RE: The AI CSA in the West (Full Version)

All Forums >> [Current Games From Matrix.] >> [American Civil War] >> Forge of Freedom: The American Civil War 1861-1865



Message


Gil R. -> RE: The AI CSA in the West (6/22/2009 4:53:51 AM)

Janh,
I meant to respond, but forgot. See below.

quote:

ORIGINAL: janh

Hi everyone,
I hate to make any statement given such a heated debate already.  jsc991 seems to be expecting what I did when I bought the game -- a simulation.  I was hoping for something that would be quite detailed, sort of like WITP (War in the Pacific, or even WITP-AE) is to older (or this other hex-based) Pacific strategy games.  But I learned that FOF is more a game, intended to be enjoyed.  There is not much micromanaging (like assigning Brigade leader, which would have been cool to get more of a personal relationship to the armies you build up and lead into the fight), which means you can play it pretty easily and a game is quite fast. 

However, some things I liked a lot, like the economy part or the maps are very nice (I played the last patch before the April one, FYI).  Other things I missed: BG leader assignments, smaller provinces (counties!) on the strategic map so you and move your army in different columns to your target (like shenandoah east and west forks routes, for instance).  Or have smaller provinces in general (Fredericksburg split in FB, Wilderness, Gordonsville, Centreville etc).  Also the detailed battle maps could be nicer, have a more realistic road pattern and other strategic features that AI would consider more in forming its lines; or some historic maps for certain provinces. 

Being able to assign brigade commanders permanently is at the top of our list of things we would like to do if there's a FOF expansion. Not an easy change to code, which is why we haven't done it for a patch, but we do think it would be a great addition.

I also agree that, ideally, we would redo the map at some points if we were to come out with a new and improved FOF. And if we did that, reworking some of the Virginia provinces would be number one on the list. Having smaller provinces overall, though, would not be something we'd want to do, since that would completely change the game and require massive amount of new programming and testing.

Historic maps for provinces is something suggested before. It would certainly be a great addition, though not an easy one. We're keeping the idea in mind, though.


Also I found that AI in detailed battle didn't seem to keep units of division somehow formed, but mixed brigades of different formations (sure historically that happened too, but in most cases brigades of a division stayed at least somewhat together).  One thing that annoyed me a bit about detailed combat was the minimum unit size, below which they rout.  Many BG's historically were small in battles, like 500-800 men units seens for instance for Lees tired army at Antietam. 

Possibly things we'll change at some point. Personally, I'd rather see the routing occur when a percentage of men are lost rather than a fixed minimum being reached. Unfortunately, I only reached this opinion long after the game had been released!

Also, artillery but in a few cases historically was concentrated into formations of 3000 men (see Alexander at GB). Initially most artillery was brigade artillery, but the Confed (initially Lee) reorganized it in 1862 into battalions of ca 300 men and 20-30 guns, which then were employed battery wise. The Union later did the same, and BG artillery was not so common afterwards anymore.  These heavy 3000 men artillery BGs did way to much effect in concentration of fire and damage than in history I felt (quite a few battle against a nurmerically far superior AI I won by having concentrated all my economy on building ART and getting the indirect fire attribute -- I kept shifting those 10-15 parrott and 24p howitzer BGs by train forth and back east and west; if you look at historical sources, artillery unless using canister at close range, created more noise than casualties, contrary to what many people want to believe. Just read up, Official Records of the War of Rebellion is a really good source and available at the Library of Congress)!

Since the game is at the brigade-level we decided to go with artillery brigades, even though that's not how things were done, especially at the beginning of the war. In retrospect, it would have been better to limit all artillery units and require an "Artillery Brigade" upgrade that would enable them to grow in size and firepower. But this is an idea we had long after release.

As for ranges, you're right that less damage was done, but guns firing at a distance could still have a major effect on the battle. Or just the fear of them could, since for example Little Round Top's value was largely its usefulness for placing artillery where it could bombard troops as far as a mile away.


However, I fixed both the BG issues with a Hex editor, and found the 3 or 4 places that I needed to lower the thresholds.  That worked will with detailed combat, and I could have a reasonable artillery battalion for each div and one for corps reserve.  However, AI never managed to organize so many arty units and organize the containers properly -- was probably just not designed to do this.  Would be nice if that could be added as an option in the next patch.

