Divisions

[McNaughton]

I have been running tests to see how or if the AI uses divisions. Here are my results.

Control: I gave the CSA AI 12 Free Divisions via events in August 1861. By September 1861 I saw the following results...

75% of the Divisional HQs had forces under their command, 25% were at 'full strength' as in viable combat forces. Given time, I expect that this rate will increase to 100%.

However, here's also what the AI did...

#1. Used a lot of militia. About 1/3 of Divisional troops were militia, even though regular troops remain in independent brigades.

#2. A very important issue, I have not seen the AI add a signle Divisional leader. It keeps their leaders outside of divisional command, even if they are Brigadiers.

#3. On its own, it builds very few commands. From 1861-January 1862 in another test the AI Built just two divisions, while I had 17 (including the 5 I started with).

The AI uses divisions, indeed are more effective when they have them. Also, they are more aggressive in the given timeframe. I noticed that formed into divisions, forces out West were way more aggressive, actually attacking my fortified towns (with varying levels of success) instead of 'stumbling'.

I suggest the following.

A) A greater importance for the AI to build Divisional HQs. They use them, even as is, and are way more effective when they do.

B) Get the AI to devote Brigadiers to Divisional Command. The unit is even more effective when given a brigadier in the division.

C) Have a priority of unit deployment to Divisions, with militia/volunteer at the bottom (if there are 3 Regular, 3 Militia, 3 Volunteers it uses all of the Regular first, then the Volunteer, then the Militia, presently it appears to be random).

D) The AI devotes units built to the frontiers that are nearest to them. This means all units built out East will fight out East. This is a major stumbling block out West, since units like HQs (only buildable at capitol cities) will always be deployed in the East and will never (or rarely) ever appear out west.

I have found giving the AI free divisions to help them a significant amount. If we get the AI to build divisions (and deploy commanders to them), the formidable AI will be even more of a challenge.

[tc237]

Excellent work!!

[pasternakski]

I have found giving the AI free divisions to help them a significant amount. If we get the AI to build divisions (and deploy commanders to them), the formidable AI will be even more of a challenge.

I agree with tc237. Great stuff here. Thank you.

I have one small additional comment. I am not sure that the ultimate solution to improving the AI is "free" divisions. In fact, I would hate to see this become so, as the result, the way I see it, would be to turn the AI into a player with capabilities and advantages superior to those you the human are accorded. I have often lamented this in other games, where those responsible for the design have given up on competent AI design and given you the choice of either being dissatisfied as a solo player or moving on to PBEM.

Not that there's anything wrong with PBEM, of course, or for that matter FtF. I play computer wargamers extensively both ways, with a great deal of enjoyment. I am also fully aware of the AI 'enhancements" that are already built into the game, and have no big difficulty with them - so far (the only one that troubles me a little conceptually is the "fog of war" cheat, but it works so well in helping the AI's ability to compete, I use it).

I just want to see what the mutated squirrel and his pals can accomplish with "Athena" on a straight-up basis before we move too far in the direction of sacrificing humanity on the cold, hard marble slab of the computer's "Temple of Limitations."

[LAVA]

#2. A very important issue, I have not seen the AI add a signle Divisional leader. It keeps their leaders outside of divisional command, even if they are Brigadiers.

Are we really sure that this makes a difference?

For some reason I seem to remember Pocus saying that during battles what matters is what special qualities are brought to the fight by the different leaders, thus, the AI will include good leaders into a Corps organization, even if they have no subordinate command. My very limited experience with the game would suggest that is true. Don't know what is under the hood, but in the battle results which are presented, you appear to get the leadership bonuses from the overall commander and then all the special qualities from all leaders in the formation thrown in as modifiers.

Nice research, BTW!

Ray (aka LAVA)

[McNaughton]

Thanks,

Leaders incorporated in Divisions help the command limitations. With leaders in a division, it affects the command rating, while brigadiers just in the corps do not (as far as I have seen). I ran tests, looking at total efficiency (maximizing troops and command rating), and generals in divisions are better than divisions without Generals and Generals sitting in the same corps without a command. While still good that they use leaders, it is better if they are integrated into Divisions.

Regarding the event giving Divisional HQs, it was merely to test if the AI can use Divisions rather than as a complete answer to the problem of the AI and Divisions. Now we know what they can do, we just have to get them to build them, and deploy them to all theatres (as I fear deployment is based almost solely on the closest front).

However, until (or if even possible) AGEOD manages to implement an AI that builds divisions and deploys them to all theatres, it may be a good implementation to have these divisions appear via event, at applicable cost of manpower, money and war resources...

Optimally, I would love for the AI to do all of this on their own, without events. However, until then, we could do so via events. We could also then test other behaviours the AI encounters so when they are taught how to build divisions, we already know how they use them (i.e., if they put too many militia in, if they keep them only in specific theatres, etc.).

I am going to test out giving them a HQ in the west, to see if that further improves combat performance and aggressiveness.

---Back to testing---

Another thing I have discovered. When a division is 'savaged' in combat, the AI is smart enough to move the unit to a 'safe town' and start rebuilding damaged units. In fact, the AI kind of became passive once the divisional HQs arrived at the front, I figure that they were in the middle of reorganization. After reorganization, the units in the West were significantly more aggressive (and successful). Those troops in the East appear to be consolidating their forces to a greater extent, but, individually their corps are more efficient in command ratings, but, are individually more powerful. It is as if they are waiting until their divisions are full before beginning offensive operations. I will run the game into 1862 tomorrow and see what they do when their units are full.

[marecone]

Excellent work, indeed!

