AI CSA Weirdness

[jscott991]

So, I've started to get the hang of this game and, as is my pattern, I've been playing it almost non-stop in my free time. I've noticed a few things and I'm a bit concerned that the game doesn't have very long-term legs.

1. Why does the CSA AI move its commanders back and forth and leave them separated from their armies? This is particularly common with Lee. Lee seems to switch back and forth between the east and the west and the AI uses him as a reaction piece. He never is stacked with any significant number of units. This means that Lee NEVER attacks. The huge, dual ANV's just sit in Richmond and some other Virginia province and react back and forth. This makes playing as the Union somewhat boring.

2. Just like in FoF, the AI strips the west to create a huge army in Virginia. Is this turtling common? It seems to happen in all my games once I take Memphis. The AI doesn't really create significant armies in Tennessee and Mississippi, preferring, instead, to have about 150,000 men in VA.

3. Why does the AI build so much heavy artillery? In all of my games, they move a ton of batteries into every fortress along the northern Miss. and lose them as they retreat out of Paducah, Humbolt, and Memphis. It's wasteful.

Has anyone ever seen the AI use Lee effectively to mount attacks?

Defend the west?

Create three armies?

[Ralph1961]

I too have been spending more time on WBTS. I am almost at the point that I can play a competent Union side. The only way to really enjoy these types of games is PBEM, there is no way an AI can handle a game this complex. AI is just a learning tool, sad but true in almost all games.

[jscott991]

It is true that the AI can't be expected to be the equal of a human, but the very basic flaw of not maintaining a front on the Mississippi or in Tennessee seems a bit excessive.

How hard would it be to program an AI that doesn't put most of its forces in Virginia?

[Treefrog]

Ralph Harper has it right, the AI is a learning tool and the challenge comes in PBEM. The North commanded by AI is no real challenge, same for the South.

As you are both newbs, you may wish to play each other.

[Ralph1961]

I would gladly play jscott. I have some pbem experience, not a very good player but getting there.

[jscott991]

I'm not much of a PBEM'er, sorry. I also have modified my leader file quite a bit to try to get historical Peninsular officers to appear.

[Treefrog]

Ralph, you can pm me about a game, although I tend be rather slow sometimes due to work commitments.

[jscott991]

The AI is truly terrible. Some of it is just bad programming (moving the generals back and forth repeatedly, and ensuring that no stack consistently gets initiative, for example) that could easily have been avoided. Some of it is negligence (failure to garrison large, PP-rich cities like Savannah, Charleston, and Wilmington). And some of it, for whatever reason, is typical of any Civil War game (huge ANVs that sit in Richmond, while the Union overruns the west and Tennessee with ease).

All of this could and should have been fixed in early patches. You can say PBEM is better and that the AI will never be the equal of a human player, but to leave it in this shape was ridiculous. How hard would it have been to make sure the AI maintains a logical balance of forces between Virginia, the Mississippi, and Tennessee? How hard would it have been to make sure the AI doesn't move its generals back and forth, forth and back every turn, keeping its armies from gaining initiative? And how hard would it have been to simply hard-code the CSA AI to keep 1-2 infantry units in Savannah, Charleston, Wilmington, and other ports to make it next-to-impossible for the the Union to vulture these cities up so easily?

Frankly, I feel like I'm playing only against the initiative die rolls, which only calls attention to how random this aspect of the game is. I advance if I hit a certain die roll and I bog down if I fail it. The CSA's army has nothing to do with it. And the CSA AI should be the easier one to program!

Frankly, the reviews were very misleading on this point.

[Longstreet_slith]

Hi jscott991,

I remember your posts on the FOF forum about the AI problems with FOF. Now that you have played both games, I would be interested in hearing your assessment of the AI of both games. Which one would you say is better? There are obviously some problems with both, but would love to hear your sense of which AI is going to play better.

Thanks!
-Longstreet

[bigus]

ORIGINAL: jscott991

The AI is truly terrible. Some of it is just bad programming (moving the generals back and forth repeatedly, and ensuring that no stack consistently gets initiative, for example) that could easily have been avoided. Some of it is negligence (failure to garrison large, PP-rich cities like Savannah, Charleston, and Wilmington). And some of it, for whatever reason, is typical of any Civil War game (huge ANVs that sit in Richmond, while the Union overruns the west and Tennessee with ease).

All of this could and should have been fixed in early patches. You can say PBEM is better and that the AI will never be the equal of a human player, but to leave it in this shape was ridiculous. How hard would it have been to make sure the AI maintains a logical balance of forces between Virginia, the Mississippi, and Tennessee? How hard would it have been to make sure the AI doesn't move its generals back and forth, forth and back every turn, keeping its armies from gaining initiative? And how hard would it have been to simply hard-code the CSA AI to keep 1-2 infantry units in Savannah, Charleston, Wilmington, and other ports to make it next-to-impossible for the the Union to vulture these cities up so easily?

Frankly, I feel like I'm playing only against the initiative die rolls, which only calls attention to how random this aspect of the game is. I advance if I hit a certain die roll and I bog down if I fail it. The CSA's army has nothing to do with it. And the CSA AI should be the easier one to program!

Frankly, the reviews were very misleading on this point.

The AI can be... and is better than you think.....