I'm not sure just what you're asking for.

The strategic AI of the game seemed to be quite sound to me, but surely not capable of defeating a human without bonuses.   Well, that is the sad story for most games though -- guess AI is the thing potential customers least see -- fancy graphics, maps, millions of weapons and stuff seem to sell these days.   Though I like Take Command 2 Manassas a lot with its detail, historic realism and reasonably good AI, it seems also the next part, Gettysburg will be "just another map" (sorry for the sidetrack) -- like Arma and Arma 2 will be for me to Operation Flashpoint.  Maybe one day I will buy another game of these series -- maybe Arma 13 and TC12.

Anyway, before I made my small hex mods, I usually played CSA and the Union AI would give me some good times.  I didn't like to give it unrealistic advantages in the detailed combat, but economic bonuses were ok for me - after all, I wanted to have some realism left.   One thing that annoyed me was that I tried to style my ANV along the lines of the historic OOBs, and that meant something like 50+ BGs in 10 divisions or so (yeah, I did change the max limits for the containers).  Yeah, huge, but I didn't replenish many BGs so altogther it were usually some 7-80000 men.  I kept them split up, protecting WV, Shenandoah, Fredericksburg, the Peninsula and Norfolk, typically.  Only in case of imminent battle or offensive I concentrated.  
Although I did something similar in the West, having also there historic troop strengths (~100.000 men across from far West, Tennessee, and the South), by mid/end 1862 AI tended to send its large formations to the East, accumulating often more than 300.000 men in W. DC. and PA.  I tuned around the parameters that can be modded (there must be a thread somewhere that MM guys gave me a hint on how AI computes its behavior), but it didn't help too much. 
Maybe AI was confused by my many, though small BGs in Virgina area. That that got worse after modding the BG limits, and tuning down the camp production rate, is probably not surprising and surely my issue.  Well, anyway, I not always saw this.

Sometimes AI did pretty well in the West, particulary if its armies in the east stayed idle and didn't loose too many times against me.  AI seemed to concentrate in the east if he lost several major battles there, or if he bled down his forces in the West.   I guess Strategic AI could be tuned down a little more to be less agressive and focus a bit more on defending its own cities, which I felt it wasn't paying too much attention to in the West.   Maybe that would help.

Now don't get the impression that I didn't like this game: I got many fun hours out of this game, and can say it is pretty nice.  I still play it once in a while, and would suggest everyone who doesn't absolute require the historical realism and detail of a WITP-AE or TC2M to buy it.   But hope for a FOF2 that will adress some things I would hope for and turn this game more into a historic simulation.  Maybe, maybe not.
Best regards,
Jan

Thanks for all your feedback. We definitely would like to do a FOF expansion, though I don't see starting work on it in the coming months. (We have multiple still-secret projects that are top priorities at the moment.) But if we do one, at least some of the changes you'd like to see should be in there.
















jscott991 -> RE: The AI CSA in the West (6/26/2009 2:58:44 PM)

The movement east on Captain with the above settings is even worse than on First Sergeant, because the AI's numbers in the east become greater (he has more total men so he has even more troops above any reasonable needs in Richmond).  I won't post another round of screenshots, because no one seems to care.

The July 1861 scenario is bizarre.  The huge Union economy is strange (I can't ever spend my labor or iron because there's just too much of it) and the Union's manpower advantage over the CSA is even larger than in November 1861.  I heavily outnumber the CSA by early 1862.  The AI, however, plays a different style.  He is hyper-aggressive with small containers.  However, in all three of my July 1861 starts, the AI's army containers all end up in the east fairly quickly.  The migration of actual troops isn't as pronounced, but the presence of those containers in the east shows that the AI bias is still present.

Of course, this is only a problem I'm seeing.  I have the one copy of FoF where the AI over-favors Richmond.




Erik Rutins -> RE: The AI CSA in the West (6/29/2009 1:28:38 PM)

jscott991,

If you have one, could you e-mail me a save file from not long before the AI does its shift to the East? If we can see that happening, we can see why it's making that decision. I suspect one of your other settings is exacerbating things, but I haven't had time to do any independent research because of my workload at the moment. All I can say is that with the settings I played at, this problem was pretty much gone by 1.10.10. Obviously it's still there for you but we need to figure out why.

