Kamikaze Confederate HighlandersÂ…

[Micah Goodman]

This post contains an easy way to defeat the AI…. If you don’t want to know them then stop reading.








































I posted in another thread on the Confederate AI tendency to rush north in an overly aggressive attempt to crush the union. The problem is the Confederates do not attempt to take Washington but instead focus on Pittsburgh or Harrisburg. In my last game I surrounded and destroyed four separate Confederate armies totaling approximately 70,000 men by the spring of 63. The first group was destroyed taking Pittsburgh in early 62. The second was a group of 5,000 militia in late 62 again trying to take Pittsburgh. The other two well, at least one occurred in early 63 when a Confederate army of approximately 25,000 men as they skipped across Kentucky and took Louisville. This was at a time when the Confederates held no territory north of Nashville. And by no territory I mean no provinces in all of Kentucky were under military control by the Confederates. With the loss of those troops and my appointment of Grant to the Army of the Potomac in early 63 and the quick capture of Richmond because the Confederates were obsessed with having an army prepped and ready to take Pittsburgh effectively ended the game in 63.

For clarification I had the AI aggression set to low and AI fog of war set to off, on the recommendation of another forum member. Here is what I believe the fundamental reason for this tendency of the AI to rush north to its death; the AI is programmed to look for Victory points. The AI see’s Pittsburgh, a healthy vp city guarded by only 6,000 militia troops and a city garrison and it says to itself… go there… take points. It then marches north disregarding military control of territory, distance from supply zones and more importantly it disregards upcoming seasons. So, the AI builds forces in the spring rushes north in the fall and freezes to death in the winter. Also, the AI obsesses over captured coastal fortifications, throughout almost the entire game the AI parked a force containing approximately 500 strength points off of Ft. Pickens Florida, (despite a huge US fleet of over 20 ships stationed around the Fort) despite New Orleans being captured, the AOP capturing Richmond and virtually no Confederate forces in the east for the rest of the game.

Please bear in mind, I mention this in the hope that the AI will be addressed in a future patch or more realistically in AGEod’s Civil War 2. My strategy focuses on destroying the enemy army. Territory is only a secondary objective, as I believe that without an army there is no chance to protect the territory. Some reading this post are probably saying, well crank up the AI aggressiveness and the game will be harder to beat. While this is true but I am looking for a reasonable recreation of the Civil War and if the Confederates are even more aggressive than they are now then the game will be even less of a representation of the war than it is now. With the way the game model is (accurately) set up to favor the defender the Confederates lose one of the south’s equalizers, that of defensive firepower. As I test, I played with the same settings as the south and took Washington twice and ended the game before the end of 61 because the North failed to adequately defend Washington because the Union AI is obsessed with North Western Virginia. Again, because it sees easy victory point cities and rushes in while I go skipping into D.C. Though I admit it is nice to prop my muddy feat on Lincoln’s desk as I smoke a stogie as the blue bellies flee from Southern might.

I don’t think people realize how important unobstructed supply lines are in this game. I didn’t until after about my fifth test game when I realized that my armies did not suffer attrition as badly if they occupied cities that could trace military control back to a supply source. If your armies occupy a city that can trace an unbroken line of controlled territory back to a supply source your armies will far less on your wagon trains in the winter and will suffer far less attrition. I mention the supply issue because the AI doesn’t seem to understand this. If it launches large invasions across several territories it can’t trace supply and so the army runs through supplies quickly and will typically only have one or two good fights in it before it collapses. If you station even relatively weak forces in the enemy’s line of retreat, can then send another force to engage and crush the enemy. The AI seems to have no concept of supply lines or how to protect them. So, in conclusion, the AI needs some more muscle for the next release. It needs to ‘see’ enemy forces better and more importantly it needs to 'see' it’s supply lines or lack thereof.

[Gray_Lensman]

deleted

[Micah Goodman]

I thought the AI fog of war advuce was suspect, but I have had similar issues with it turned on as well. I typically do not heavily garrison cities other than Washington D.C. and that might be the issue as well, but playing as the Union I use an offensive strategy overall and use my troops for the front line rather than the hinterland. The fort AI issue I know for a fact happens regardless of AI fog of war settings. The AI obseses over Ft Pickens and in all of my games goes crazy trying to retake them. Regardless, the AI does launch large deep attacks into enemy territory without having clear supply lines. And because of game mechanics, modeled correctly I believe, the AI starves itself to death.

