Does the AI Cheat?

[marquo]

I have cut off the Shenandoah Valley for months, maybe even a year; the CSA forces there refuse to leave presumably because of victory point cities. The problem is that they are cut, supply cannot get in, and they simply do not degrade. My USA forces take massive hits and even evaporate when OOS after several weeks. Does the AI cheat when it comes to supplying it's units?


Notice the magnificent 3 region jump Jackson did - he has been isolated, OOS and yet managed to bond over/through my forces. He, too, has been cut off OOS for months yet manages this. Does the AI cheat when it comes to supply and moving it's forces?

Thanks

[ATTACH]26152[/ATTACH]

[James D Burns]

Each circled area has a depot in it, so they all produce some level of supplies each turn. I think this also allows replacements to be drawn, so they are probably undersupplied and taking supply hits but replacements refill the lost strength each turn.

If you were in defensive posture Jackson may have gone passive and evaded combat to slip through your lines. Check the text messages and look for a repot with a failed percentage die roll for your troops attempting to intercept him. If you don’t see a report like that I’m stumped how he got through.

Jim

[GraniteStater]

Define 'cheat'. Seriously, what one means by the term is important.

Looking at your screenie, what I see is several CSA stacks with really low Cohesion and next to no Supply. Warm bodies are somewhat reduced, also.

You need not fear these formations, I would say.

As for Athena doing 'knight's moves' - examine Before & After carefully, especially adjacency of Regions, ability to move through Regions occupied by hostile forces (enemies on Defend do not necessarily prohibit movements or engagements), etc.

There are a lotta Rules in this game & it takes some time to get a good 'feel' for it. I'm more a 'feel' guy, IOW, does the program model the historical considerations well?

IMO, it does, very well.

[marquo]

Conceptually how do replacements refill out units which are cut off - and how long can depots supply units when ringed by regions which do not allow the passage of supplies? The do not have infinite capacity, do they?

Cheat = the AI given advantages under the hood which are not available to human players; things liked enhanced ability to survive OOS, advantageous bonuses to die rolls, etc.

[GraniteStater]

A) 'Dunno' is the short answer. Good observation, though.

B) Well, every AI cheats - they have to, it's just code with probabilities and priorities attached. In this game, you can specify How Much Cheating Allowed.

[James D Burns]

It’s a game engine limitation. Depots are production facilities, therefore they produce whether cut off or not. It would be far too complex to try and fashion a rule that isolates production facilities, you’d have to come up with some draconian rule that states they have to trace a supply line to the capital or something, and it’s far worse to ring the capital and end all production across the entire map than it is to simply let them produce their meager supply each turn when cut off.

No way is a single depot (even if maxed to level 4) going to produce enough supply to feed an Army let alone a Corp, so I doubt it’s worth the effort to try and jury-rig a fix into the game for something like this.

I should also note that Moore WV (might be Strasburg though, ammo graphic is unclear which region it goes with) has a couple of ammo stacks shown, so at some point it probably had a large amount of supplies on hand that have been eaten through. Give it a few more turns and you may see severe damage to the guys trapped inside, it all depends on the size of the depots present. If they’re level 1 then they aren’t making much supply each turn. I think they need to be level 3 to produce ammo, so it's possible they're pretty large depots.

Jim

[Pocus]

You can't receive replacements if you are not in full supplied status, so I doubt here...

As others said, there are depots and it seems to me cohesion is not that high. Also, don't forget about in-region foraging.

The AI don't cheat on supply or weird 'teleport' move. She has slight to significant bonuses on move speed or cohesion recovery (but only if she could recover in the first place) depending of the game level. No combat bonus either.

[GraniteStater]

BTW, looking at the screenie: it appears Jackson's path was Falmouth - Culpeper - Nelson. It pays to scope out what's adjacent to what - most especially in NoVa. Believe me, I know - I was taken to school on this part of the map by some very good players. There are several "interesting' areas on the map: 'Paduchaland', NoVa, the Arkansa River valley, central Tennessee - gee, all the areas where people might brawl!

