Combat Generals...

[Fox]

Something I've noticed in allot of my games and have been meaning to bring up for a while is the tendency for the AI to move very high level HQs (Corps Division and the like) right up into the front lines.

Below is a recent example I captured in the "Race for Bastogne" scenario where the two highest ranking HQs for the American side (VIII Corps and 28th Infantry Division HQs) got moved right up to the scenario start front-line on "skyline drive". As much as I admired their offensive spirit this was a somewhat disadvantageous place to put themselves as they ended up ~25km behind my leading units and got cut off and destroyed, which no doubt produced severe problems on the AI's ability to resist for the rest of the game.

Day 1 05:30 (scenario start, both images were taken after manually surrendering so there are no intel issues)


Day 2 15:14


I'm not sure of any reason the AI would see to move these units so far forward, any ideas on the cause/solutions?

 

[jimcarravallah]

Something I've noticed in allot of my games and have been meaning to bring up for a while is the tendency for the AI to move very high level HQs (Corps Division and the like) right up into the front lines.

Below is a recent example I captured in the "Race for Bastogne" scenario where the two highest ranking HQs for the American side (VIII Corps and 28th Infantry Division HQs) got moved right up to the scenario start front-line on "skyline drive". As much as I admired their offensive spirit this was a somewhat disadvantageous place to put themselves as they ended up ~25km behind my leading units and got cut off and destroyed, which no doubt produced severe problems on the AI's ability to resist for the rest of the game.

Day 1 05:30 (scenario start, both images were taken after manually surrendering so there are no intel issues)
View attachment 5177

Day 2 15:14
View attachment 5176

I'm not sure of any reason the AI would see to move these units so far forward, any ideas on the cause/solutions?

Pg 147 of the 1.1 Version of the Game Manual discusses Orders Delay, which is reduced or increased, in part, by the distances between units issuing orders, and subordinate units receiving those receiving them.

 

[Kurt]

That's not the answer he was looking for Jim , he simply wants to know why these high level " rear area ^ units are used by the enemy AI in front line combat . The answer is simple , the AI or rather its set of instructions / algorithms are flawed and in need of fixing .

 

[jimcarravallah]

That's not the answer he was looking for Jim , he simply wants to know why these high level " rear area ^ units are used by the enemy AI in front line combat . The answer is simple , the AI or rather its set of instructions / algorithms are flawed and in need of fixing .

So, what the manual says regarding distances affecting how quickly orders are transmitted has nothing to do with the issue?

Is it possible the algorithm affecting the AI decision logic over values the impact of those distances, and perhaps results in the behavior?

Is the manual's nominal description of the process used to form AI behavior simply to be ignored in favor of "it's flawed and needs to be fixed?"

I asked a long time ago whether the software documentation used to define the AI algorithms could be released and was told that documentation is both proprietary (competition sensitive) and in some cases non-existent. Next best source of that information appears to be the game manual(s) for the various software packages. If that isn't a source of information regarding how the AI functions in the player's behalf, then Dave wasted a hell of a lot of time recording it when it would have been easier to simply say "it's flawed, and we'll have to fix it."

 

[Kurt]

Your first point about coms distances dose not explain why base units and arty units also appear to be in enemy AI front line .
I do agree that we as dev's need access to to decision making processes regarding AI behavior .
Your reference's to the manual and the question of whether Dave has " wasted a hell of a lot of time " I can't answer .
The fact however remains that the AI is indeed flawed and in need of fixing .

 

[Fox]

So, what the manual says regarding distances affecting how quickly orders are transmitted has nothing to do with the issue?

Is it possible the algorithm affecting the AI decision logic over values the impact of those distances, and perhaps results in the behavior?

Is the manual's nominal description of the process used to form AI behavior simply to be ignored in favor of "it's flawed and needs to be fixed?"