[Pocus]

Give me some time, and the AI will progresses. For now, I'm still mostly devoting my time to bugs and numerous user interfaces requests, so the AI don't get much (perhaps one hour there and there per week!).

a leader in a division do add something, as McNaughton said. This is strange that the AI don't merge generals into divisions, this is supposed to work though...

[McNaughton]

It may be that I haven't given them enough time to incorporate leaders, since the AI is not as efficient as a player in reorganizing (it still does pack in the troops). It still has some leaders in command of brigades (even a two star general), as well as two star Generals supporting a three star General (Jackson is still in Johnson's stack as a supporting commander, not as his own corps commander).

I am really impressed at what the AI does with the tools it is given at its 1.02 status. They do use divisions very well, and are doing a fair job in reorganizing them.

It is tough to get the AI to do what you want it to do, as in one situation you want it to do A, in another it is best if they do B. Modding "Hearts of Iron" AI was a severe annoyance!

[jimwinsor]

Division creation is at the same time one of the most mentally challenging yet ultimately rewarding tasks a player undertakes in this game. It took me days to master (what I think are all) its intracasies. Hence I tend to be very understanding of any AI shortcomings currently in this area.

They take a long time to build (and perhaps transport west), so you have to anticipate their need by 3-4 turns.

Then you have to ship them off to the "assembly areas." For example, Bowling Green in the west for the Union is two turns from DC by rail; all reinforcement brigades destined for new divisions in the "middle" west go their for assemby. Not to mention any unassigned western brigadiers.

Cairo looks good, but is actually bad; it is exactly 32 days by rail from DC. The town two spaces north (I forget the name) is actually better as the "far" west assembly area...it's get there in precisely 2 turns.

When a precious HQ arrives at the assembly point, the cookin' begins! If their is an elite unit present, include one (but no more!). Add a sharpshooter (again, only one!). Perhaps even a marine/sailor for that Pontooner icon as well. Make sure the 16 combat elements are properly combined arms to, with at least one artillery and cav element in the mix.

So, you should have 16 combat elements, one HQ element, and one leader element (=18), all combined to give you an Elite Pontooning Sharpshooter with inf, cav and art all in the mix. The perfect division!

Well, not quite. There is also the traits of your brigadier to consider. If he is an Artillerist, for example, you might want to include a few more heavy guns (better than 6 lbers) in the mix too, for example. Artillerist is a stack ability, so doing this helps ensure any stack the div ends up in is fairly well artillery laden. Common sense stuff like that.

If a brigadier is not yet available, leave a space open for one when one does show up.

So yeah, lots of stuff going on here. Heck, I shoulda put this post in the Strategy section....

[pasternakski]

Great post, Jim. I hung on every word. As Bob Dylan sang once:

Then, she opened up a book of poems and handed it to me,
Written by an Italian poet from the thirteenth century.
And every one o' them words rang true, with a fire like burning coal
Flaming off of every page like it was written in my soul
From me to you...
Tangled Up in Blue.

[McNaughton]

I absolutely love creating my divisions.

I plan their development months in advance, and become a lot like McClellan (a nasty side effect!).

I can organize an effective army, but am loath to risk them in battle! This game makes me very cautious, as I know that even a well designed corps can be wiped out if put in the wrong situation.

I have designed my Eastern Divisions, based on a certain pattern, very artillery heavy. My Western Divisions are have more cavalry than artillery.

I fight 1861 using the forces on the map. I spend 1861 organizing my divisional HQs and newly produced brigades as the core of the 1862 army. I build 3 Divisional HQs every other turn (along with support, my corps are always stocked with medical, engineer and signals, with the occasional balloon).

The AI can manage and build divisions if given the HQs. All that needs to be done is to get the AI to have those HQs and they will be flying!

[DEL]

McNaughton

Let me for the moment be Lincoln to your McClellan and tell you "If you are not using the army, I should like to borrow it for a short while."

Back on topic
Encouraging to see that given HQ's the A.I. is capable of putting them to good use. As I've said in the other topic on removing HQ's, it would be a real shame to see them removed, especially given all the effort that has already gone into having them in AACW. Like you I take my time building an organized force. I'm never in a rush to end the war. Long lasting war is much more fun. There is alot of fighting that must be done!

[McNaughton]

Well, by 1862 over 50% of the divisions were full, with the remaining about half full themselves... To me, this says that the AI can build divisions, however, they tend to run them around too much during the winter months, giving them casualties they shouldn't be experiencing (the AI needs to go into winter quarters a bit better than they do). Also, even though there are 8 leaders in one of the corps, none of them are applied to existing divisions.

Also, one CSA division out west is at around 450 strength... As strong as my strongest division (mine has a leader though!)...

Imagine if the AI applies leaders?

The way the AI works in building divisions is the following...

All HQs are sent to 'large stacks' of troops. They are purposely moved toward these large stacks and start filling up. If there aren't enough troops, the HQs then move on to the next stack, and fill up there. Their divisions are fearsome, even without leaders.

If the AI can be taught to create Divisional HQs during the winter months, when they should be in winter quarters, they will have all the time in the world to create their new formations to use during the spring-summer-fall campaign season. They needed about 2-3 months to start up all 12 divisions I gave them. Even the weakest of the divisions put the AI in a better position than if they didn't have these divisions (1 division of 2 brigades is better than just 2 brigades on their own).

I personally think that the AACW AI can use divisions and army HQs, it just needs to build them and apply leadership. It isn't as if the AI has no clue as to how to build divisions with HQs. In fact, it is fairly intelligent, as it moves HQs to stacks, not stacks to HQs. Some divisions were even formed while on the move.