[TheDoctorKing]

I think the basic problem is the fact that the CSA doesn't really pay any price for invading the north. Human players do the same thing, though they are better at looking out for their supply lines. Why did the CSA historically refrain from invading the north (real north, that is, non-slave states) except for a few brief raids and Lee's one major invasion of Pennsylvania that got about 20 miles north of the Mason-Dixon line? The reason is that they would have paid a very high political price. Certainly if the war had started with a southern invasion of the north (as it frequently does in this game) there would have been much less sympathy for the southern cause in the north and in Europe. Northern voters would have had no difficulty electing pro-war candidates in 1862 and 1864. Nobody would have complained about a draft. Volunteers would have turned out with great enthusiasm. And any hope of European intervention would have faded if the war was perceived as a southern attempt to seize power in the country as a whole. I wish there was some way to create an event that would reflect this. Instead of a few free regiments, the USA should get dozens or hundreds of conscript points, a big national morale boost, and a big decrease in the foreign intervention level every time significant CSA forces (more than a couple of regiments of cavalry) enter a non-slave state in the USA (Kentucky, Maryland, and Missouri are OK).

[Jarkko]

DoctorKing, nice to see somebody feels exactly the way I do about that silliness When I did bring up the matter here I was asked to make a mod with such events if it so much bothers me.

And I seriously tried, I did.

Sadly I am a complete idiot also regarding modding and events All I was successfull with was making the game crash to desk-top and/or locking up the computer so that it needed a cold-boot

Maybe AGEOD someday releases AACW2 and one could hope for seeing in the campaign game too strategies based a bit more on historical options. This excellent game for sure would in my opinion deserve that

[Micah Goodman]

The problem with writing an event is that south will be punished for invading the north if the AI is not changed then the south invades the north and get whatever punishment is written into the game. The Confederate AI seems to disregard supply lines and weather and only 'sees' force levels in corresponding enemy cities. If it sees a weak spot it rushes north in a haphazard manner.

Last night using the 1.12 patch I ran a test game. I played as the north. By November 61 the situation was this,

In the east... the Confederates had taken Grafton West Virginia and held it with a strong (approximate 600 strength point stack under Jackson) the rest of the Confederate army in the east was being besieged by the AoP under McClellan in the town just south of Harpers Ferry. I was fearful that the Grafton force would march north and take a lightly defended Pittsburgh

In the Kentucky-Ohio area... two turns prior I was driven from Bowling Green Kentucky by a superior enemy force and chased to Henderson Kentucky. I managed to send Grant down from Louisville with a force that once combined with the Bowling Green force roughly equaled the Confederate force. I was sweating buckets. I face imminent disaster in Kentucky. Also a 200 point force was besieging Elizabeth Kentucky. This force had been detached from Jackson in Grafton WV. Waiting for the rebel hammer to fall I advanced to the next turn. This was the only move that I thought was silly, but not fatally so.

At this point I was pleased with how the game was going as it seemed that the AI was being aggressive but not overly stupid. And then it happened, the Confederate force facing Grant split with half remaining in Henderson and the other half under Johnston (the good one not the poor one) marched north to Hard Kentucky, (where Munfordville is located), I ordered Grant to attack the force in Henderson and order most of my strength to move from Frankfort to Louisville.

In an attempt to figure out what the AI was thinking I loaded the AI saved game file I noticed that the force under Johnston was moving to Frankfort so I reloaded the game and cancelled the move from Frankfort to Louisville. When I reloaded the game and advanced the turn the force under Johnston had instead moved to besiege Cincinnati. The rebels in Henderson were pushed back by Grant and the enemy suffered moderate losses. The force I sent to relieve the siege at Elizabeth engaged and captured Confederate siege artillery because the rest of the Confederate force had died do to attrition apparently.