Regards,

- Erik




jscott991 -> RE: The AI CSA in the West (6/30/2009 1:16:17 AM)

The save game you are asking for might be tough to find. I have a lot of savegames where the AI migration is complete (the screenshots show this), but my autosaves have all been replaced with July 1861 games or a restart trying advanced supply.




jscott991 -> RE: The AI CSA in the West (7/14/2009 12:59:47 AM)

Like an old soldier, "concern" over this issue has just faded away.

If it weren't for the terrible looking maps and graphics, I'd try another Civil War game I think. Sadly this game has me hooked a bit, despite the fact that its about as much of a Civil War game as Chess is a simulation of medieval combat.




Mr. Yuck -> RE: The AI CSA in the West (7/14/2009 2:04:17 AM)

I hear that the AI in chess might be up to snuff.




Gil R. -> RE: The AI CSA in the West (7/14/2009 2:26:23 AM)


quote:

ORIGINAL: jscott991

Like an old soldier, "concern" over this issue has just faded away.

If it weren't for the terrible looking maps and graphics, I'd try another Civil War game I think. Sadly this game has me hooked a bit, despite the fact that its about as much of a Civil War game as Chess is a simulation of medieval combat.



Concern isn't fading, and has never faded. I have previously told you that we are considering the matter internally; I have also said that we have no plans for a FOF patch anytime before autumn, simply because we have other projects (two of them with contractual deadlines). Nothing has changed. Just because I have expressed doubt regarding how serious a flaw this is does not mean that we are not taking the issue seriously -- we don't want any flaws, if we can help it.




jscott991 -> RE: The AI CSA in the West (7/14/2009 3:54:32 AM)

You're right.

Armies of Northern Virginia in excess of 100,000 men on EVERY difficulty should be the default state of any Civil War game.

And this is leaving aside mention of the migration issue, which makes a total mockery of the AI's ability to play a competitive game. Let's just assume that migration NEVER occurs. That still leaves the huge size of the CSA army in the east even BEFORE western troops flock to its side.

Does anyone NOT see ANVs of this size by the end of 1863 (usually by the end of 1862) in any game playing as the Union using quick combat/instant combat on ANY difficulty level?

I've tried Captain, Major, all the sergeants and the AI will always have an army of this size. How is that the Civil War? How does routine 1:1 odds (or worse) for the Union in the east not qualify as a serious flaw?




jscott991 -> RE: The AI CSA in the West (7/14/2009 3:55:53 AM)


quote:

ORIGINAL: Mr. Yuck

I hear that the AI in chess might be up to snuff.



As long as it abandons half the board to defend some random square, it's good enough for me.




cesteman -> RE: The AI CSA in the West (7/14/2009 4:20:42 PM)

Now I think your getting way too sarcastic. Gil said the problem is being looked into. You have to be patient.




jscott991 -> RE: The AI CSA in the West (7/15/2009 12:28:46 AM)

Gil always prefaces or postscrips his remarks by saying its not a serious problem.

It is a serious problem. Erik R. has mentioned in his posts that AI turtling was looked into before, so I get two totally different vibes.

Playing as the Union and not using detailed combat leaves you with two choices:

1. You can play a difficulty level where the Union will enjoy a manpower advantage in line with the historical balance of power and win the game by taking advantage of the migration issue. The AI will migrate in every First Sergeant game. It will migrate on Captain as well. And it will pack the east whether it migrates or not on all the difficulties in between (I just wanted to see sample ANVs, so I didn't play these games beyond mid-1862) producing a near 1:1 situation in the east, even if the Army of the Potomac numbers over 100,000 men. Factor in the 27 brigade limit on QC/IC and Lee will never be seriously outmanned or even outmanned at all.

2. You can play on a high difficulty level and, maybe, not face the migration issue. However, the war you're fighting will bear absolutely no resemblance to the Civil War, because the CSA's combat boost and economic levels will produce an unrealistic balance of forces and require ludicrous odds to win a QC/IC (3-1 is no guarantee on these levels, especially during the first 2.5 years of the war).

I am convinced that this game was primarily tested playing as the CSA, on difficulty levels well above captain, using almost exclusively detailed combat. That would explain a great many things (the insane quick combat results because of rallies, the migration issue, the USA v. CSA balance of forces, the brigade surrender issue, etc.).