I think they did it on purpose - seriously. The line of the Rappahannock in Virginia is almost designed for a MTSG defense wanting to hold F-burg over to Culpeper and C-ville.

[pgr]

It’s a game engine limitation. Depots are production facilities, therefore they produce whether cut off or not. It would be far too complex to try and fashion a rule that isolates production facilities, you’d have to come up with some draconian rule that states they have to trace a supply line to the capital or something, and it’s far worse to ring the capital and end all production across the entire map than it is to simply let them produce their meager supply each turn when cut off.

No way is a single depot (even if maxed to level 4) going to produce enough supply to feed an Army let alone a Corp, so I doubt it’s worth the effort to try and jury-rig a fix into the game for something like this.

Jim

Well It may not matter too much for a sparse region with 1 level 1 depot, but what about a fully developed Richmond? I currently am playing a solo game where I methodically isolated the region and have it completely surrounded. The problem is that the large CSA force inside is so dug in and well led it is impossible to dislodge, but the Richmond area is so rich that they are as happy as clams inside the ring and will never be out of supply. (Which in reality, once Petersburg was lost Richmond became untenable.)

I think the big problem is that towns and cities don't consume anything. Why not cap general supply production for all towns and cities, as in a size 1 town produces as much as a size 20 city, but that towns and cities consume general supplies as a factor of there size. Set it up so that a level 5 town represents the break even point, above which a city eats more than it produces. Industrial structures should only produce ammunition and money, not general supplies. Depots would remain unchanged, and would serve as the nodes in the distribution system.

This would make it a fundamental liability to stay in a cut off city, because the army is competing with the inhabitants for general supply. Starving cities would reduce national moral (think of Richmond bread riots) and be less productive. Finally, you would have to protect your rural areas to feed your forces. (Making the loss of the Shenandoah Valley a big deal for the ability of the ANV to supply its self for example)

So I say make cities eat so that no single region can become a self-sufficient island.

[G-Burg Bullet]

Excellent observations and suggestions, pgr. I agree with you about major cities. Units therein that are otherwise cut off from supply can exist indefinitely within the city. This is pretty unrealistic. I like your ideas.

[GraniteStater]

Hate to differ, but I think these statements need more examination.

Let's take the historical record (it's what the application is trying to model, right?).

I can think of two sieges: Vicksburg & Petersburg.

Vicksburg was pretty airtight, once Grant settled in. Length? Six weeks, about. IIRC, one big problem Pemberton had was low stocks to begin with. So, they were down to rats and prairie dogs at the end. Pemberton failed (among other CSA commanders in the general vicinity) to secure his Stuff at Jackson, which fell to Sherman. Grant moved fairly quickly (advantage of surprise, once he landed on the left (eastern) bank of the Mississippi) and got to Jackson first in strength, at least enough to seize Jackson decisively and quickly.

So, Grant faced a dispirited garrison with very low supplies. It still took six weeks - or 3 turns, maybe 4, depending on how you would set the benchmark in the code.

Petersburg was what - nine, ten months? Take a look at the maps of the operations around P-burg then. Grant didn't have it surrounded. He kept inching to his left - in the spring of '65, Five Forks settled the issue. Why didn't he envelop it in the first place? The answer is, he would've if he could've, but couldn't, so he didn't. He couldn't because he had a sizeable opponent, well led and determined. To suddenly rush to the left and start digging works - the oppo ain't gonna let you just do that without interference. So, it was step-by-step and is part of the reason the siege took as long as it did, other factors being time of year, weather, etc. Yes, P-burg was not isolated, like V-burg, but the general supply situation for the CSA in Richmond and environs was getting pretty bad. Still, Richmond, by itself, had stores. The big problem was transporting stuff from outer farmlands, via RR, etc. Cavalry raids, anyone? Five Forks settled it, 'cuz it was spring, Grant could maneuver, and now he had, essentially, flanked P-burg. It wasn't Out of Stuff (still connected to Richmond) - it was untenable (Grant couldn' t be seriously impeded from isolating P-burg after Five Forks), which is why Lee withdrew. BTW, I lived in Dinwiddie County as a kid and attended school in P-burg.