At this point I could not play further because I had a corrupted save game and decided to reinstall the game with the new patch. Again, the AI is to preoccupied with grand sweeping offensives. Most of which failed because of the worsening weather and lack of supply lines. If the Confederates had concentrated on Bowling Green, held it and built up it's supply lines I would have faced a stiff fight to retake it instead of waltzing in with a weak militia force under Buell the turn after the rebs took it.

The AI issue is one thing, but the other is rivers. Rivers are modeled well when it comes to transport but poorly when it comes to moving across them. It is unrealistic to believe that a force of 12,000 men under Johnston could have moved several hundred miles across hostile territory and constructed enough boats to cross the Ohio River in hostile territory, cross the river, and lay siege to Cincinnati in 14 days. If rivers were an adequate barrier to movement that might help the AI issue as well.

I don’t mean to sound overly critical of the game, this is THE BEST Civil War game I have ever played and I realize that a major AI rework is probably not in the works but hopefully it will be taken in consideration for any future versions of the game.

[Coffee Sergeant]

I'm still looking for a way to *not* easily defeat the AI.

[TheDoctorKing]

The problem with writing an event is that south will be punished for invading the north if the AI is not changed then the south invades the north and get whatever punishment is written into the game. The Confederate AI seems to disregard supply lines and weather and only 'sees' force levels in corresponding enemy cities. If it sees a weak spot it rushes north in a haphazard manner.

Exactly: the CSA AI needs to be "aware" of the risks, and to avoid them. I have no idea how to do this as far as coding is concerned.

The truth is, I don't really like to play against the AI. I play the AI just to figure out how things will work and learn the system but "real" games I play against human opponents. But even human opponents make these enormous invasions as the CSA. The reason is that they _work_. The CSA can put a big hurt on the USA, take cities, push their VP and NM totals up, make foreign intervention more likely, etc., by invading the north. I'm arguing historically that the result would have been the opposite - southern public opinion would have been dubious at best (many of Lee's troops refused to leave Virginia even to enter Maryland in the Antietam campaign), northern public opinion would have been outraged, and foreign opinion would have become _less_ favorable to the south if they had invaded the north with anything like an intention to capture territory. This was not a coup by southern interests to take over the government of the US but a separatist movement to divide off a part of the country. Lee's invasion of Pennsylvania in the Gettysburg campaign was controversial in the south, and it was clearly just a raid, although a big one. He had no intention of capturing territory in Pennsylvania although he was thinking about capturing Baltimore and Washington (both located in slave territories).

I'd say that you could use an approach in modding like the Struggle For A Vast Future mod uses for Kentucky. Make the free states non-enterable for the CSA until they buy a "unit" entitled "invasion of the north", then remove the restrictions for a limited period of time. The "unit" costs NM and conscript companies (representing southern public disaffection and desertions from the CSA army) and also gives NM and conscript companies to the north and reduces foreign intervention. Then you have an event that fires if the CSA forces win a battle in the north or captures an objective city in the north that gives them extra NM above what they ordinarily would have won, and perhaps a foreign intervention bonus.

[Jarkko]

I'd say that you could use an approach in modding like the Struggle For A Vast Future mod uses for Kentucky. Make the free states non-enterable for the CSA until they buy a "unit" entitled "invasion of the north", then remove the restrictions for a limited period of time. The "unit" costs NM and conscript companies (representing southern public disaffection and desertions from the CSA army) and also gives NM and conscript companies to the north and reduces foreign intervention. Then you have an event that fires if the CSA forces win a battle in the north or captures an objective city in the north that gives them extra NM above what they ordinarily would have won, and perhaps a foreign intervention bonus.

In my opinion it should be a political decission (if South wants to attack Kentucky or North early on). Attacking Kentucky or the North should be allowed only after taking that decission. If the decission is taken, National Morale would drop, conscripts would be lost and foreign intervention chance would drop by *a lot* (and of course CSA would now be free to attack).

The decission should in fact have two "variants":

1) "Secure Kentucky". South is allowed to attack Kentucky freely, but the NM , conscription and FI hits are not that heavy. The costs should reduce as the time passes, reaching zero at the end of 1861.