Mr. Yuck -> RE: The AI CSA in the West (7/15/2009 1:15:34 AM)

JScott,

My only thought at this point is that you've found some perfect strategy of your own that makes the AI turtle. What are your early goals?

The game was absolutely not playtested only as the CSA.

I'm unclear-how did the July-start work out for you? Have you noted what CSA Nat'l Will is at when they start turtling? I've yet to see this, and I've  been trying since I came back to this game and saw this thread.






jscott991 -> RE: The AI CSA in the West (7/15/2009 3:28:49 PM)

There's no trick to my "perfect" strategy.

It boils down to "don't attack in the east if the CSA is in Fredericksburg."  Build your troops up intelligently.  Move post to post in the west.  Focus on getting your main AoP to 36 brigades immediately and then creating a second army in the east that should be 15 or so brigades as soon as possible.  Couple this with four western independent commands (a corps in Missouri usually under Hunter to protect Topeka and Jefferson City, plus eventually take Little Rock; a corps in Kentucky, usually under Buell; a main army under Grant that will lead your sieges at Cairo, Ft. Henry, Donelson, Nashville, Memphis, and then down the Miss. River; a second army, usually under Halleck, operating in Kentucky and central Tennessee to cover Grant's sieges) and the AI will just give up.

The AI shifting at least one western army container east happens so fast (early to mid-1862), that I don't understand why people can't replicate these results in 20-30 minutes of play.  I can do it every time. And do people not notice that the ANV evacuates Fredericksburg to sit in Richmond on the first or second turn of every game? Isn't that a tip that something is off? Why would the South give up its impregnable defenses in Fredericksburg to sit in its capital?

The July start is not that much fun.  The Union economy is ENORMOUS.  The CSA is totally overmatched.  The AI is completely different in a lot of respects, as I mentioned above; he is more aggressive, pushing little containers all around the map to their death.  But he doesn't stand a chance.  I couldn't even spend all the resources I had as the North and my armies were much bigger than my November 1861 forces.




Erik Rutins -> RE: The AI CSA in the West (7/15/2009 5:31:50 PM)

Just FYI, we looked at your saves, we see the issue, we're discussing what can be done and when. Given that this is not typical for other players and that FOF was pretty much at the end of its support cycle, it's not really something that takes higher priority than the other projects the dev team is working on now. However, thanks to your reports and saves, I think we'll be able to do something about this in the long run, though ultimately that will depend on Eric.

With that said, making broad assumptions based on your personal experience is shaky ground. Look at how many responses you've gotten here from other players saying "I haven't see this". This isn't because everyone is playing the CSA. A lot of testers focused on the Union.

I personally played the Union primarily throughout testing and played primarily quick battles both solo and for PBEM (except when specifically testing detailed battles) because I didn't have time for a lot of detailed battles. I found the early manifestations of this, which by the time a couple of updates were out were completely resolved _in my games_. Obviously, some of this still remains and it's exacerbated by your settings and play style, which is what seems to make it more prevalent than it is for other players.

The upshot is that your typical FOF player isn't experiencing this and hasn't seen it. You see it all the time, we now have your saves and we've got it on the list of things to address when we get more FOF development time, which may not be very soon unfortunately.

Regards,

- Erik




jscott991 -> RE: The AI CSA in the West (7/15/2009 10:32:02 PM)

Please tell me the settings that are causing this and I'll change them.

Which settings should I choose that will cause the ANV to hover around 60,000 to 80,000 men instead of 100,000 to 160,000?

Which settings need to be on to cause the AI to stay in Fredericksburg instead of retreating to Richmond (and, again, does anyone not see this happen on the first or second turn and again after almost every major battle)?

Which settings should I use to cause the AI to group its western divisions into armies instead of using little containers that float around the map and ultimately end up in Richmond?

I've tried combos of most settings at this point. I don't think the settings cause this.

And if my playstyle (no suicide runs into Fredericksburg) is causing it, I'm baffled.




Anthropoid -> RE: The AI CSA in the West (7/16/2009 3:36:31 AM)

Have not played this game for a very long time, but might be feeling the urge to reinstall and patch up. Making me remember how challenging the AI can be sometimes . . .




cesteman -> RE: The AI CSA in the West (7/16/2009 6:47:46 PM)

Go for it Anthropoid! I stepped away from the game just because I was burned out. Well worth it.




