RE: An AI Player's AE Concerns (Full Version)

All Forums >> [New Releases from Matrix Games] >> War in the Pacific: Admiral's Edition



Message

Mike Scholl -> RE: An AI Player's AE Concerns (12/16/2007 12:33:43 AM)


quote:

ORIGINAL: Kull
Assuming the equation "More time = better AI" is true, then you'd be hard pressed to find an AI player who wouldn't LOVE to have that added as a check-box option. In the same vein, keep in mind that some of us actually can afford to buy (or already own) pretty serious machines, so the added processing time wouldn't be that much of an issue. And making it optional ensures the game doesn't become unplayable for the owners of ancient clunkers.

If it's true that much of the AI amounts to scripting, then I also support the suggestions of those who are begging for periodic script reset points, to give the AI a better chance of adjusting to changes over time (i.e. changes on the map).



"More Time = Better AI" is only true if the AI is written to make use of it. I doubt the current 2by3 AI would be any better if you gave it a year. You are going to have to wait until WITP II and an AI that's actually written to make more intelligent strategic decisions for this to be the case....



Admiral Scott -> RE: An AI Player's AE Concerns (12/16/2007 5:49:08 AM)

I dont care if the AI took ten times longer if it will play better.
Its still much, much faster the PBEM.

Us hardcore wargamers are very patient anyways, you have to be to play a game that can take years to finish!

Having it as an option sounds good, but I would want the best AI no matter the time wait.
I have a fast computer.



Mike Scholl -> RE: An AI Player's AE Concerns (12/16/2007 7:32:08 AM)


quote:

ORIGINAL: Admiral Scott

I dont care if the AI took ten times longer if it will play better.
Its still much, much faster the PBEM.

Us hardcore wargamers are very patient anyways, you have to be to play a game that can take years to finish!

Having it as an option sounds good, but I would want the best AI no matter the time wait.
I have a fast computer.



Do you read English? I said no amount of additional time would help the current AI. It simply isn't "wired" that way. For any real improvement you will have to wait for WITP II and a total re-write of the code. Sorry if I seem insulting..., but your "reply" to my post sounds as if you never read it at all.



berto -> RE: An AI Player's AE Concerns (12/16/2007 2:10:00 PM)

quote:

Do you read English? I said no amount of additional time would help the current AI. It simply isn't "wired" that way. For any real improvement you will have to wait for WITP II and a total re-write of the code. Sorry if I seem insulting..., but your "reply" to my post sounds as if you never read it at all.


Do you speak about the WITP AI as a knowledgeable insider? Better, someone who has seen and actually worked with the low-level AI code (i.e., is a member of the WITP code development team) and is technically qualified to pronounce, "No, this is not possible"?

If so, okay, but are there not at least parts of the current AI that are arbitrarily terminated for time reasons? Is not the current AI modular enough that at least a few small parts of it might be redesigned in this way (without doing a total AI rewrite)?

We're just asking.

Or maybe not asking, just openly wishing for/speculating about AI improvements along these lines? It gives WITP designers, AE or II, some things to think about. What's the harm in that? How does it serve to just abruptly shut this discussion down?



Mike Scholl -> RE: An AI Player's AE Concerns (12/16/2007 2:48:07 PM)

I've spoken to, and ask similar questions of, several programmers who've tried to deal with the WITP code over the last few years. And I've gotten the same answer I'm giving you. There is a very finite limit to what can be done to the AI with the current coding...., too much of it is "hardcoded", and it's "all over the place". Much of what we see as "AI effort" is just "scripting".

I'm sure the "AE Team Programmers" will do whatever they can to make the "AI" function within the new Admiral's Edition. But from the responses I've recieved, there isn't much hope of making it "better" than the current version...., just getting it to reccognize the new map features and units and be able to function with them as well as the AI in the current game will be a challange.

According to the "poll results", apparently at least half of the current players find the AI enough of a challange that they play "only against the AI"..., so they must find it adequate to their needs. For the rest of us, it would seem that PBEM is the answer until some group gets the time, money, and interest to re-write the code from the ground up and build in enough flexibility for a better AI to operate with.



Mike Scholl -> RE: An AI Player's AE Concerns (12/16/2007 2:57:39 PM)


quote:

ORIGINAL: berto
Or maybe not asking, just openly wishing for/speculating about AI improvements along these lines? It gives WITP designers, AE or II, some things to think about. What's the harm in that? How does it serve to just abruptly shut this discussion down?



