AI problems (Full Version)

All Forums >> [Current Games From Matrix.] >> [World War II] >> Norm Koger's The Operational Art Of War III



Message

Thorwald -> AI problems (10/10/2007 5:13:32 PM)

The PO seems to do more harm than good to itself in many turns. If you wish to play against a better AI opponent it would sometimes be better to just interrupt its turn and end it before it can make any attacks.

Some problems I have identified with the "strong" AI:

1) Desperate offensives which lead to 10:1 casualty ratio. The problem can be identified in any scenarios

2) The AI can rarely make succesful air attacks. Usually it loses a lot of planes while I suffer 0% casualties. The air staff assistant is also useless because of this tendency.

3) Tendency to withdraw their forces with the wrong strategy and destroy them. They lose huge amounts of troops as they are engaged or evaporate. Troops which have been engaged are also very easy to destroy. Their troop movements are in general strange as they move them vertically.

4) Strangely, the PO does not fortify properly. It never makes entrenched or dug in units "fortified" even if it does not move them.

I was just victorious in Operation Blau 1942 playing as the Germans in "challenge mode" (I never made more than one round of attacks per turn) on turn 21 because of these problems. I encircled and destroyed the Soviet South and SW fronts with my panzers while the Soviets wasted their strength in the central sector with useless attacks.



JAMiAM -> RE: AI problems (10/10/2007 6:52:41 PM)

Hi,

A few things to point out.

It's not really the AI that is at fault here, it is the PO. This is a very important distinction in TOAW. The PO, or Programmed Opponent is only as good as the designer can make it, and in the case of this scenario, the designer frankly did a poor job of it. The Soviet Formations have ridiculous objective tracks set to them, showing that he really didn't understand the way that the AI will attempt to move, attack, and defend along the tracks. Next, he has the Soviet formations set to Attack general orders, so they will be more aggressive, when they should be more defensive. I did not dig any deeper to see if he used multiple objective tracks for the scenario but it is doubtful and even if he did, the way he's got things cookie-cut at the beginning of the scenario doesn't bode well for any changes down the line. Finally, this is a fairly large scenario. This means that not only is it more difficult to do a good job of programming the PO, but changes from the "expected" can quickly drive the PO off the rails. This is due to the complexity of the situation, and is exacerbated by the fact that human players will often approach things differently than the designer anticipates, and then the PO is hard-pressed to react to situations not already anticipated.

If you're playing against the PO, it is best to choose smaller scenarios. Better yet, though, is to play against live opponents. They will most likely be better than any PO. Treat the PO as a training dummy, and then, go out and find a real dummy...errr...opponent, to play...[;)]



Thorwald -> RE: AI problems (10/10/2007 7:23:06 PM)

Thanks for your response, JAMiAM. I have noticed that in almost all scenarios the PO loses units because of evaporation and engagement. It wants to move units although it shouldn't and I don't remember any scenario in which it fortifies properly.

Smart use of air power never happens either and the air staff assistant seems to be useless. Are my concerns familiar to you?



JAMiAM -> RE: AI problems (10/10/2007 8:03:34 PM)


quote:

ORIGINAL: Thorwald

Thanks for your response, JAMiAM. I have noticed that in almost all scenarios the PO loses units because of evaporation and engagement. It wants to move units although it shouldn't and I don't remember any scenario in which it fortifies properly.


Ralph hasn't been able to teach Elmer how to manage the HQ/Artillery rearguard withdrawal technique. Thus, the computer will always be at a disadvantage in that case. The AI is much better at attacking, than defending, and Elmer struggles to find and tread the fine line between sitting still, to hunker down for a static defense, and adjusting his positions to prevent being outflanked.

quote:

ORIGINAL: Thorwald
Smart use of air power never happens either and the air staff assistant seems to be useless. Are my concerns familiar to you?

The smart use of air power is often a weakness of human players, as well, and Elmer, though he's made some strides in the right direction, is still way outclassed when it comes to using air.




Thorwald -> RE: AI problems (10/10/2007 8:20:48 PM)

Jamiam, generally it would be better if "Elmer" did not make bombardments / airfield attacks at all. Can this be programmed to scenarios? Also, people surely want that their air staff assistant does not do any attacks.



Silvanski -> RE: AI problems (10/11/2007 5:59:51 AM)

I am looking into the PO of Operation Blau 1942, with Brian's consent... The German PO is almost where I want it ... I have an idea or two to beef up the Soviet PO



Curtis Lemay -> RE: AI problems (10/11/2007 6:23:50 PM)


quote:

ORIGINAL: Thorwald
3) Tendency to withdraw their forces with the wrong strategy and destroy them. They lose huge amounts of troops as they are engaged or evaporate. Troops which have been engaged are also very easy to destroy. Their troop movements are in general strange as they move them vertically.

4) Strangely, the PO does not fortify properly. It never makes entrenched or dug in units "fortified" even if it does not move them.


With all due respect to James, I really think these two aspects aren't due to the scenario objectives and could be helped by items 7.8 and 7.12.1 in the wishlist. If I could ever get Ralph to try them, I think they would help the PO a bunch.

To recap, item 7.8 would require the PO to never completely vacate an enemy ZOC (with a few exceptions). That would stop the endless disengagement attacks the PO suffers, would greatly help clearing up enemy units in the PO's rear (or at least not leave those units as loose cannons), and would greatly increase the chance that surrounded enemy units remain surrounded - thereby evaporating instead of escaping upon attack.

The reason the PO rarely gets to Fortified status is because it has to set up attacks just to evaluate them. Most get canceled. But, currently, when attacks are canceled, the units don't return to their previous deployment - they are left in mobile mode. From there, the best you can do is get back to Entrenched, not Fortified. If 7.12.1 were implemented, that wouldn't be the case. Human players would benefit too, but the PO would be the big winner.



Thorwald -> RE: AI problems (10/12/2007 12:06:01 AM)

quote:

ORIGINAL: Curtis Lemay
The reason the PO rarely gets to Fortified status is because it has to set up attacks just to evaluate them. Most get canceled. But, currently, when attacks are canceled, the units don't return to their previous deployment - they are left in mobile mode. From there, the best you can do is get back to Entrenched, not Fortified. If 7.12.1 were implemented, that wouldn't be the case. Human players would benefit too, but the PO would be the big winner.

However, strangely the PO does not even dig in the units although they have movement points left. It leaves many mobile units to the front line. Often I can see only 1-2 entrenched units in a large stack.



Page: [1]

Valid CSS!




Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI
1.109375