Strange AI Behaviour (Full Version)

All Forums >> [New Releases from Matrix Games] >> Gary Grigsby's War in the East Series



Message

Mentor -> Strange AI Behaviour (12/17/2011 4:17:51 AM)

Hi All,

Here is the situation:

Me Axis vs. AI Soviet, challenging difficulty. Turn 89 (Feb 25, 1943). I started the game on version 1.04.36, and beta patched within a turn or two of the patches being made available. I think I switched to the 1.05 version by about T10 or so.

I was happy with my '41 campaign, although I did not capture Lenningrad, Moscow, Voronezh, Stalino, or Rostov, that was not the goal. I was aiming to destroy Soviet units and weaken the Red Army for the blizzard counteroffensive. The blizzard was rough, I was pushed back about 6-8 hexes, but I fought for every hex and made the Soviets pay. By the spring my infantry was in shambles, but the mobile units were rested. The Red Army was a mess, suffering huge casualties in the winter in many failed attacks.

In the summer of '42 I drove north to link up with the Finns and pocketed about 1.2 million of the best Soviet units around Lenningrad. It took 2 armies half the summer and the entire fall to finally reduce the pocket.

I rested in the mud, and then launched into a winter offensive against Moscow. I encircled the city, and again perhaps 800k of men in strong Soviet formations. I just finished capturing Moscow in the screenshots I am about to show. Ever since Moscow was surrounded the AI has been acting strange. Screenshots below will illustrate, bear with me as I'm posting the entire front.

First, here is the Leningrad sector. This is how I would expect the front to look in the winter of 42/43. Some stacks of good Soviet units interspersed with some mediocre divisions. The AI has been making strong and successful attacks here, but no real threat.

[image]local://upfiles/40459/26CB4F6D23A74986A375B192706675E3.jpg[/image]



Mentor -> RE: Strange AI Behaviour (12/17/2011 4:21:49 AM)

North Moscow sector. Again relatively strong Soviet formations are making successful counterattacks on my tired units. Those Soviet mech corps are strong.

[image]local://upfiles/40459/5716DCCB12A948F3A4BCADA9438F1EA0.jpg[/image]



Mentor -> RE: Strange AI Behaviour (12/17/2011 4:24:06 AM)

South Moscow. More strong units doing what I expected, countering. These guys are ripe for encirclment however.

[image]local://upfiles/40459/22E74B0F211A4251ADA7C32DE26FEC31.jpg[/image]



Mentor -> RE: Strange AI Behaviour (12/17/2011 4:26:41 AM)

Voronezh region. The Soviets have been very active here, making strong attacks on my weak and exposed infantry. A panzer army is finally just arriving to lend support and should stabilize things.

[image]local://upfiles/40459/3C18FFF37BCA4285B6BDC254B36A13A3.jpg[/image]



Mentor -> RE: Strange AI Behaviour (12/17/2011 4:30:37 AM)

For the Don Bend area I have zoomed out just to illustrate the scale of things. This shot is after my recon for the turn, what you see here is most of what is in the area. WTF? This has been happening steadily ever since I closed the ring around Moscow. As far as I can tell, there is only 2 rifle brigades between me and Stalingrad.

[image]local://upfiles/40459/AC85DDA6ADD740BFB3288955A07CBC86.jpg[/image]



Mentor -> RE: Strange AI Behaviour (12/17/2011 4:37:02 AM)

Finally, Crimea. Same strange behaviour here. The AI had strong stacks (>20 CV) guarding the approaches, then suddenly just opened the door and let my allies in.

The Soviets, while suffering big defeats in the past 6 months, still have an OOB of 6.0 million men. They should be able to hold the line. As it stands now 1st Panzer Army is going to roll up the front from Voronezh, merrily pocketing mechanized units as it goes until there is nothing left south of Moscow.

Is this unusual behaviour from the AI? I was not expecting such a cakewalk in Feb '43 on challenging.

[image]local://upfiles/40459/C33BBF3CD4CB4D2CA6BBC0C722A15A21.jpg[/image]



Mentor -> RE: Strange AI Behaviour (12/17/2011 4:43:18 AM)

Or is it a devious trap laid to get me to overextend?

[sm=00000619.gif]



Joel Billings -> RE: Strange AI Behaviour (12/17/2011 6:29:51 PM)

Please send a save at the end of your turn or better at the start of the AI turn to 2by3@2by3games.com. It's possible that the computer has lot more men/units than it can handle, although it should try to put some kind of line together.



carlkay58 -> RE: Strange AI Behaviour (12/18/2011 1:40:20 AM)

Perhaps it is the lack of the Moscow Command directing the defense? Stalin is despondent upon the loss of his favorite bed? I actually put more belief into the "devious plan" category.

Seriously though, I have found that when the Axis threatens or takes Moscow, the AI seems to abandon the southern flank to mass troops to retake Moscow. This may be what is happening in your game.



Redmarkus5 -> RE: Strange AI Behaviour (12/18/2011 3:21:55 PM)


quote:

ORIGINAL: carlkay58

Perhaps it is the lack of the Moscow Command directing the defense? Stalin is despondent upon the loss of his favorite bed? I actually put more belief into the "devious plan" category.

Seriously though, I have found that when the Axis threatens or takes Moscow, the AI seems to abandon the southern flank to mass troops to retake Moscow. This may be what is happening in your game.


Too much complexity in the engine. The more complex it becomes, the more bizzare the outcomes will be - Chaos Theory applies and the current approach is destined to fail.

For a solution (in a brand new game engine, not a recycled one) the principles of fluid mechanics provide a better set of rules for governing how an AI should respond to enemy penetrations and align its forces. We need to move beyond chits sitting in hexes for this class of game and examine how combat units are actually situated on the battlefield and how they are really represented on operational military maps; frontage, depth, orientation, axis of advance, logistics train, etc.

So, a division might be stretched to cover 20 miles of front here, while a Corps might focus on a sector of only 2 miles over there. Stacking and ZOC do not work sufficiently well at WiTE scale to address this. In Unity of Command they have been done away with altogther, which is a further retrograde step.

A continuous AI line of varying depths (like the edge of an advancing pool of water) that treats terrain, road and rail infrastructure, strategic objectives and enemy units as factors in a 'resistance score' can advance or receed in a far more realistic fashion than hundreds of discrete 'chits'. Using fluid mechanics to model this would ensure that no gaps in the front exist, although the 'depth' of the fluid may be greater where the need is greatest. Breakthroughs can still occur where that depth is low.

Probably sounds like the illusions of a drunkard to most, but think it through for a while ;) Some new creative thinking is desperately needed in this field.



Joel Billings -> RE: Strange AI Behaviour (12/18/2011 11:10:23 PM)

We purposely make hex grid games. I agree that there are other styles of games that could be done, but that's not what we are doing. All games have limitations and tradeoffs. We admit we are dinosaurs, but it's what we know and like. At the same time, Gary and I both loved Sid Meier's Gettysburg and the old SSI Fighter Command, which are not hex games and have a real time element. Even though I think SMG is one of the best computer games ever, I can tell you all the things I don't like about it (especially regarding the AI), but many go to the heart of the system, so changing them would ultimately change the game and not necessarily for the better.

I find your thesis depressing (that the more complexity the more we are destined to fail). I choose to keep trying even though we are destined to fail. But then, The Myth of Sisyphus was always my favorite story in college. [:)]



gingerbread -> RE: Strange AI Behaviour (12/19/2011 12:30:39 AM)

I would have guessed Zorba who recommends a dose of madness in order to try.  [:)]



Page: [1]

Valid CSS!




Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI
0.953125