I'm not trying to "shut down" anyone's hopes and dreams, BERTO. Just pointing out that even with the best of wills and intentions, the "Admiral's Edition Expansion" has to deal with the limitations of the game itself. Just trying to "inject some reality" in regards to folk's expectations. Hope and dream all you want..., it's everyone's right, and I've certainly done my share.



jwilkerson -> RE: An AI Player's AE Concerns (12/16/2007 3:05:25 PM)


quote:

ORIGINAL: Mike Scholl

I've spoken to, and ask similar questions of, several programmers who've tried to deal with the WITP code over the last few years. And I've gotten the same answer I'm giving you. There is a very finite limit to what can be done to the AI with the current coding...., too much of it is "hardcoded", and it's "all over the place". Much of what we see as "AI effort" is just "scripting".

I'm sure the "AE Team Programmers" will do whatever they can to make the "AI" function within the new Admiral's Edition. But from the responses I've recieved, there isn't much hope of making it "better" than the current version...., just getting it to reccognize the new map features and units and be able to function with them as well as the AI in the current game will be a challange.

According to the "poll results", apparently at least half of the current players find the AI enough of a challange that they play "only against the AI"..., so they must find it adequate to their needs. For the rest of us, it would seem that PBEM is the answer until some group gets the time, money, and interest to re-write the code from the ground up and build in enough flexibility for a better AI to operate with.


Actually as one of the programmers in question, I'd have to say I disagree with parts of Mike's post. While the AI system in WITP does certainly pose challenges and limits to what we can do, things are not hopeless. And we did have the recent idea to pull the lists of units and bases out of the code and put these into the editor. We are doing this primarily to speed up our process of replacing the old units (slot #s) with the new units (slot #s). But we get a side benefit of having the AI now exposed to the editor, which will help make it easier to change for both our team and the moderators. So if the AI is easier to change will it be easier to "improve"? While having a full blown scripting engine [which we do not have] would make it far easier to add variations to an "AI", putting the AI data into the editor will save time for us and thus free up more time to tweak things in the code. For instance we do have some plans to add some new modules for invasions of India, Oz and NZ. But bottomline, improvements are possible in the current system.



Nomad -> RE: An AI Player's AE Concerns (12/16/2007 3:26:11 PM)

This is good news Joe, I truely thing that some modders can make AI only senarios for each side that will fare much better than the standard senario that has to support both PBEM and AI play. I am not looking for a truely intelligent AI, just one that does not do quite so many stupid things. Thanks again Joe and team.



JeffroK -> RE: An AI Player's AE Concerns (12/16/2007 11:51:16 PM)

Thanks Joe,

Its good to see that an improved AI isnt impossible.

Also, I have played a mod by Dingo where the AI gave you a good tussle, my view is that the AI in shorter scenarios is easier to program than in the full war scenarios.

From Mike
"According to the "poll results", apparently at least half of the current players find the AI enough of a challange that they play "only against the AI"..., so they must find it adequate to their needs. For the rest of us, it would seem that PBEM is the answer until some group gets the time, money, and interest to re-write the code from the ground up and build in enough flexibility for a better AI to operate with"

Maybe a wrong assumption, and appears to be denigrating the AI players. Many AI players dont want to put up with the hassles many PBEM games entail, want to get through a couple of years in the game and to try "different" things without the whine of "Thats not in the house rules"



witpqs -> RE: An AI Player's AE Concerns (12/17/2007 12:00:52 AM)

Right on, JeffK. The AI was only a challenge until I learned the game itself. I may try PBEM when AE is out, but the pace of game play and being tied to an opponents schedule make that a hassle. As far as the AI goes, to modify and old saying and make it family friendly, "Any beer is good beer."



wworld7 -> RE: An AI Player's AE Concerns (12/17/2007 1:10:51 AM)


quote:

ORIGINAL: witpqs

"Any beer is good beer."


Let's try not to be too crazy here...



Sabre21 -> RE: An AI Player's AE Concerns (12/17/2007 4:19:56 PM)


quote:

ORIGINAL: JeffK

Thanks Joe,

Its good to see that an improved AI isnt impossible.

Also, I have played a mod by Dingo where the AI gave you a good tussle, my view is that the AI in shorter scenarios is easier to program than in the full war scenarios.