2) "All out war". South is allowed to take any and all actions (and thus is seen as the aggressor, as it attacks cities in north, pillaging and ravagin what it can hold etc) anywhere on the map, and effects on NM, conscripts and FI should be severe. Again the costs should reduce as time passes.

[Micah Goodman]

Most of the issues raised in the above two posts are valid and well thought out. But to implement them the AI has to be fixed. Until it is writing any event or speculating on changing the game is pointless. If you write the event's and the CSA invades northward time and time again it will further cripple the South and in the long run make it even easier to defeat the CSA than it already is. The AI for both sides will rush headlong into the breech over and over and most player's can surround the AI forces, and kill huge stacks of them as they rush blindly north (or south). Any event you write will trigger because the AI will invade the north. The AI won't see the event because it isn't trained to see it.

If you play against a human opponenet you can impliment a house rule of no Confederate forces may enter any non border states until say 63 and no Confederate forces may enter Maryland until 62. Having house rules with another human oppenent is easy. Writing AI is not.

[Rafiki]

Even if the points you make about the AI are valid, I have to disagree with

Until it is writing any event or speculating on changing the game is pointless.

AACW has an active crowd of PBEM'ers. I'm sure they even now could appreciate some events dealing with these issues

[bk6583]

Well I felt I had to chime in as well. I finally had to stop playing AACW for the reasons cited above. In all of the games I have played against Athena, regardless of the patch or mod or AI settings, the South always launched invasions of the North primarily in or through PA and its environs (heck, Bobby Lee was up around Erie in one game), was always cornered and destroyed, resulting in victory for the North in 1863. I've been hovering around this forum in the hope that at some point this would be alleviated, but alas, it appears not. It also appears that there are some smart people on this forum (I mean that sincerely) that do not believe this is a problem, so I suspect this 'go North young man' tendency of the Southern AI will never really be addressed.

[sval06]

Well I felt I had to chime in as well. I finally had to stop playing AACW for the reasons cited above. In all of the games I have played against Athena, regardless of the patch or mod or AI settings, the South always launched invasions of the North primarily in or through PA and its environs (heck, Bobby Lee was up around Erie in one game), was always cornered and destroyed, resulting in victory for the North in 1863. I've been hovering around this forum in the hope that at some point this would be alleviated, but alas, it appears not. It also appears that there are some smart people on this forum (I mean that sincerely) that do not believe this is a problem, so I suspect this 'go North young man' tendency of the Southern AI will never really be addressed.

So, you should play PBEM....

Harder but very funny

[soundoff]

So, you should play PBEM....

Harder but very funny

+1 Against a human player the Grand Campaign really rocks

[Jabberwock]

I've been hovering around this forum in the hope that at some point this would be alleviated, but alas, it appears not. It also appears that there are some smart people on this forum (I mean that sincerely) that do not believe this is a problem, so I suspect this 'go North young man' tendency of the Southern AI will never really be addressed.

I still have faith that this will be fixed, the tool needed to do it was only [relatively] recently made available to modders.

[Clovis]

I'm working on it. The current release has some results.

[bigus]

I'm working on it. The current release has some results.

Yes the AI.ChgLocalInterest command is nice (if this is the command your talking about) but it would be better I think to have the AI coded to be more aware of it's supply. With the new commands comes the problem of changing all the scenarios by adding events (to all the campaign games). The events would have to be dynamic and change almost all the time depending on the situation.
I thought there was a patch at one time where Pocus had coded the AI to be more aware of supply during harsh weather. I wonder if it's possible to tweak the code so the priorities are more like Supply first, then Enemy forces and finally Production centres (or objectives). Or something similar to this.

[TheDoctorKing]

Human players can be more careful of their supply and not get their armies wiped out in their foolish invasions of the north but it is still a "gamey" tactic in that it really wasn't an option for the historical commanders, at least not as frequently as it occurs in the game. Certainly, no Confederate army would ever have gotten close to Erie PA or NYC. The actual game (as opposed to just the AI) should be changed to make an invasion of the north a much less attractive option for the CSA. Otherwise CSA players in PBEM games will go on doing this.

[Gray_Lensman]

deleted

[Micah Goodman]

Gray,
you bring up some valid points about a Confederate invasion or possible invasion of the north was not out of the realm of possibility but for the political reasons you stated unlikely until at least mid 1862.

