advantage to top-tier techs, which is not bad. The c. is necessary for the AI to get to the Advanced techs.

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

I don't suppose there's a way (i.e. enough room) for a race to flag a tech as "obsolete" (i.e. T+10 when you have T+30)

I could probably erase the inferior techs from the researchable list of a palyer once he finds a superior version, but that would count for changing the game rules, I guess.

massively devalue them for trade

Have you really traded IT+10 (or any other IT)??
The tech branches allowed to be traded/given even if the receiver has the superior version are: ECM, BC, ARMOR, Auto-repair, Shields, Engines, Beams, misls, Bombs (incl Bio), Pulsars, Projectors, Special shields and Unique techs.

So trading IT is definitely a bug... The above said, the list should be really shorter, I guess it was meant to help the AI design, but the trade-off is high.

[RefSteel]

I could probably erase the inferior techs from the researchable list of a palyer once he finds a superior version, but that would count for changing the game rules, I guess.

Hmmm ... the main effect it would have would be to bring Advanced Tech into play sooner, I think, though I can think of a few other (rare!) circumstances when I would consider researching a truly obsolete tech (BC, Armor, Shields, weapons, and ship specials never become truly obsolete, I think...)

(Good points on the top-tech-tier options, by the way. Could the AI be persuaded to weigh tech choices similarly to the way they weigh the value of techs in trades? Of course, they would have to someday research techs for which they'd never trade at all, in order to reach advanced techs; I'm not sure how feasible this would be.)

Have you really traded IT+10 (or any other IT)??

... Now that you mention it, I don't think I've ever had the option to trade IT+10 to anybody (I was inventing an example). In fact, I recall that in Imperium 16, I lost the ability to even give it away to a race in the course of a single audience, presumably after gifting them better Terra technology. I know I've traded for IT+10 when I had +20 (as recently as Imperium 22) just for a miniscule uptick in production and miniaturization, but that's presumably not the same thing. I was sure the range tech I traded for it was also obsolete for them, but I guess I could have been wrong. (I know I've never received a tech that wasn't itself obsolete for me in exchange for a tech I believed was obsolete for the AI already.)

[kyrub]

(Good points on the top-tech-tier options, by the way. Could the AI be persuaded to weigh tech choices similarly to the way they weigh the value of techs in trades?

The trade_gift_price routine itself is unusable, the logic is different (the AI basically overvalues the Uniques = refuses to trade them; so it would focus maximally on them... bad choice). A new special weigh routine is something I was reffering to as needing a stadium space. So - no.

Speaking of space I managed to make our minibox a 4x bigger one (it is still not very big). I have two alternative plans:
- either try to combine the old top_tech system with the algorithm I have described above, the latter working exclusively in the Weapons field + maybe some special treatment for Psis
- or combine the new algorithm with an unused value x03 from the RESEARCH.LBX that would fonction as a special label for the key techs. The RNG will loop twice (only) if it draws a non-labeled tech.

The question is, which one fits the box... if any.

(I know I've never received a tech that wasn't itself obsolete for me in exchange for a tech I believed was obsolete for the AI already.)

The AI should not trade to gain those obsolete techs. If it does, something is wrong. I'd like to see some confirmation - which is probably hard to have.

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

A question for owners of Moo OSG:
Is there any mention about which tech groups/branches are checked when the individual tech trees are created? There are planetary_shields, missiles, IRC, but something is wrong with the Range tech check... So I'd like to know what the authorities say on the matter.

[sargon0]

A question for owners of Moo OSG:
Is there any mention about which tech groups/branches are checked when the individual tech trees are created? There are planetary_shields, missiles, IRC, but something is wrong with the Range tech check... So I'd like to know what the authorities say on the matter.

The OSG says - "every player's Limited Research List must include:
1. One Planetary Shield technology
2. One Robotic Factory Control technology
3. One missile (ie. either a missile, rocket or torpedo) technology
4. Either Range 4 or Range 5 fuel cells
If each of these does not appear on a player's Limited Research List, the computer will simply keep rerolling it until all the conditions are met. This way, no player will be without at least one of these critical defense, exploration or industrial technologies"

However except for the last condition (which happens because they are the only two options available in their segment) I believe players have reported instances of all of the other supposed banned situations ie. missing all Planetary Shields, all RCn or all missiles. I never knew if this was a plan never implemented or one coded that does not work. Sounds like you are saying the latter.

[kyrub]

(triple post)

[kyrub]

I thought the Range condition was a bug because - you've said it already.