From Mike
"According to the "poll results", apparently at least half of the current players find the AI enough of a challange that they play "only against the AI"..., so they must find it adequate to their needs. For the rest of us, it would seem that PBEM is the answer until some group gets the time, money, and interest to re-write the code from the ground up and build in enough flexibility for a better AI to operate with"

Maybe a wrong assumption, and appears to be denigrating the AI players. Many AI players dont want to put up with the hassles many PBEM games entail, want to get through a couple of years in the game and to try "different" things without the whine of "Thats not in the house rules"


I have to agree with you Jeff. My decision to be an AI only player has nothing to do on how good or bad the AI is. I've been playing computer wargames for 25 years and the only time I will ever play a human player any more is someone I personally know and typically that will be hotseat. Most human players I've encountered don't even compare to a weak AI when it comes to strategy and the added hassles of having to wait a week or more for a turn, people cheating or just quitting after a couple bad turns is just a waste of time. No, I gave up on human players years ago after too much frustration. Hopefully not everyone's experience's have been as bad as mine, but for me, AI is the only way to go.



mlees -> RE: An AI Player's AE Concerns (12/17/2007 7:25:43 PM)

I am currently playing as Allies vs. Japanese AI in one of El Cid's RHS scenarios.

I am only up to the third week of March, '42. I have sunk almost 300 ships with my land based air in the NEI, and the PI.

1) One of the most critically stupid things the AI seems to be doing is sailing it's ships without air cover. (My B-17D's based in Manila have attacked shipping docked, but not disbanded, at Formosa, and there is always little to no CAP.) Can this be changed within the scope of AE?

2) Another minor silly thing the AI does is that it is stubborn. If my air forces (or subs) succesfully damage a ship or TF, the AI continues to sail it towards it's original mission destination, instead of pulling it back to safety. I think that the AI should be "aborting" TF missions where ships become seriously damaged, except for the most critical operations or high value targets. (I am not sure, but I suspect that the AI continues to utilise ships after they become moderately damaged, where a human would be withdrawing them from service for repair.) Is this change possible within the scope of AE?

3) I did a snap invasion of Marcus Island in Feb '42. (Wake did not fall.) I plan to build up that airfield, and use my multiemgined bombers to put the smackdown on shipping heading from the Home Islands to Truk/Rabaul. The AI has, so far, not responded. I speculate that it hasn't, because the VP value of Marcus Island is still kinda low. However, a human player would immediately see the (future) threat that Marcus Island may pose. I think that the AI should be trying to secure the outlying islands that are within range of it's high value SLOC or bases. Is this possible (other than through jacking up the VP value through the editor) with AE?

4) The AI does not appear to try and suppress enemy airfields agressively enough. If the AI made serious efforts to shut down Clark, Manila, Rangoon, I would be in serious doo-doo. It did all right with Singapore, but all of my other major air bases seem to be attacked halfheartedly. (Is the AI leaving too much stuff on Naval Attack/Naval Search? I haven't loaded and inspected the dark side, yet...) Is there a change possible here?

Thanks!



JeffroK -> RE: An AI Player's AE Concerns (12/18/2007 12:12:18 AM)

As soon as you use a mod you deviate from what the AI is "scripted" to do. But as vanilla is so poor its a burden you have to paly with.

Re Marcus is.  AI is buggered if you go outside a reasonably historical approach, best way to Japan is via the North with Marcus as your southern flank. (The poor weather helps to take out the Zeke & Betty as threats, I;d put up the USN Battleine against the IJN Battleline). You really have to try the historical lines of approach



el cid again -> RE: An AI Player's AE Concerns (12/18/2007 3:18:28 AM)

Good comments. Actually, I say there is no AI at all. What we have is hard code - not intelligent evaluation - except for certain very specific exceptions.

So what "AI" does is merely go where history went. That kinda sorta works for Japan - because it is predictable where it needs to go for a while. But - yep - the moment you diviate from the plan and do something NOT done in history - AI is wholly blind to it - since it does not really think about much of anything. It says "unit x, go to hex 1234" kind of thing.

AI can be slightly programmed. To the extent that AI is part of running logistics, we can tell it to use certain points by values we set in the editor. I find 90,000 is a sort of clue for the AI - it means "in this scenario this is a place to send supplies etc TO" - even if it is not in stock. I find port values and airfield values adding up to 7 or more help AI know something similar. And of course putting a HQ there helps AI know something similar. I find you can assign an objective to a unit - and AI will try to send it there. But it is stupid - it won't assign the right command! So in an RHS AI oriented scenario all the commands are set - as clues to AI where these units need to go - if different from where they appear.