See what I'm saying? I don't have an issue with larger cities holding out. Remember, the regions can be a couple of hundred square miles or more. You can have opposing units in the same Region & No Combat. The maneuvering and fighting within any given Region is an abstraction, over which the player has little to say. We're not Generals, we're stage managers - we coach, encourage, set the rehearsals and raise the curtain - but what happens on Wednesday night on stage can be very different from what you rehearsed.

I don't have a big issue with even large forces hanging in and hanging out. To me, it's all a sense of proportion. I would question the Army of NoVa hanging tough and long in an isolated Manassas under siege.

Which is why you do see besieged forces attempting breakout battles. And why you see surrenders. I would have to re-read about sieges myself, but I think a lot of the original AACW mechanics are still in place and if one played AACW long enough, one could say the siege possibilities were pretty fair.

[James D Burns]

Great post Granitestater, I concur completely. Many players forget we are simulating battles that occur in regions usually larger than several counties. The notion that if you simply march into a region you instantly surround an opponent is a horrible abstraction failure of the system and designing in ways to survive that kind of game engine weakness is part of what makes the game so enjoyable.

I think where players are running into frustration is when they approach the system with a mindset if I do X then I expect Y every time. That kind of mathematical certainty is a long standing part of board wargames of old, but should not be in a game that simulates massive army+ sized battles in regions that span huge tracts of land. For that kind of a system to work you need uncertainty and a myriad of possible outcomes if you want the game to feel like the civil war.

What I would like to see is a force size requirement to isolate cities based on the city’s population size. It makes no sense that a city the size of New York or Richmond can be isolated by a division or two and forced to surrender. Large cities like those should require tens of thousands of troops (possibly hundreds of thousands) to completely isolate them to a degree that no supplies could get in. If that kind of a requirement existed than I’d be all for making it easier to starve out defenders, but as things stand now it’s WAY too easy to reduce fortified cities already, making it even easier would hurt the simulation in my view.

Jim

[GraniteStater]

Yup.

[pgr]

GraniteStater I see your point, but in reality there is not that much of a difference in how Vicksburg and Petersburg played out. Once cut off, the place will fall. The main difference was that, as you pointed out, Grant was able to completely invest Vicksburg quickly. In game terms, he took the region containing Vicksburg and besieged the city. Over in a matter of turns.

The siege of Petersburg lasted so long precisely because Lee made it very difficult for Grant to cut his lines of communication. Up until 5 Forks, Lee had open rail links south, but as soon as they were threatened, he had to abandon the place. The essential point is that the confederates held out so long at Petersburg because they continued to receive supplies from the outside. They were not living exclusively off of the production of the cities of Richmond and Petersburg.

I agree that some regions of the map can represent vast areas, but the game regions in Va are relatively small. The region around the city of Richmond represents its immediate surroundings and wouldn't have been able to support its civil population let alone the Lee's army. So if a player has managed to completely occupy ALL the regions around Richmond (taking Petersburg cutting all the rail lines etc) there should be no way any sizable force (let alone a force of 70,000) should be able to survive exclusively on the output of Richmond.

Of course if I was really worried about this, I would actually run the numbers of what a 70,000 man army consumes verses what a fully developed Richmond produces...

[GraniteStater]

"The map is not the territory."

No map, or model, has 100% fidelity. Newtonian physics is awesome as an explanation of basic kinetics and dynamics - until you start to approach relativistic velocities. Then Newton's model is in a degree of error and Einstein's model is more applicable.