On a game level almost any invasion of the north in 1861 by the AI is doomed to failure. First there are no supply centers between Harpers Ferry and Pittsburgh. This makes the supply lines from Virginia to Pittsburgh very tenious.

Let's take a hypothetical and say that the AI was smart enough to have military controll from Harpers Ferry to Pittsburgh and captured Pittsburgh in 1861. I as a human player have just won the war. How? I can leave the fixed forces in Washington in place march on Harpers Ferry, capture it and then wait for the rebel army to either starve to deat or march back south in a weakened state for me to destroy.

I understand that many people play PBEM and I would love to do the same. However, my schedule is so eratic that it would be unfair for me to start a game that would be erratic at best. So, I am left with the AI. If I could write code I would be more than glad to help mod the game. Sadly, my limited skills were taxed just by writting events, much less AI for HoI2... I think this game is much more complex.

From: The DoctorKing

Human players can be more careful of their supply and not get their armies wiped out in their foolish invasions of the north but it is still a "gamey" tactic in that it really wasn't an option for the historical commanders, at least not as frequently as it occurs in the game. Certainly, no Confederate army would ever have gotten close to Erie PA or NYC. The actual game (as opposed to just the AI) should be changed to make an invasion of the north a much less attractive option for the CSA. Otherwise CSA players in PBEM games will go on doing this.

I agree, Confederate invasions should be much more difficult. Making movements across rivers should be more difficult than it is in game for one, and secondly there should be an event that fires in 1861 until at least the spring of 62 that punishes the south if it invades the north. Oh, also, the force needed to blocaked a river should be much lower than it is currently. One river gunboat or ironclad would prevent any Confederate force from maintaing bridges or boats across a major river. But, as I stated earlier, human players can mutually agree on house rules stating no invasions by the south over the Ohio River, no invasion of Maryland until the summer of 62, ect. Problem solved. If you write an event before you fix the AI, you will compleately wreck any game for a player playing against the AI. Eventually, it would be nice to have an event inserted that does this but IMHO not until after the AI is modified.

Not knowing how the AI is written I would suggest the AI be written with these peramters,
Agression levels set by date, and season.
Range of nearest enemy army and military control of the area.
Enemy control of certain cites. IE if the North controls New Orleans for example, ths southern AI should want to have a nearby army of at least a cetain percentage in strength of the enemy army.
It should ignore most captured Forts along the cost, as long as the enemy keeps only a minimal enemy force present in the fort and instead beef up nearby coastal cities, as was done historically.

These of course are just suggestions to help shape the discussion and are not nessecarily the best solutions to the problem.

[Gray_Lensman]

deleted

[Clovis]

Yes the AI.ChgLocalInterest command is nice (if this is the command your talking about) but it would be better I think to have the AI coded to be more aware of it's supply. With the new commands comes the problem of changing all the scenarios by adding events (to all the campaign games). The events would have to be dynamic and change almost all the time depending on the situation.
I thought there was a patch at one time where Pocus had coded the AI to be more aware of supply during harsh weather. I wonder if it's possible to tweak the code so the priorities are more like Supply first, then Enemy forces and finally Production centres (or objectives). Or something similar to this.

Partially. But I've got since a rather long time a large diminution of CSA raids on North by changing a bit the theaters, which are the geographical cadre for AI thinking.

[Gray_Lensman]

deleted

[Micah Goodman]

Micah Goodman:

All valid points concerning the state of the current AI behavior. My original statement was to point out the fact that aside from Jefferson Davis defensive stance, there would not have been that much opposition to a move of this sort. He was definitely hoping to secure foreign intervention, so in-game, that should be the political price to be paid. (No hope of foreign intervention and possibly some reduction in the amount of materials received through blockade runners)

Regarding AI programming, currently it is super complicated because parts of it are in the game engine (executable) itself, and these are accessed only thru some very complex event manipulations. The game engine part of the AI has some problems that need to be addressed first such as the perceived over-value the AI places on some target regions. Whether this is caused by too much value being placed on "production" capabilities, or VPs, or a combination of these items and others, I'm not sure just yet, but until these "value" decisions are brought more in line, it seems to be too much of a challenge to overcome the AI choices.

edit> Game engine AI changes have to be done by the programmer/developers and is not something I can do. My abilities currently lie within the realm of overseeing and adjusting the various database files that support the game engine externally.