My question is - will they upgrade AI - or anything else - for WITP I? We have disabled nations (one per side),
lots of hard code in slots that need not be there, stuff like that which could be changed. I suspect it won't be now.



Charles2222 -> RE: An AI Player's AE Concerns (12/19/2007 12:22:14 PM)

quote:

ORIGINAL: Captain Cruft

I think it might be better to just admit that it's impossible to make an AI for something as complex as this that isn't utterly useless. Then we could have the thing earlier.

That won't happen of course. Although this is a real labour of love, it's still a commercial venture. Though for how much longer this industry can keep selling something that doesn't exist I'm not sure. As long as people keep buying I guess ...



No, you got the cliche all wrong. The cliche is that gutting the AI gives more options to the game, not less time making the game.



Andy Mac -> RE: An AI Player's AE Concerns (12/19/2007 3:35:33 PM)

We will do our best within the constraints we face we can do no more.




NormS3 -> RE: An AI Player's AE Concerns (12/19/2007 11:27:09 PM)

You have done more for us than we can ever repay you. Please keep up the fantastic work. I have not bought a game of any kind since you dealers got me addicted to Witp.



hvymtl13 -> RE: An AI Player's AE Concerns (12/20/2007 10:49:18 AM)

Another AI only player here. Not that I could not or would not PBEM, but I purchased the game with the intention of gaming against the AI. Sometimes 8 or 10 hours straight no worries. I can only say at this point I hope you included a generous amount of hari kari swords for these poor Japanese AI Commanders of ground, air, and naval forces. I have dishonored them badly and I am sure that many if these poor AI devils have taken the more honorable route and gone to visit ancestors rather then return to the homeland. I mean, I'm expecting the AI to announce any day now- Oh hell no, this is just plain wrong I quit! lol?
So anyways, when you guys build a new AI the next time please find a hard core pbem player and let him coach the programmer on what the AI Needs to do. This one is like a moped- fun to ride, but nothing you want to brag to your friends about.
Don't get me wrong I'm not hatin on you guys. Matrix has and is doing some cool games. I hope you are serious on the AI improvments to this one in the upcoming add-on. It sorely needs it and I'll be happy to chime in on suggestions when/if you make a thread for one.
Cheers



Andy Mac -> RE: An AI Player's AE Concerns (3/11/2016 3:22:53 PM)

Wow looking for the AI scripting thread I wrote all those years ago and I came across this pre release thread wow how time flies !!



Lecivius -> RE: An AI Player's AE Concerns (3/11/2016 4:04:53 PM)

Phht, what's 9 years? [;)]



Trugrit -> RE: An AI Player's AE Concerns (3/11/2016 4:55:47 PM)


I've never played any game longer than I've played AE.

Much of that time I've spent playing you Andy.

Many thanks



DD696 -> RE: An AI Player's AE Concerns (3/11/2016 10:32:43 PM)

Holy Crap!

Saw this and wondered who dug up this pounding I took over hoping there would be a vast improvement in the AI from the original WiTP to AE.

AndyMac, definitely the culprit of this Resurrection of the Dead.

But, more importantly, the one who vastly improved the AI capabilities and ensured that this great game would find the players it needed in order to be a success.

I am sure that a great many players cannot thank you enough - I know that I can't.

Thanks again, Andy. It isn't perfect, but it is vastly better!



PaxMondo -> RE: An AI Player's AE Concerns (3/12/2016 12:05:06 AM)

Reading those many posts by Mike ... do miss him here .... Absent comrades! [sm=00000436.gif]



stuman -> RE: An AI Player's AE Concerns (3/12/2016 8:00:48 PM)

Thanks to all who have made this game so enjoyable !



Rising-Sun -> RE: An AI Player's AE Concerns (3/13/2016 8:55:34 AM)

I also noticed when I had my CAP on and the next day I put those pilots on rest, then the bombers came. Then days later in the same location, finally got a chance to upgrade those fighter planes, then while they are put on standby, the bombers came. Anyway I just wondering if the AIs were cheating, I simply put the fighter group CAP and rest between days, the bombers only came when the fighter pilots were resting.

So this time I use a bait if they want to cheat, I give them what they want alright. So I put crappy air group on that airfield and another airfield nearby with heavy CAP, well the bombers came alright got chewed up on that spot where the planes are resting.



Page: <<   < prev  1 2 [3]

Valid CSS!




Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI
3.84375