RHS AI abilities (Full Version)

All Forums >> [Current Games From Matrix.] >> [World War II] >> War In The Pacific - Struggle Against Japan 1941 - 1945 >> Scenario Design



Message

ReDDoN45 -> RHS AI abilities (8/20/2006 9:01:10 PM)

Hi,

can someone give me a hint please, how well the AI handles the job in the RHS-EOS scenario or isnīt it intended for play vs. the AI, at all.

If playable vs. the AI, which side will be more interesting, i.e. which side will be handled better by the AI?

Thank You!



Accipiter -> RE: RHS AI abilities (8/20/2006 9:23:49 PM)

I was told that RHS-EOS can be played against Jap AI. Apparently there is no good AI for Allies. So it is Human vs. Jap AI or Human vs. Human.



Mifune -> RE: RHS AI abilities (8/21/2006 12:00:38 AM)

Human opponents are always better. Of course AI is always good to learn with.



bstarr -> RE: RHS AI abilities (8/21/2006 12:33:39 AM)

Last I heard, he Cid was having a lot of problems adapting RHS to the AI. And if he hasn't moved Palembang back to location slot #511 I guarantee it won't work properly against the AI.

I wouldn't start an AI game against the AI using RHS. CHS is much more thoroughly tested versus the AI - in fact, I did some of the testing myself. On the other hand, if you're wanting to experiment with RHS vs the AI, be sure and give Cid feedback.



el cid again -> RE: RHS AI abilities (8/21/2006 3:57:52 AM)

RHS is running its 108th AI vs AI game. Test 104 is running into 1944.
I can do about a year a day on my new test machine (which runs so fast you cannot read the messages most of the time).

AI is doing rather better with the economy than expected. Some skeptics flat out said the game would crash if we didn't give AI "free supply" at many points - and it is fortunately not true. It has taken time to figure out HOW to get the AI to do what it should do - and it is sometimes way too efficient. But it is basically functional because the foundation concepts were sound - only a failure to take the time to work out the numbers made economics wholly twilight zone.

AI is not doing very well - as predicted - with Allied strategy. But it also is an idiot on the Japanese side. For some reason - if Japan fails to take Hollandia - AI sends convoys there - and builds a big air base -- wholly undetected and unmolested - and unescorted! Right through a sea of Japanese bases! Lots of wierd stuff like that.

Palembang is not changed - and will not change. It was wrongly programmed as non-malarial in the first place - and Baguio City is wronly programmed as malarial (and not even given a place in the location file - much less vast mineral and food resources). You cannot make Baguio City a port - it is land locked - and making it a level 8 airfield is as much nonsense as making it malarial or not a resource hex. I have - however - been messing with Palembang - and it may be I have AI dealing with it better. I did fix the problem of convoys thinking Baguio was Palembang.
Part of this fix was to change the slot to one AI wants anyway. Another part was to increase the airfield/port sum - as suggested - to eight. But hard code issues may still obtain until these are cleaned out (which is tentatively planned). I will continue to test and work on this - possibly the most difficult problem we have.



bstarr -> RE: RHS AI abilities (8/21/2006 5:24:45 AM)


quote:

ORIGINAL: el cid again

RHS is running its 108th AI vs AI game.


And that's going to be your problem. AI vs AI does not give an accurate representation of what happens when a human player is involved. I can't recall the thread, but just the other day one of the designers was lamenting the fact that most of the problem with the AI today is that too much reliance was made on AI vs AI testing.

Unless you're designing an AI vs AI game, you've got to have human input.



el cid again -> RE: RHS AI abilities (8/21/2006 11:32:35 AM)

I don't disagree - but I was told to run AI out as far as possible before subjecting humans to a system sure to break (by a Matrix programmer). So I did. I now am running human games on both sides. And these are producing useful eratta reports. But so far mostly no major problems.
It will take a long time to get human games into 1944. But you wanna do one - let me know.



el cid again -> RE: RHS AI abilities (8/21/2006 11:36:24 AM)


quote:

ORIGINAL: el cid again

I don't disagree - but I was told to run AI out as far as possible before subjecting humans to a system sure to break (by a Matrix programmer). So I did. I now am running human games on both sides. And these are producing useful eratta reports. But so far mostly no major problems.
It will take a long time to get human games into 1944. But you wanna do one - let me know. I was responding to your charge "CHS was much more thoroughly tested AI vs AI." I don't know if it was or not ? But I have run up to four tests at a time for half a year real time - and got over 100 true game runs - not counting test bed runs. It would not shock me to learn this is more than CHS had. It is certainly better than nothing.
Now it seems that you are against this testing - which is it? Is CHS right to use AI vs AI testing? Either it is good for CHS - and RHS - or it isn't.
Granted it isn't going to show some things - it shows one thing human games cannot - what happens after years of play - without taking years to get there.




bstarr -> RE: RHS AI abilities (8/22/2006 1:08:46 AM)

quote:

ORIGINAL: el cid again

I don't disagree - but I was told to run AI out as far as possible before subjecting humans to a system sure to break


You made no mention of future human vs AI tests. The way you were talking I thought you were running these AI vs AI tests instead of running human vs AI.

quote:

I was responding to your charge "CHS was much more thoroughly tested AI vs AI." I don't know if it was or not ?


I said CHS was more thoroughly tested against the AI; meaning Human vs the AI. You said that you're running these AI vs AI tests before subjecting humans to the mod and CHS has been around for a long time, so it's a safe bet.

quote:

Now it seems that you are against this testing


In reply to my comment "I wouldn't start an AI game against the AI using RHS. CHS is much more thoroughly tested versus the AI" You said:

quote:

RHS is running its 108th AI vs AI game. Test 104 is running into 1944.


It should be obvious from the fact that I was answering someone who was interested in playing against the AI that I was referring to a player vs AI game. You responded that your game was sound because you had over 100 AI vs AI tests; you never mentioned AI vs human tests. I never said I didn't aprove of AI vs AI tests. I was simply stating that AI vs AI tests are no substitute for AI vs human tests (especially when the question is, "is the game ready for AI vs human play) AI vs AI tests can pick out the major glitches, but they aren't sound enough to say it's ready for someone to get started on such a long and detailed game.



el cid again -> RE: RHS AI abilities (8/22/2006 6:22:45 AM)

quote:

ORIGINAL: bstarr

quote:

ORIGINAL: el cid again

I don't disagree - but I was told to run AI out as far as possible before subjecting humans to a system sure to break


You made no mention of future human vs AI tests. The way you were talking I thought you were running these AI vs AI tests instead of running human vs AI.



REPLY: Understood. I thought it was known (but possibly now forgotten in burried threads) that RHS is DESIGNED for human play. While we do make things workable for the AI when possible - compromises when required favor PBEM. The only exception is the EOS case - which is somewhat optimized for AI playing Japan. Since AI does a halfway fair job as Japan - and since many players like to be Allies vs AI - EOS is a scenario for those who want a "tougher" Japanese side. It is ALSO a scenario for those who want more "play balance" (this is not a balanced war folks) and those who want Japanes "toys" - all three going pretty much hand in hand. Nevertheless, RHS is a PBEM scenario set, and for that reason we always have had a lot of human testing. There are a number of human guinea pigs who ran manual games - in spite of their awful early state - and report on things that work and things which don't - so we could make the set you see today.



Page: [1]

Valid CSS!




Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI
0.59375