Bullwinkle Posted January 18, 2003 Share Posted January 18, 2003 I was befuddled/amazed at the AI in 2 recent games. I play AI only because 1. I don't have the time for 1-on-1 and 2. I don't know how 1-on-1 play even works. I've never played anyone any game over the internet and would be lost in doing so. Anyway, to my point. I'm using v1.06 with settings Beginner and +1. I was playing the Axis (FOW, FF, Scorched Earth on). In my first game I took Poland, the low countries, Denmark, & France rather rapidly, so I saved the game at that point. Russia came in late and the US even later. While moping up Russia, the US/UK attempted D-Day but to no avail. I simply covered all available landing hexes with Italian corps and that was that. Their air power scattered itself over various targets and so was ineffective. It wasn't much different in Russia, where I was able to go into the Balkans with a few corps and take the oil reserves. Even in situations where the Russian unit only had to move 1 hex to retake a mine or oil field, it stayed in place. OK, so on my second game I started where I had saved the first - right after the fall of France. This time Russia & the US came in much earlier. But what surprised me was the change in their tactics! The Russians heavily defended the Balkans and always retook available mines & oil fields when they were vacated. When D-Day occurred, the Allied air power focused on those units inhibiting invasion and so were able to get several units ashore. This never happened the first time. I won the second game too, but how & why did the AI behave differently? I really don't believe that it actually learned, but I have no other explanation for the change in play. Any ideas??? Link to comment Share on other sites More sharing options...
JerseyJohn Posted January 18, 2003 Share Posted January 18, 2003 I've noticed similar things on resumed games. Apart from your connection problems, regarding games with a computer opponent: Against the AI you might have a better game with the Fog off as suggested by Old Man as it enables the AI to see you're units and make better decisions. Also, it plays a better game with the setting at Expert and you'll find yourself improving quickly as you'll be playing against better computer strategy. I think the lower skill levels may be more random as well, accounting for the vast difference when your game was resumed. Or -- it could be your computer is learning. Ever see that Movie Collosus, the Forbin Project! -- they were also surprised. [ January 18, 2003, 12:44 AM: Message edited by: JerseyJohn ] Link to comment Share on other sites More sharing options...
AndrewT Posted January 18, 2003 Share Posted January 18, 2003 Bullwinkle, playing 1-on-1 by the Internet or on a local Lan is easy, it is no different to playing against the AI in terms of how the gameturn runs. People make much better opponents. See the 'TCPIP Quick How To.txt' file that comes in the current patch. [ January 18, 2003, 12:46 AM: Message edited by: AndrewT ] Link to comment Share on other sites More sharing options...
Compassion Posted January 18, 2003 Share Posted January 18, 2003 JerseyJohn Good reply and kudos for name checking Collosus... trivia q... what was the Russian computer called? Anyway, if playing against the AI, playing with FOW off is preferable no matter what level you are playing at unless you like dumb single unit advances. I also like to tryto not look at the computer's setup during my turn. Harder as the Axis, but still... Link to comment Share on other sites More sharing options...
JerseyJohn Posted January 18, 2003 Share Posted January 18, 2003 Compassion ".......2x2=4....7-5=2.... We are Collusus ..... We are Guardian ..... We are One..... We are Unity .....where is Dr Forbin ...." You are obviously a fan of great but under appreciated movies. Good Question. My first answer was Protector Trivia Q. back, what was the name of the Russian computer's creator? [ January 18, 2003, 02:59 AM: Message edited by: JerseyJohn ] Link to comment Share on other sites More sharing options...
pzgndr Posted January 21, 2003 Share Posted January 21, 2003 how & why did the AI behave differently? This is always an interesting question. Hubert has previously attempted to explain his unique fuzzy logic AI which is very good as AIs go, and now may be a good time to hear some more about it and how it might be improved for SC2. There's no AI learning that takes place between games, but I wonder if there is any learning taking place during a game. Do various weighting factors get adjusted along the way or what? Could iterations of AI vs human games be used during playtesting to optimize weighting factors and develop a broad range of AI strategies that are as good as playing a human? Link to comment Share on other sites More sharing options...
Recommended Posts