The distances in the manual are to my understanding there to specify the distances at which orders delay will start to increase if you exceed them. Whether the AI uses these distances in order to make a decision where to place its HQs is a separate issue. Regardless I think it's fair to say that this decision making process has some problems as it ends up placing High level command units in dangerous front line locations that wouldn't be selected by real commander.

I asked a long time ago whether the software documentation used to define the AI algorithms could be released and was told that documentation is both proprietary (competition sensitive) and in some cases non-existent. Next best source of that information appears to be the game manual(s) for the various software packages. If that isn't a source of information regarding how the AI functions in the player's behalf, then Dave wasted a hell of a lot of time recording it when it would have been easier to simply say "it's flawed, and we'll have to fix it.

I don't think there is a documentation problem

 

[Fox]

Your first point about coms distances dose not explain why base units and arty units also appear to be in enemy AI front line.

The problem as far as I've noticed it is confined to HQ units. The base & arty units that are located next to my forces are in the screenshots are there because my units overran their positions.

 

[jimcarravallah]

The distances in the manual are to my understanding there to specify the distances at which orders delay will start to increase if you exceed them. Whether the AI uses these distances in order to make a decision where to place its HQs is a separate issue. Regardless I think it's fair to say that this decision making process has some problems as it ends up placing High level command units in dangerous front line locations that wouldn't be selected by real commander.



I don't think there is a documentation problem

What you think and how the military simulation / military training industry operates are two different things.

Dave and I had the discussion back when Panther was hoping to become a source for developing simulation and training software to support the US Army and I was working with a separate vendor on inserting realistic logistics effects into existing Army battalion and above commander training simulations. The issue boiled down to a lack of resources necessary to produce the documentation with a hope that going forward, a vendor procuring Panther's expertise would help fund the development to resolve the issue.

The discussion morphed into development of logistics distribution modeling you're using today in CO2, supported in part by some flow diagrams and decision logic trees used to document the changes necessary to make that distribution more reflective of real world situations.

 

[Kolbex]

What you think and how the military simulation / military training industry operates are two different things.

Dave and I had the discussion back when Panther was hoping to become a source for developing simulation and training software to support the US Army and I was working with a separate vendor on inserting realistic logistics effects into existing Army battalion and above commander training simulations. The issue boiled down to a lack of resources necessary to produce the documentation with a hope that going forward, a vendor procuring Panther's expertise would help fund the development to resolve the issue.

The discussion morphed into development of logistics distribution modeling you're using today in CO2, supported in part by some flow diagrams and decision logic trees used to document the changes necessary to make that distribution more reflective of real world situations.

So what you're saying is it's real world SOP to move your HQ up to the front line and get overrun?

 

[jimcarravallah]

So what you're saying is it's real world SOP to move your HQ up to the front line and get overrun?

No.

One component of how quickly an order can be addressed is measured by the administrative delay time between when the order is conceived at headquarters and when it can be delivered to a subordinate unit for execution. A component of that administrative delay time in the World War II era, is the distance over which the dispatch has to travel, particularly if portions of the communications grid is maintained by courier-delivered dispatches rather than radio or pre-strung telephonic communications.

Orders delay takes into account the distance component of administrative delay time and the algorithm that calculates that delay time based on the type of headquarters issuing the order may over evaluate the distance component effects and thus cause AI-controlled headquarters units to move closer to front lines to reduce the delay time.

Absent a programmer's evaluation of the algorithm, one means to determine if the distance component is over weighted, and thus causing the unit to move closer to the front to reduce the orders delay time is to have documentation which someone not familiar with the programming language could review for the decision logic that affects a headquarters' AI behavior.

One hint there may be a linkage is included in the game manual, which defines a relationship that includes distance as one of the controlling factors for calculating orders delay time. How that factor is calculated with other aspects of orders delay time (receiving unit cohesion, fatigue, complexity of the order, the cumulative administrative burden at the headquarters unit, HQ and receiving unit staff efficiency, commander capabilities and the like) could result in the faulty AI behavior.