As someone who has professional IT experience - you can't have everything. You can't take five years to make a 'perfect' representation, you have to hit the market at some point, you're in business. They have done the best they could. I have my wishes for features and stuff, but they didn't include me in the design meetings (shudda though ).

I don't know how much you have played AACW (the first game). In my experience, sieges were fair. Now, I didn't siege that much, for several reasons, but what I saw was reasonable. Remember, there weren't that many sieges from '61 - '65.

And don't assume that a siege is an 'I win' for the besieger. There were attempts to relieve Vicksburg, but they were repulsed. Throughout history, sieges have been won or lost by the besieged & the besiegers. Very often, sieges were ended because the besieger's logistics compelled the lifting of the siege. A well designed city/fortress could, very often, hold out, sometimes for an astonishingly long time. This doesn't really happen in the game, AFAICS - one probably needs to relieve the siege from 'outside'. Still, the ability to endure a siege can be a lot more than one may suppose.

The biggest import of Petersburg was not the siege itself. The biggest import was that Lee was cooped up - IOW, Mr. Tricksy Fox, you can't maneuver now, all your tactical wizardry is out the door. Lee knew in the autumn of '64 that he couldn't hold off Grant in that situation forever: too many men, too much Stuff, more men and Stuff coming. Lee needed a relief column - he didn't get it.

Again, two full Corps holed up in Podunk, Tennessee holding out for six months - I would have my reservations. Level 8 cities - if you want to show this isn't modeled right, I would trot out my Historical Research and show that it's not modeled correctly.

I don't have a beef with sieges in the game - and, even if one can show that the model is askew, well, I wouldn't hold my breath waiting for code changes - these guys are awesome at support and patching, but even they want to have weekends and drink beer and such. They did a pretty good job, I think. I really wouldn't worry about the sieges too much - in your example, maybe it's the application's way of saying, Hey, ya gotta come in here and take me on in Richmond. I do believe Cohesion, etc., will drop eventually - but very slowly in the bigger structures.

OTOH, I could be entirely wrong. Just the way I see it.

[James D Burns]

So if a player has managed to completely occupy ALL the regions around Richmond (taking Petersburg cutting all the rail lines etc) there should be no way any sizable force (let alone a force of 70,000) should be able to survive exclusively on the output of Richmond.

But there’s the rub, you can do all that with nothing more than a single regiment in each surrounding region, so it goes both ways. Yes it’s tough to starve out an army that has a huge 2000+ stockpile of supplies on hand in the city, but it is also far too easy to isolate that same army with tiny forces that couldn’t realistically hold a barn or two let alone a county or more worth of acreage.

For the most part the system works pretty well, but there are extreme examples on both sides that see cities fall too fast to forces that shouldn’t be able to take them and you also see cities last for months on end when logic says they should probably have fallen sooner. The problem I see with the proposal is you are going to swing the scale too far towards reducing cities fast and things already move along too quickly in game as it is. If anything we need to slow down the games tempo, not speed it up.

I wouldn’t be opposed to a rule that cut the output of supply production for a region in half if it was unable to trace an unbroken path to two other supply producing regions. But barring that I think reducing overall supply output or forcing cities to consume huge amounts of supply will simply produce an effect that destroys armies sooner and sees games start to end in 1863 instead of 1864 as most do now.

Jim

[Laernius]

Remember too that Vicksburg and Richmond had huge logistical support from the navy. Vicksburg was shelled from the river as well as the land to ensure that there was no safe place for the enemy to hide. The idea being, if I'm not mistaken, by Grant that it would be too horrifying to hold onto for long. At Richmond, Lee's cousin was serving in the task force bottling up the James River, cooperating with the Army of the James in Fort Munroe. There was much more going on than just land based operations. Everytime I think the siege mechanics are fucked, I try again with mortar boats and perhaps a sizable, separate force acting as a distraction and I see the same results that occurred in the real war: 2-3 months at the latest. Perhaps you are not using the right means to effect a more accurate timeframe. More is not always better and sometimes neither is different, too.