4. Either Range 4 or Range 5 fuel cells
(which happens because they are the only two options available in their segment)
....
I believe players have reported instances of all of the other supposed banned situations ie. missing all Planetary Shields, all RCn or all missiles.

Magnifique. Thanks for reminding of those complaints! I've read them but totally forgot it. There's a bug, but only in the planetary_shield check (as far as I can tell; there is another one in the missiles check, the Stingers are omitted (while torpedoes are not on the list at all), it is rather negligible).

As for the shields, they mistook the Force fields for the Planetology, so the routine checks for the presence of at least one from:
- controlled inferno landings (lvl 12)
- atmosferric terraforming (lvl 22)
- terraforming + 60 (lvl 32)
- advanced cloning (lvl 42)

It looks like a minor thing but I bet that it makes a solid difference in the large games with max players, because the whole lot is rerolled for any mistake in any player's tech tree. We probably witness a generally increased presence of those Planetology techs in all trees. On the other hand, we see the games where one of the players totally misses on planetary shields tech group.

It is definitely worth repairing.

[kyrub]

(triple post)

[RefSteel]

The trade_gift_price routine itself is unusable, the logic is different (the AI basically overvalues the Uniques = refuses to trade them; so it would focus maximally on them... bad choice).

Is it a good thing that the AI overvalues "uniques" though? If the insanely high value they attached to such things were reduced to something approximating the average real (relative) value of such uniques, would that a) be a good thing and b) make the trade_gift_price routine usable for this purpose? On the other hand, both of the proposals you list (if they'll fit!) sound very clever and intriguing....

The AI should not trade to gain those obsolete techs. If it does, something is wrong. I'd like to see some confirmation - which is probably hard to have.

I think I just confirmed (that I seem to have seen) it actually: In Imperium 22, I got contact with the Sakkras in 2338 when I colonized Kronos, 5 parsecs from Sssla (their nearest world to any of my planets). I had only R4 at the time, so I assumed the 'zards had R5.



Then, in 2353, I traded them R4 (when I assumed they already had R5) for IT+10 (I had already had IT+20 for a number of years; this was a pure miniaturization/production/score trade, that added little even for those purposes).



In 2364, I got my first spy report, and found the Sakkra with R4 and R5 but no other propulsion technology. My next Sakkra spy report of which I still have a copy after that was from 2382, when they had both IS and R6.

So, either: a) It took the Sakkra 54+ years to research R5 and contact range is bugged somehow (unlikely, I think) or b) they traded (obsolete junk) for obsolete R4 when they already had R5 (in this case, it just had to take them ~30 years to research e.g. Inertial Stabilizer).

ALSO: Great work spotting the planetary shields bug! (Looks like a forum hiccup caused a triple post though....)

[sargon0]

So, either: a) It took the Sakkra 54+ years to research R5 and contact range is bugged somehow (unlikely, I think) or b) they traded (obsolete junk) for obsolete R4 when they already had R5 (in this case, it just had to take them ~30 years to research e.g. Inertial Stabilizer).

Well my Sakkra still hadn't completed R5 after 70+ years but I'm not boasting! Since the first tech in a field is more expensive the 1250 cost for R5 (with both R4 then R5 is faster for AI) can take quite a time for a small AI on hard (Sakkra 3PE without range) if they do not focus on it. Mine had only reached 1969 or 28% although the number of turns with failed hits must have made this incredible bad luck.

You are right about your Sakkra having R5 though since, apart from greeting, they need range for Whynil which is 4 from Sssla. I am sure I have seen this as well where an AI exchanges for an obsolete tech (although do not have any examples to hand).

Kyrub: Let me add my congrats on finding the planetary shield requirement bug. You keep on helping explain why the game does not follow documented or expected paths.

[kyrub]

You are right about your Sakkra having R5 though since, apart from greeting, they need range for Whynil which is 4 from Sssla. I am sure I have seen this as well where an AI exchanges for an obsolete tech

I will look at it closely again.

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

The easiness of 'ask for DOW' diplomatic option is so obvious, it has become a part of the RBO exploit list. AI players change their long lasting alliances without turning an eyelash and they ally to smite their yesterday's best AI friend.

There's a simple bug, the routine counts a special penalty to succes based on current treaty (NAP or Alliance) between AI_asked_to_DOW and AI_target_of_DOW, but it forgets to substract the result from the diplomatic core reaction (it still substracts their current relations score, though).

The penalty is not negligible - the bonus of the Human race more or less compensates the AIs alliance penalty. Consequently, we were always having the humans' diplomatic advantage when we were trying to plot a war alliance against an enemy.