Page 2 of 2
RE: Length of game.
Posted: Thu Sep 30, 2004 6:05 am
by Hentzau
Heard of something similiar to this AI library of games idea in a recent football video game, don't remember which one though. A bit different in that the football game tracks your moves and then you can swap that file(or whatever) with your friend so he can play against an AI "you".
RE: Length of game.
Posted: Thu Sep 30, 2004 2:54 pm
by mlees
Wait a sec!
Not all of us players are budding Moltkes![:(]
If the AI has a library of hundreds of games to draw from (including the razor sharp matrix beta testers), I'll NEVER win a single player game!
Make the difficult/hard AI setting tough, but let the easy/baby-in-diapers AI setting be easy.
And no, I can't learn to beat the AI by just playing over and over. I have a hard time figuring out what I did wrong, (and a living opponent sure as heck won't tell me!) and what was just plain bad luck. If I lose more than a few games, I end up moving onto something I can win (without cheat codes). [:D]
RE: Length of game.
Posted: Thu Sep 30, 2004 4:34 pm
by *Lava*
Hi!
There is a "difficutly" setting for the game. So you will be able to give your side full advantage if necessary.
[;)]
Ray (alias Lava)
RE: Length of game.
Posted: Sat Oct 02, 2004 6:32 am
by FrankHunter
Its still too early to know how good or bad an opponent the AI will be due to it requiring a lot of example games before it can do more than wing it.
RE: Length of game.
Posted: Sat Oct 02, 2004 12:06 pm
by wodin
This game above all others is the one Im waiting for. Even more so than CL and as Ive become a Squad BAttles addict even more so than the next in that series (unless thay did a WW1 version).
RE: Length of game.
Posted: Sat Oct 02, 2004 4:05 pm
by pzgndr
I guess I'm still leery about this concept of a games library of hundreds of games. That cannot mean hundreds of completed games with the latest gold version upon release. So games in the library based on early versions with whatever bugs may have a potential strategy waiting there which is more or less questionable now. Looking ahead to future patches with expected game tweaks, that library may become a bit dated.
Don't get me wrong, I like the concept. I'm just suggesting that this learning mode for the AI could be an on/off option for players to use if and when they desire to engage it. The default AI and the games library could be updated together as future patches are released. There could also be a way for players to submit their completed games, especially human vs human games, to help populate the generic games library over time. I think having the AI learning mode is a neat idea. However, let's please ensure we have a very solid foundation of relevant games in the library and inexperienced players don't risk recalibrating the AI while they themselves are in "learning mode."
Games like this take a while to appreciate the true depth of grand strategy decision making. I would feel much more comfortable being able to upload the latest AI and games library as they become available, and at some point when I am ready then engaging the AI learning mode. It's just a request. I'm still looking forward to the game regardless. [:)]
RE: Length of game.
Posted: Wed Oct 06, 2004 6:15 pm
by FrankHunter
I have some of the same worries about it. But it wouldn't be difficult to allow the importation of other people's AI's, either replacing or merging with your own. As I watch the AI I'm happy with much of it, for example it maintains a line, it doesn't leave gaping holes if it can help it, it puts resources into R and D, it keeps its forces supplied and reinforced etc. Sometimes it demonstates an ability to go for the kill and other times it doesn't. To me it doesn't matter if it beats humans, I just want it to try different strategies so the player has something he can play against when not using it for PBEM.
RE: Length of game.
Posted: Wed Oct 06, 2004 9:34 pm
by Hanal
Frank, you made a very interesting comment: "To me it doesn't matter if it beats humans, I just want it to try different strategies so the player has something he can play against when not using it for PBEM." Now this makes me wonder if the game is for me....I am not a PBEM player, and if I cannot get a reasonable game from the AI, and at hard difficulty levels, actually have a chance to lose, then I think that GoA, which I have been waiting for, for a long time, may not be for me....I need a challenging AI since I do not have the time or inclination to play PBEM....If the AI is going to be a weak sister in this game, to just be used by PBEM players to work out different strategies, then the market for this game may have been narrowed a bit...
RE: Length of game.
Posted: Fri Oct 15, 2004 10:43 pm
by SirRodneyOfGout
I'd also primarily be a solo player for this game, and I think this type of AI is even more incentive to buy it.
Too many game AIs are based on a "best game" approach. The developers program an AI that will beat inexperienced players, that is until they play enough to understand how the AI works. These games are like puzzles, but once the puzzle is solved the game is no longer interesting to play.
Too many people are looking for a perfect game from an AI. I do play some PBEM, and have played head to head boardgames since the 70s. Nobody, but nobody, plays a perfect game. I've seen lots of head scratching mistakes, made many myself. So no expectation that a programmed AI would do better, or would always beat me. The only problem is that an AI will repeat its mistakes, and the pattern can be quickly determined.
So I want an AI that can make different high level choices each time I play. Where and when to attack, what to research, etc. As long as the AI is also reasonably competent in unit handling, supply and other low level routines. That way it is as close to playing a live opponent as possible.
I really hope this works, because I don't think I'm alone in wanting this kind of an AI.
RE: Length of game.
Posted: Wed Oct 20, 2004 6:29 pm
by FrankHunter
JP, if I could write an AI that would beat a human I would. But generally a human can recognize patterns and make intelligent guesses that are way beyond any AI I could write.
One thing a human does in strategic games is that over time he builds a library of actions. Things like optimum setups and opening moves get used over and over. I wanted the AI to be able to do the same which is why there's a limited amount of learning. The German AI for example will make a decision based on Schlieffen or von Moltke (the elder) as his overall opening strategy. The Austrian AI will choose between Serbia and Russia as its target. The AI will have a library of setups and opening moves available to it but these are not static scripts. Instead, they are behaviour that is learned from the player and can change. Often it will be the case that the AI will be in a situation its never seen before exactly, so it will look at the past games that most closely match and make a guess or if it can't even do that it will fall back on the hard-coded AI which plays conservatively.
If the AI isn't challenging after awhile, the difficulty level allows you to pump up the available resources available to it, classic cheating but there it is.
RE: Length of game.
Posted: Wed Oct 20, 2004 8:24 pm
by Hanal
Thanks for the reply Frank and I understand the difficulty you must be facing regarding the AI....people shun the idea of AI cheats as if they destroy the integrity of the game, but, as you said, if a player beats the AI regularly after a time, then cheats could be used to level the playing field a bit...I have no problem with the concept and will use them in order to continue playing the game....I know now that I really cannot pass on this game.....geeez...I flip flop like---, um, best leave politics out of this forum!...[:D]