Rekm41 -> RE: The AI CSA in the West (7/16/2009 8:02:02 PM)

I agree I cannot seem to put the game away either.1/2 way through 63 and the south is giving a good fight as the AI. Just started a PBEM game with cesteman and looking forward to that one.

Ron




steel god -> RE: The AI CSA in the West (7/17/2009 1:42:42 AM)

Not trying to be a smart ass or anything, but I would suggest that the one setting you need to adjust is the one where you select a local game over a PBEM game.  As a veteran war gamer with 30+ years playing, and numerous credited play tests under my belt I can say with confidence that no AI developed for any war game will ever be even half as good as a human opponent.  Find a living breathing opponent and put that AI play behind you and you'll never worry about such things again.

Respectfully.




jscott991 -> RE: The AI CSA in the West (8/1/2009 6:34:14 PM)


quote:

ORIGINAL: steel god

Not trying to be a smart ass or anything, but I would suggest that the one setting you need to adjust is the one where you select a local game over a PBEM game.  As a veteran war gamer with 30+ years playing, and numerous credited play tests under my belt I can say with confidence that no AI developed for any war game will ever be even half as good as a human opponent.  Find a living breathing opponent and put that AI play behind you and you'll never worry about such things again.

Respectfully.



I don't want the AI to be that good. I just want it to be semi-competent. Maintaining a western front is a pretty basic quirement for a Civil War game.




Longstreet_slith -> RE: The AI CSA in the West (8/3/2009 9:58:09 PM)

I agree with jscott991. I too have experienced what he has very clearly described in many posts. An adjustment so that the AI does not abandon the western front would be nice. Please consider doing this!

-Longstreet




vonRocko -> RE: The AI CSA in the West (8/4/2009 6:23:35 PM)

Why do people always say "play pbem" as an excuse for bad AIs'? When I got time to play,I do not have time to set up a pbem,and I don't want to play at the convenience of another person.What happens when the other player isn't around?I want to get more than one turn in.It is not possible to have another person at the ready any time I want to play.
These games need a decent AI. The fact is,the majority of players play against the AI. So like it or not, The AI is THE single most important issue in a wargame.




Randomizer -> RE: The AI CSA in the West (8/4/2009 8:32:45 PM)

Dumb question perhaps but here goes...
In the Data folder there are a couple of files of interest.

The ACWProvinces file Column K (in M$ Excel) is headed "Theater" and the following provinces are assigned a value of 5 out of a range of 0 to 5:

Shenandoah
Fredericksburg
Rappahanock
Petersburg
Norfolk
Cumberland
Annapolis
Potomac_River
James_River

Meanwhile the ACWConstants file has an entry (row 100 in M$ EXcel) reading as follows:
AI_OutOfTheaterMultiplierOnValuex100 15

Now is it merely a coincidence that the AI seems to gather in the provinces with highest "Theater" value?

Will admit to not having conducted any experiments editing either the "Theater" value or the "Out of Theater Multiplier on Value" since I tend to keep my Eastern Union armies fairly small and generally see continued Confederate AI presence in the West. Since I could find no text or information detailing what these values actually do, I figure that changing either or both of the above would likely have major unintended consequences.

Still I wonder if what jscott991 is seeing is in the Confederate AI is related to either or both of these values and so perhaps this AI behavior might be configurable by the player.

Perhaps one of the designers can enlighten us. Thanks in advance.

Best Regards




ericbabe -> RE: The AI CSA in the West (8/6/2009 6:47:21 PM)


quote:

ORIGINAL: Randomizer
Dumb question perhaps but here goes...
In the Data folder there are a couple of files of interest.

The ACWProvinces file Column K (in M$ Excel) is headed "Theater" and the following provinces are assigned a value of 5 out of a range of 0 to 5:

Shenandoah
Fredericksburg
Rappahanock
Petersburg
Norfolk
Cumberland
Annapolis
Potomac_River
James_River

Meanwhile the ACWConstants file has an entry (row 100 in M$ EXcel) reading as follows:
AI_OutOfTheaterMultiplierOnValuex100 15

Now is it merely a coincidence that the AI seems to gather in the provinces with highest "Theater" value?

