Hard AI vs Fast AI
-
Under what circumstances does the Hard AI play significantly better than the Fast AI?
-
@RogerCooper Primarily in any map that uses lots advanced unit options which don't directly impact HP and strength as the Fast AI uses just a simple HP and strength formula to estimate battles rather than the battle calc.
-
@redrum So on most maps, Fast AI is good (and much faster). Does the Fast AI handle one-use units properly?
-
@RogerCooper what do you mean by one use units? Suicide units?
-
@redrum Yes, suicide units
-
@RogerCooper I believe it would vastly overvalue suicide units as I think it'll just take their strength and not consider number of rounds it can fight. So it would value gas in Dom NML at strength 4 and overestimate how powerful gas units are. This would probably cause it to attack with gas in bad places and retreat from enemy gas even if it can hold areas because it thinks it would lose.
-
@redrum So if there are suicide units around, I better use Hard AI.
-
@RogerCooper Does the Fast AI handle multi-hit units well and make adjustment for units that can repair?
-
@RogerCooper Yeah, Hard AI should do considerably better if there are suicide units in the map.
I think Fast AI should handle multi-HP units fine so battleships on many maps for example. If anything it might undervalue them a bit.
-
@redrum I hope the Fast AI is improved to adjust for suicide units and other special cases. The Hard AI is too slow on larger maps.
-
A few things i observe about Fast AI (correct me if i am wrong)
Fast AI is more aggresive and does not care much about TUV loss. This is a good thing if you achieve some stragetic goals but lose some TUV. In what seems like an impossible attack, Fast AI attacks with very little land units and air units. It wins but forget to leave land units behind to get that territory. So in the end its just a lost of TUV. Maybe the problem is with casualty selection.
-
@Lord-Bevan It also seems that the Fast AI is more aggressive but the Hard AI also tends to make attacks without bothering to take the territory even if by losing 1 aircraft it could take a factory.
-
The battle calculator stops working above 512 inf. Maybe changing it could improve things.
-
I am reactivating this old thread to post the results of some experiments with the AI's and see how they compare in actual games.
My first experiment
Scenario World at War Variant - v3
https://axisandallies.fandom.com/wiki/World_At_War_Variant_-_v3
Version 2.6.14696
Results Hard AI did better. When the Allies used Hard and the Axis the used Fast, the Axis was too aggressive in Eastern Europe initially, which the Russians exploited successfully, overruning much of Eastern Europe.Both AI's played the Americans poorly, building land units and transports without bringing them across the Atlantic.
-
Scenario Classic Iron Blitz 2nd Edition (with no Russian first turn attack)
https://axisandallies.fandom.com/wiki/Classic:_Iron_Blitz_2rd_Edition
Version 2.6.14738
Results Another victory for the Hard AI. As the Axis it won in 15 turns. As the Allies it won in 26 turns.The Fast AI messed up as the Allies by attacking Germany with an insufficient force, which was destroyed. By the time it rebuild its forces, Russia had fallen to the Japanese.
Unlike the previous game, both AI's handled the naval war well.
-
Scenario Dragon War
https://axisandallies.fandom.com/wiki/Dragon_War
Version 2.6.14759
Results Evil won in 25 turns with the Hard AI and 30 turns with the Fast AI. Both AI's were unable to handle the multiple resources and most powers had unspent PU's at the end. -
Scenario World War II V3 - 1942
Version 2.6.14770
Results Fast AI won in 9 turns as the Axis, the Hard AI won in 8 turnsThe Fast AI successfully pulled off some risky attacks as the Axis, but still took longer to win the American even taking Italy for one turn.
With 4 victories in a row, using different scenarios, I can conclude the Hard AI is better.
