Broken Axis AI Decisionmaking

[Vonrich]

Hi Alvaro,

I ran into a situation where, playing as Allies, I took Copenhagen before the Germans could. This may be resolved in a new version because I know you made Copenhagen easier to take for the Germans, but what happened was that they would send a Pz Corps in, kick my troops out to the north, take the city, then withdraw. I was then able to walk right back in, causing them to redeploy the Pz corps. The end result was that I was able to hold France. Paris was never captured.

Even if the Copenhagen problem itself is solved with the map redesign, it still serves to highlight a problem with the AI assessing threats to territory it wishes to control.

Von

[Vonrich]

See attached

[AlvaroSousa]

If you are playing the A.I. what difficulty level?

[Vonrich]

Historical. I didn't think the difficulty level would change decisionmaking ability, only make the units beefier. Is that not correct?

[AlvaroSousa]

From what I understand Veteran with Optimal supply is a good challenge.

Currently I am playing 4 games at once for balance. After that I will reexamine A.I.. The A.I. is comparable to SC's A.I. A good player should easily defeat it at normal levels. Hell I stuff the Axis A.I. in SC3 Europe in France.

[Vonrich]

Please don't take this the wrong way...

Needing to up the unit strength is a poor replacement for intelligence. It frustrates intelligent decisionmaking by forcing overloading. Yes, intelligence will usually beat strenth, but a fair fight where intelligent decisionmaking is the only factor is far more satisfying. I know this is why we can play against humans, but a challenging artificial intelligence without artificial strength can be equally as satisfying. To be fair, it's not really AI but an optimized algorithm...because a true AI will beat us every time now, but you know how I mean

[AlvaroSousa]

Not offended.

Games like SC and WP require a very high level of deep thinking that an A.I. is not capable of doing.

Take Civilization. They have a whole team behind it and the A.I. doesn't play well militarily.

Unity of Command is a completely different game with a fixed battle field on a time table. It is much simpler to make an A.I. for that. I read an article on how they approached it and it is the same way I would.

Unity of Command is more like chess with a fixed set of moves and finite complexity while SC or WP is more like poker which is based on math and a mountain of variable with an very expansive field of decision trees.

But I like working on A.I. so after game balance is noted this will be my next project.

[Vonrich]

Totally understand. Thanks, Alvaro.

[pzgndr]

ORIGINAL: Alvaro Sousa
Games like SC and WP require a very high level of deep thinking that an A.I. is not capable of doing.

Well, and you know this, the AI does not think at all, it executes scripts. With enough clever scripting and sufficient randomness, the AI can appear to be thinking like a clever opponent.

With regard to your comments about SC's AI, perhaps true for the generic game but I've scripted my A3R mod over the years to provide challenging Axis and Allied computer opponents. Granted, I've been adjusting and rewriting the scripts for about 15 years now through multiple game updates, so it's not trivial. But, it's not impossible! [;)]

[Vonrich]

You could have an entire team dedicated to doing that and still not offer the same experience as playing against real intelligence as that decision tree approaches infinity

[AlvaroSousa]

I am the entire team.

[PanzerMike]

Go team Sousa [:D]

[AlvaroSousa]

Well from the coding side. I do a lot of the graphics but I hire out stuff I can't do. Like I really like my music composer. The music for the Pacific is better than WarPlan Europe.

[Edorf]

ORIGINAL: Alvaro Sousa


But I like working on A.I. so after game balance is noted this will be my next project.

Thumbs up[:)]

[iDachs]

ORIGINAL: Alvaro Sousa

I am the entire team.

A talented team at that, one man or otherwise.