RESOLVED:AI Research pattern

[unmerged(21168)]

AI Research pattern

I been monitoring how the AI does research and I noticed a few things. The main reaons he is always behind in equipment is that he spends all of his research IC mainly on the 20IC cost research levels and not the items within it. By 1941 the Russians had all of the armor techs that are on the main level but very few to none of the inner stuff. In addition England and Russia had the techs for Improved planes already but never did all the research for basic planes or the improved planes within it. They have some of it going on but not all of it.

 

[unmerged(8622)]

Montok, what was the basis of this report? Discuss this with other players in the GD. If you all come to a concensus, then post a detailed account in here for our review, or better yet, post such a post in the Suggestions forum. Thanks.
-SS

 

[unmerged(21168)]

Shadow I do not understand what you are trying to say here? Since you have come back to these forums you seem to have a very aggressive attitude toward people who are just trying to help out with the game and report things. If I am reading into your text incorrectly please tell me but even your hello I am back post was very aggressive.

Lothos/Montok is the same person and Vulture will tell you I am extremely active in finding issues with the game and have reported alot of issues/bugs before anyone else has. If you need me to provide you more detailed information please ask and I will do my best to give you what you need.

Reason I have two names is one at work and one at home and I never remember my password hehe.

 

[Vulture]

SS, I have to back up the man here... He's very active in bug reporting. And has found many a bug for me.

This seems a particular buggy AI feature to me.

 

[unmerged(8622)]

Well I am not trying to derail your efforts, if Vulture would vouch for you then I change my tone with you. You'll just have to understand that I am weary of people bombarding this forum with stuff. Thought you were one of those types. My apologies.
Im just a no-nonsense type. No offense intended.
Carry on.
-SS

 

[Vulture]

Hmmm so basically, the AI puts a too high priority on the gold levels? Could you check in an handsoff game or so for a few other majors like USSR, UK & Germany?

 

[Lothos]

What I been doing Vulture is playing Germany most of the time and having 3 HoI games running at the same time. I save my main game periodically and then load up the save game on one of the other HoIs and see what the AI was doing. I checked the following countries


England
Japan
USA
Italy
Russia

They were all concentrating mainly on the (gold) 20 IC techs (or main level techs. I did not check smaller nations cause they tend to be very much behind.

Do you want me to check some of the minors as well that I have no contact with (cause I bombard my allies with tech every 7 days hehe). I can check countries like Mexico, South Africa, Australia etc... when I get home tonight from work.

P.S. Shadow no worries man I totally understand.

P.S. Vulture when i get home I will run a hands off game till 1939 September and check again.

 

[Vulture]

Aye...

I'd like to know the following things:

1) Does this behaviour continue when the war starts
2) Does the AI research gold techs in specific tech trees? Or in all trees? In other words, does the AI pay attention to the tech distribution in its AI files?

 

[Nikolai II]

Originally posted by Vulture
Aye...

I'd like to know the following things:

1) Does this behaviour continue when the war starts
2) Does the AI research gold techs in specific tech trees? Or in all trees? In other words, does the AI pay attention to the tech distribution in its AI files?

I'll add agreement - I played Italy until -42 and then swapped to USA for continued challenge and they had almost as many 'gold' techs as they had applications

(I haven't made through testings though, just chipping in )

 

[Vulture]

Ok I'd like more observations... But I start to see the pattern here

 

[unmerged(14683)]

Idea...

Not sure if that would help, but during the work on C.O.R.E. AI R&D pattern, I've noticed several things that might need correct.

Those are:

1) The default (no directives) AI will always develop theories rather than applications.

Solution: This one can be easily fixed by specifying detailed tech paths.

2) I'm also not sure if the problem with prefering gold techs over applications is not connected with serious interconecting of the pre-requisites, like adding doctrines to the pre-req list. Human players are rather good in coordinating research, but AI miserably fails there. The effect is that sometimes the only tech that is researchable is the golden tech, since all the applications need some high level doctrine.

Solution: Better balancing of the R&D ratios (it will only partially solve the problem though).

3) The AI research directives (ie armor = 2.750) are rounded down (so armor = 2.0), so fractions don't work at all.

Solution: Well, it doesn't need any (although it would be nice to see it corrected in 1.06 ), as long as we know about it.

4) Endgoal parameter doesn't work at all.

Correct me if I'm wrong, but I've never seen that work.

Solution: In fact, endgoal parameter is usually not needed (since we want AI to R&D, not to stop it). What is really needed is the list of techs, that AI should never develop (like paratroopers, rockets and other equipment that AI never uses). I know it's not a wishlist, but again - it would be great to see "endgoal" parameter working in any way.

5) When at war, the AI reduces IC spend on R&D (a good thing). Unfortunately, I've noticed dangerous pattern in that - in case of the USSR is practically stops R&D at all, when enemy walks into their core territory (not tested with other countries, but I believe it works for all, seen similar thing with "Seelioned" UK). That can result in sudden stop of Soviet development for 2-3 years (it's especially true in case of games vs. human player).

So far observed R&D/production ratios (after paying CG and suppy upkeep) are:

- in case of peace/agressive war - 50/50,
- in case of enemy on core provinces - 0/100 (or 1/99, but basically something like that),
- in case of total lack of manpower 100/0.

Solution: Besides the best one (making the ratio modable by moving it to the AI file), there can be more stages of R&D ratios, dependant on the VP lost. Generally, enemy in core provinces should trigger 30/70 ration, not the current "panic mode".