winky51
Posts: 164
Joined: 1/18/2005 Status: offline
|
In my game the German AI attacked the Soviet AI and the Soviet AI overran them the 1st few turns (july 1941). Thre were no German troops on the border. Remember though a game on this scale it is unbelievably difficult to make a good AI. They are trying to do it for World in Flames and personally I think they should just put out a networkable version. In GG World at War, latest patch, the AI is fairly decent but eventually you beat it without giving it uberproduction. If they can't make a perfect AI in that area based game then how the hell can they do it for a division or corp level game. They only way I can see developing a good AI for a computer game like WW2 is 1st off make sure its corp level at worst. Then set specific strategies for the AI. Lets say the computer is playing Germany. It picks an option at start of game. 1. All out vs Russia, ignore England 2. Russia 1941, keep England at bay (heavy atlantic battle with subs) 3. Russia 1942 from eastern front and Turkish border, Med strategy 1941 to protect Italy (except gibralter) 4. Russia 1942, Take Gibralter and Egypt 1941 5. Russia 1941 winter attack + med strategy (assumes Germany took out France fast) 6. Sealion, Russia 1943 (to take territory and delay Russia from crushing them close to border 1944. 7. Sitzkrieg, this is where Germany takes out Gibralter, all the med and attacks Russia in 1943 just to take borders so it takes a looooong time for the allies to take it back. Very boring but effective for axis to win on VPs. Now you have your strategic goals as the AI. Because you know what you are doing you can perfectly build for it and through everything you got into your strategy. Now as the programmer really you need someone who knows these strategies for the game so you can program the AI to do it. These are the basic German strategies used in World in Flames. The simply fact is whatever you want to accomplish with a military force is to through all of your strength at the selected objective till its taken, then do the same for the next one. WW2 showed both sides not doing that. And I find most AI opponents dont do that. Take Africa for example. The right play was for the Axis to take Malta, place a lot of troops in Lybia, and crush the English. Instead they sent a small force to the desert that got its supply convoys shot up to hopefully accomplish their goals. You can't do everything 1/2 ass. Japanese in the Pacific same things. They send out just enough to accomplish the mission and basically coinflip with the americans to see who wins the naval battle. Coral Sea, why 2 carriers, you knew how many the Americans had. Midway why the complex plan with 8 carriers. Put all 8 carriers in the main strike force and take the fooking island. Give the AI a goal, stick to the plan, build for it, move on to next goal. Makes for a better AI.
|