Gray Lensman,
I agree with punishing the south for an early invasion of the north. I figured that this type of AI work would requrie a major rework and that is why I figured that these changes would probably only occur with a new release of the game. It would be nice if I was wrong but I do not expect it to happen. Overall, I think the game is awsome with that one exeption. I just wanted the programers to be aware of some concerns I had with game. I purchased the game about a month and a half ago and have been pleasently suprised with the pace of improvments and hope it continues.

[Daxil]

I think the basic problem is the fact that the CSA doesn't really pay any price for invading the north. Human players do the same thing, though they are better at looking out for their supply lines. Why did the CSA historically refrain from invading the north (real north, that is, non-slave states) except for a few brief raids and Lee's one major invasion of Pennsylvania that got about 20 miles north of the Mason-Dixon line? The reason is that they would have paid a very high political price. Certainly if the war had started with a southern invasion of the north (as it frequently does in this game) there would have been much less sympathy for the southern cause in the north and in Europe. Northern voters would have had no difficulty electing pro-war candidates in 1862 and 1864. Nobody would have complained about a draft. Volunteers would have turned out with great enthusiasm. And any hope of European intervention would have faded if the war was perceived as a southern attempt to seize power in the country as a whole. I wish there was some way to create an event that would reflect this. Instead of a few free regiments, the USA should get dozens or hundreds of conscript points, a big national morale boost, and a big decrease in the foreign intervention level every time significant CSA forces (more than a couple of regiments of cavalry) enter a non-slave state in the USA (Kentucky, Maryland, and Missouri are OK).

There also needs to be some functionality to penalize Union players who do not invade the South in the East. As is, there is a 10NM hit if you dont do something by 1862 June (?) and then nothing. Therefore it is a viable strategy for the USA player to squat on his main cities and wait till he has appropriate leaders. This would have resulted in extreme morale loss in the North imo. The best way to remedy it is by masking and randomizing leader abilities but since so many players want authenticity, a morale hit is the alternate method. The question is how to implement it. You'd probably need another layer imo.. maybe a diplomatic layer with specific objectives, masked from your opponent- "Lincoln demands an attack by such and such date" or "the people of Pennsylvania demand your protection. Recapture Scranton by turn? or suffer a morale hit."

I know it cant be implemented in 1, but maybe AACW 2.

[Gray_Lensman]

deleted

[oldspec4]

Very

[Clovis]

I'd say that you could use an approach in modding like the Struggle For A Vast Future mod uses for Kentucky. Make the free states non-enterable for the CSA until they buy a "unit" entitled "invasion of the north", then remove the restrictions for a limited period of time. The "unit" costs NM and conscript companies (representing southern public disaffection and desertions from the CSA army) and also gives NM and conscript companies to the north and reduces foreign intervention. Then you have an event that fires if the CSA forces win a battle in the north or captures an objective city in the north that gives them extra NM above what they ordinarily would have won, and perhaps a foreign intervention bonus.

The real problem is AI. For my approach of KY question, I've used many tricks to overcome the total impossibility for the AI to know these "new rules". AI is events blinded. I've used random factor, locked some units to give CSA AI a chance to react to KY evebts when they fire. It is giving good results because the situation is fairly limited in scope and predictable. On a more open ended context, I fear events would severely challenge AI...

[TheDoctorKing]

Great idea! Can it be extended to the northern states as well?

And in response to the earlier post by Daxil (?) I think it would be very appropriate to penalize the north for not being aggressive enough. I don't think the invasion has to come through the east, though. A strategy I have used as the USA is to stand on the defensive in northern Virginia while conducting amphibious invasions along the coast, pushing down the Mississippi, and sending significant forces to the Trans-Miss to at least threaten Arkansas. It worked pretty well until the huge CSA army in the east crushed my defenses in Alexandria and took DC... But the north should get credit for any attacks, not just in the eastern sector.

[Gray_Lensman]

deleted