American Civil War Game Club (ACWGC)

Gamers,
Berto has commented that as time becomes available, he hopes to code improvement to the AI. But just exactly what does that mean and what are things that the AI is currently doing? Saying the AI needs to be better, is not what is needed. What is needed are specific examples of what the AI is currently doing and why it is wrong. Being specific is what Berto needs to successfully upgrade the AI.

This is not necessarily a wish list, but a series of observations. These observations can be used to improve the AI. Provide examples, perhaps even screen shots. Anything will help.

Provide examples of what the AI is doing:
Moving
Firing
Melee
etc...

_________________
Brig. Gen. Rich Walker
AotW I/3/4
Scenario Designer:
Franklin, Shiloh, Chickamauga, Antietam, Atlanta, Chancellorsville, Petersburg and Shenandoah

It just so happens that I am now working on a new feature for Panzer Battles that, as a side effect, might have positive implications for the AI. If all goes well, I should be able to apply this, in the near term, to Campaign Petersburg (i.e., available in the inevitable 1.01 patch) and the other updated CWB games also.

_________________
Civil War Battles Lead Programmer, https://wargameds.com
Panzer Battles & Panzer Campaigns Lead Programmer, https://wargameds.com
Campaign Series Lead Programmer, https://cslegion.com

Most obvious error the AI makes is it can't handle the difference between artillery and infantry. Many a time I have watch the AI charge my lines with limbered artillery.

It also has a problem with leader handling in that it tends to leave leaders wandering around alone or in the front lines where they get killed. Many a game against the AI, which really only do for a quick look at the campaign scenarios, I have taken out Grant or Lee because he was wandering around in front of the infantry.

Worse logic problem it has is that the AI isn't situationally aware. It has no problem advancing a brigade or division with short frontage and no flank protection so it can be easily surrounded and wiped out. But this is a tough one to fix.

So all you have to do is link up DeepMind to the game to play the other side.

_________________
General Kennon Whitehead
Chatham Grays
AoT II/1/3 (CSA)

Artillery seems to be the most obvious. It should stay unlimbered until there is a threat it will get attacked when alone. There should be a stacking logic that says if it has over 500 infantry or cavalry in the hex it wonder limber up and run away.

Kennon pointed out the obvious about the leaders. First priority should be that it is protected and after that that it can project its command radius to as many subordinate leaders (or units if Brigade leader) as possible. Corps or Army leaders should never be in a hex that can get attacked in the next human player's turn. It should seek shelter with a unit if it is Isolated.

_________________


Optional Rules I Use in WDS ACW Games:
(by column from left to right)
Column 1: All ON except for Man. Def. Fire; Column 2: All ON except for Alt Fixed Unit Rel.; Column 3: All ON except for Art.Capt. & Prop.Op.Fire

I don't play against the AI, preferring PBEM. As such I can't offer opinions as to certain aspects of AI performance.

Might I suggest a different approach to addressing the issue of the AI. Rather than attempting to be all purpose, focusing on developing strong defensive AI first would seem to be a good path. The multitude of options required for offensive proactivity is certainly much greater than that of defensive reactivity. A defensive AI which allows the computer to "see" everything and take action accordingly (introduced as an optional rule) could make for an interesting opponent. The human could designate at the beginning of each turn which side is the defender.

_________________
MG Robert Frost