What is the status of AI? (Full Version)

All Forums >> [New Releases from Matrix Games] >> World in Flames



Message

Princessjj1 -> What is the status of AI? (10/31/2013 10:36:31 PM)

I know it's not going to be there on release. Is this something that is definitely being worked on, is it a "well, we hope so, but we don't know", or is it a "no way, no AI for you."



Shannon V. OKeets -> RE: What is the status of AI? (10/31/2013 11:03:42 PM)


quote:

ORIGINAL: Darkstar One

I know it's not going to be there on release. Is this something that is definitely being worked on, is it a "well, we hope so, but we don't know", or is it a "no way, no AI for you."

Working on the AI Opponent is the chocolate cake with vanilla ice cream at the end of my long trial and tribulations getting the Solitaire, Head-to-head, and NetPlay version done. I'll still have some cleaning up to do post-release: special rules to finish for a couple of scenarios, a half dozen more optional rules, and getting NetPlay working for up to 6 players. Once that is completed, I'll be full time on the AIO. Sometime later will I look into making PBEM a reality.

---

It might seem weird, but I haven't been able to play MWIF. I write code to make things work, and fix bugs that beta testers identify. But I don't actually start a game and play through turn after turn exploring various strategies and tactics. That takes too much time - which I never seem to have available.

But as soon as I'm back full time on the AIO, I'll be working with Peter on evaluating the setups for all the minor countries. And testing how well they do against various attacks by enemy major powers. For instance, the Poles, the Finns, the Rumanians, the Dutch, the Belgians, the Yugoslavians, the Greeks, ... Then I can switch sides and see how well the major powers do attacking the minors; writing scripts for the AIO to put together land and air forces for taking hexes and capturing capitals. That will be great fun.[sm=character0272.gif]

There is no need to motivate me. I will be absolutely gleeful.



Missouri_Rebel -> RE: What is the status of AI? (11/1/2013 12:40:44 AM)


quote:

ORIGINAL: Shannon V. OKeets

I will be absolutely gleeful.



I may have to pull this quote out every once in a while.



WIF_Killzone -> RE: What is the status of AI? (11/1/2013 12:48:11 AM)

Ha, you will know Steve's in trouble when he starts to sharpen his ceremonial sword.



wodin -> RE: What is the status of AI? (11/1/2013 3:15:43 AM)

How long do you think roughly the AI will take to implement? Two or three years? If you can give a conservative estimate..



Shannon V. OKeets -> RE: What is the status of AI? (11/1/2013 3:16:45 AM)


quote:

ORIGINAL: wodin

How long do you think roughly the AI will take to implement? Two or three years? If you can give a conservative estimate..

No predictions! That's a lesson I have learned well.



Greyshaft -> RE: What is the status of AI? (11/1/2013 3:30:56 AM)

Will the AI have difficulty settings?

Perhaps they should be named after Generals with Zhukov being Very Hard and Himmler being Easy?



pzgndr -> RE: What is the status of AI? (11/1/2013 1:24:55 PM)


quote:

ORIGINAL: Shannon V. OKeets

quote:

ORIGINAL: wodin

How long do you think roughly the AI will take to implement? Two or three years? If you can give a conservative estimate..

No predictions! That's a lesson I have learned well.


I know there's been a lot of hyperventilating recently about AI in general being worthless and AI in particular for this game being impossible. I don't agree with those assertions. It's true that some AIs have failed to impress for a variety of reasons, but some others have proven to be quite good.

There are concerns that an AI cannot possibly handle WWII grand strategy decision making on a theater or global level; ie, diplomacy, research, production, etc. The Strategic Command series AI has in fact been doing all this quite successfully for several years. But, the SC series has been growing for over 10 years now and it has taken literally years to grow the AI scripting capabilities to be as robust as they are. And of course it takes someone to write and playtest literally hundreds of event and AI scripts to get things to where they optimally need to be for good players to consider the computer opponent challenging and competitive. But it IS possible. I've done it for my Advanced Third Reich mod, but I've been at it since like 2005 with my latest update as of last year. But again, it is very time consuming to get it right (e.g., running hundreds of AI vs AI games to identify and resolve issues), which is why most AIs fail to impress because they're never given enough post-release effort to make them really good.

Here's my 2 cents. I think Steve and his team have laid a good foundation for the MWiF LAIO parser and scripting capability to handle the different levels of decision making, and they have a good idea of how to populate the scripting based on all the AI strategy threads. So I believe a functioning and decent AI could be released in a year. Will it be perfect? Probably not. There will probably need to be some more fine tuning, perhaps additional coding to implement a few more AI features, and whatnot, especially to plug any holes where players exploit AI weaknesses. So I would also expect some additional AI improvements and enhancements over the years. IF Matrix and Steve stay the course and commit to making the AI really good then I expect we could have a challenging and competitive computer opponent in a couple/few years. [8D]

Something in MWiF's favor is that there is basically just one game and the one game version. Granted there are a dozen scenarios but they are all pieces of the one grand campaign, so once the full global AI is working OK then it can be tailored to the other scenarios. That's doable. But something else not in MWiF's favor is the laundry list of different game optional rules. The AI should be expected to work OK under any combination/permutation of these options. For starters it must work well under the novice/standard/advanced sets, and that should be doable. How well the AI may handle some creative personal sets of optional rules that could conflict with one another is another matter. Hopefully that won't be a major issue but it's a potential hazard down the line. Or, it could lead to some pretty interesting games!




vonRocko -> RE: What is the status of AI? (11/1/2013 3:25:24 PM)

Knowing this was really a one man effort to create this game, I must ask what will happen if (God forbid) something happens to Steve? Will Matrix/slitherine continue work on the AI and other promised features?[&:]



Numdydar -> RE: What is the status of AI? (11/1/2013 3:39:35 PM)

For anyone interested War in the Pacific AE (which is on sale and everyone should buy it ASAP [:)]) which was released in 2009 is still having the AI updated. The last major update by Andy Mac was October 2012. A full three years after the release. So I can easily see the same thing occuring for WiF.

Also the one major advantage that WiF has over 90% of the normal computer games, is there is already a HUGE body of knowledge about strategies in the game. So like WitP, numerous scripts can already be written based on this data. Most games need this done from scratch as part of the development. Not so WiF.

WitP AE also has something I think is really great. There are numerous AI scripts but only one is chosen when a new game is chosen. So as the Allies against Japan, each game you do is totally different as each differnt script changes how the AI will behave. So if WiF could have several differnt behaviors chosen at the start of each game, that would go a long way to improve replayability just like it did in WitP.



Page: [1]

Valid CSS!




Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI
0.640625