Will admit to not having conducted any experiments editing either the "Theater" value or the "Out of Theater Multiplier on Value" since I tend to keep my Eastern Union armies fairly small and generally see continued Confederate AI presence in the West. Since I could find no text or information detailing what these values actually do, I figure that changing either or both of the above would likely have major unintended consequences.

Still I wonder if what jscott991 is seeing is in the Confederate AI is related to either or both of these values and so perhaps this AI behavior might be configurable by the player.

Perhaps one of the designers can enlighten us. Thanks in advance.


No, the theater number isn't quantitative of anything.




steel god -> RE: The AI CSA in the West (8/6/2009 8:00:31 PM)


quote:

ORIGINAL: vonRocko

Why do people always say "play pbem" as an excuse for bad AIs'? When I got time to play,I do not have time to set up a pbem,and I don't want to play at the convenience of another person.What happens when the other player isn't around?I want to get more than one turn in.It is not possible to have another person at the ready any time I want to play.
These games need a decent AI. The fact is,the majority of players play against the AI. So like it or not, The AI is THE single most important issue in a wargame.


I can't speak for others, but I can speak for me, and I always say try PBEM because when playing an AI you will always have to contend with it doing irrational and illogical things. Considering you are buying a $50 product, the AI is pretty darn good, but it will never be able to handle the infinite permutations of possible moves like even a very poor human opponent. So if just "playing" is all you are after, than AI play will certainly satisfy. But if you desire being challenged by an opponent that can think and plan as well as you, then PBEM or live on line play are your only real choices until someone figures out how to put a next generation AI into a $50 hobby game.

As I stated originally, I'm not trying to be a smart ass, I'm pointing out that it is unreasonable to expect a game's AI to behave as a human would, and if you want the experience of playing a human, than do so.

Regards.




hgilmer3 -> RE: The AI CSA in the West (8/7/2009 12:36:28 AM)

I have always seen pretty large armies in the east for the CSA.  I never really thought much about it because I'm not a person who has studied the Civil War for any length or period of time (although I have been getting more interested.)  It does seem of note that I read a brief on the Battle of Gettysburg and it didn't seem like they had very big armies.  Total casualties ended up being like 23K for Union and 25K for CSA and the total forces were something along the lines of 160K total troops both sides put together?

The AI does not handle the west well, though, especially once you build up enough troops.  I have only beaten the game on DC/QC up to 2nd lieutenant so I'm not sure how good I am as a reference.




steel god -> RE: The AI CSA in the West (8/7/2009 1:38:37 PM)


quote:

ORIGINAL: hgilmer3
It does seem of note that I read a brief on the Battle of Gettysburg and it didn't seem like they had very big armies.  Total casualties ended up being like 23K for Union and 25K for CSA and the total forces were something along the lines of 160K total troops both sides put together?

The AI does not handle the west well, though, especially once you build up enough troops.  I have only beaten the game on DC/QC up to 2nd lieutenant so I'm not sure how good I am as a reference.



Your perception on army size is correct. The Armies in the east were the largest for both sides, and the South usually had less than 70,000 men in the Army of Northern Virgina, sometimes significantly less. The Army of the Potomac hovered around 90,000 men, sometimes going well over 100,000, especially in the Spring before the campaign season started, but did not maintain that level of strength until very late in the war.

Out west the armies were considerably smaller, and usually below 50,000 men, and for the south almost always below that. The population of the US in the 1860s was about 27 Million (non-slave), of which less than a 1/3 of that was in the southern states. So using some very loose math you can speculate that of the south's 9 million population, 4.5 million were men, and roughly 40% of that total would be between the ages of 16 and 40, giving a manpower reserve of 1.8 million. From that you imagine how impossible it wold be for the south to have ever fielded 3 or 4 100,000 man armies and kept them up to strength (not even considering the manufacturing needed to equip them). So yes, armies in the ACW tended to be small compared to their European contemporaries. Also gives some good insight into what effect a European intervention would have had.

As for the AI having issues in the west, I would expect the AI to have greater problems there than in the east. The west is larger, more open, and has less natural geographic barriers to form a defense line on, and therefore has exponentially more permutations to consider when it calculates each more. The more the AI has to think the more likely it is to do something unreasonable.




Page: <<   < prev  1 2 3 [4]

Valid CSS!




Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI
0.6914063