TacAI

[Larsen]

There are a number of layers in CM games.

Strategical layer - you overview the map, analyze it and you the best way (and in case of QBs the best force selection) to achieve the best result. the player has the full control over how he wants to fight this particular battle.

Tactical layer - usually the battle consists of a number of small encounters (unless we talk about really tiny battles) that involve a company, platoon and a number of supporting assets in a particular situation that came to be during the strategic battle execution. Again, here the player has the full control over which units do what and where they move.

Unit level - this is how one specific unit behaves during the battle. And at this level for CM2 the player has absolutely no control over the specific unit. And that, in my opinion, is a big problem.

Let me explain. In CM1 units were represented by abstract entity that had some abstract firepower which depended on the distance and types of weapons the unit carried. Teh firepower would change as the unit would suffer loses. At no time we actually had to worry where and how individual soldiers inside this unit were located. Not so in CM2. From one hand we, as a player have no control how the units within the teams are spread out over a specific action spot and from another the outcome of the way the soldiers are positioned directly affects its effectiveness. Supposedly, BFC spent tons fo time making the units behave in the most natural way to relieve us from the micro managements. Well, the best I can say is that the result is mixed. 

1. One action spot is 8m by 8m. It is occupied by a team - 3-6 soldiers, or 4-6 soldiers of a support unit (MG, mortar, gun etc.). That is a lot of guys in a small, compact space. It is bad already. It gets worse when the artillery or direct HE or MG fire start hitting this particular spot. Normally, if you look from the top the teams, the way they are spread out in an action spot, resemble a small gay orgy - one of the top of another one. If it happens so that an HE shell from even the smallest mortar or a tank 75mm shell lands in the spot the whole team probably will eb wiped out. Given how the moral is affected by losses such a poor management of soldiers' positions has a direct effect on the morale of a much bigger unit. And here we can't do anything because that layer of abstraction is not available to us and is handled by the TacAI.

2. LOS. This is also a direct result of the troops' position inside the action spot. One of the guys might see something but the gunner can't. And rearranging the direction of the unit might or might not solve the problem (most often it does not). Frankly speaking the CM1 abstraction worked much better in terms of how individual units spotted and fired at targets. Here I am not talking about the borg spotting but rather the unit either saw something as a whole or did not and we didn't really need to worry about which guy in the team or vehicle saw something and which didn't. And to add an insult to this the gunner might lose a sigh of the target that did not move. Just like that. One turn it sees it and the next turn it does not. Go figure.

3. Foxholes. They prove descent cover. But... if you look closely at the units int eh same action spot as foxholes more often than not a few of them will be outside of foxholes and since CM2 is so particular about single soldiers those guys that are not int the foxholes will get killed really fast by artillery or small arms fire. Rearranging the direction in which the troops face helps putting them in the foxholes. But they tend to get out once they start firing and change the direction at which they face.

4. Armor. Here I am not sure what the right behavior should be. Maybe guy who served in the armor divisions can help. In CM1 if a tank is faced with a thread that it deems formidable it would pop smoke and reverse. Sometimes they do that successfully, sometimes they get knocked out but they tried to break the LOS with the better tank. In CM2 the default behavior is to dismount... to be killed by MG, HE or artillery fire. I just don't get it. Isn't the chance of surviving higher if the tank pops smoke and reverses rather than bailing out? And the level of troops has very little effect. I just had a crack tank with a descent armor bail out after the internal armor spalling

Summary, although CM2 is supposed to be a better, more realistic simulation than CM1 I find that this TacAI behavior is inferior to the simplistic representation of CM1 engine.

What do you think?

[chuckdyke]

16 minutes ago, Larsen said:

What do you think?

The Devil is in the detail. An infantry squad if you use the move tool you see that the TacAI spread the unit over 2 or three action squares. If you see the foxhole icon it will provide cover for only four men. The question is when to take over manual control from the TacAI. If you see an enemy tank on the UI but your unit doesn't have a tentative or full contact, I don't expect the TacAI to play the game for me. It will lose even when you can plot a LOF on the enemy unit.

 

[Larsen]

I am not even talking about a squad. Ia m talking about a team taht occupies one action spot. Even if you have a team with 3 men and you place it in an action spot with the foxholes there is a good chance that open of them will be staying in the open unless you manually rotate the squad so taht all of them get into the foxholes. That will last until they start firing and turning inside the foxholes. Then some of them will get out again. LOS is another side effect of the micro TacAI.

[chuckdyke]

Foxholes are generic. A typical example is the PAK 75, you can't put an AT-Gun in a foxhole. It is there to protect the crew. This is the method to get rid of them. The crew couldn't stand the heat. 

[Codreanu]

I would rather have Battlefront work on refining and improving the TacAI behavior rather than adding more abstraction to it. I've never played any of the CM1 games so I can't speak to how well it worked in those. I do think that a lot of learning how to play CM2 is learning how to best utilize orders at the tactical level to make the individual unit AI do what you want to more often than not, so it could use some work. 

[chuckdyke]

3 minutes ago, Codreanu said:

I would rather have Battlefront work on refining and improving the TacAI behavior rather than adding more abstraction to it. I've never played any of the CM1 games so I can't speak to how well it worked in those. I do think that a lot of learning how to play CM2 is learning how to best utilize orders at the tactical level to make the individual unit AI do what you want to more often than not, so it could use some work. 

I found that at times doing less macro management is better. The Hunt comment is typical. Just put two-way points without splitting a squad and it did the job better than splitting up and make a dozen waypoints. 

 

[Glubokii Boy]

The AI needs work...YES !

But i can see two sides to this problem...

 

1 As primarely a H2H player then yes the small unit behavior may be the most important to fix.

2. As primarely a single player i would considder the 'overall' AI performance (one or two levels up from single units) to be far more important to improve. Having larger parts

of the AI force be able to co-odinate several units and different weapons in a clever way as well as use the terrain better is in huge needs of improvements imo.

 

The fastest and easiest way to achive point 2 imo is to let a human (scenario designer) 'help' the AI when designing the scenario via improved scripting tools.

 

 

 

 

 

[Bulletpoint]

One thing I would like to see would be that if I have a team with a MG, and I then give them a facing order and click on a specific spot in the distance, then the team will rearrange to make sure as many troops as possible, and especially the guy with the MG or bazooka, gain LOS to that particular spot.

Edited February 21, 2022 by Bulletpoint

[chuckdyke]

2 hours ago, Bulletpoint said:

One thing I would like to see would be that if I have a team with a MG, and I then give them a facing order and click on a specific spot in the distance, then the team will rearrange to make sure as many troops as possible, and especially the guy with the MG or bazooka, gain LOS to that particular spot.

She doesn't do that with a facing arc?

[Erwin]

5 hours ago, chuckdyke said:

She doesn't do that with a facing arc?

"I would like to see would be that if I have a team with a MG, and I then give them a facing order and click on a specific spot in the distance, then the team will rearrange to make sure... the guy with the MG or bazooka, gain LOS to that particular spot."

Agreed... the current problem with some weapons like HMG team is that the unit seems to have LOS to a target when in actuality only the 3rd ammo bearer has LOS and can fire.  Currently, the HMG gunner will not move the HMG so that the HMG will get LOS and fire at the target.

[Larsen]

1. The strategic AI is very hard to make in a game like CM than the TacAI. There are just too many choices of units that could be used in a different way. TacAI is much easier to get eight because the scope is much smaller.

2. The units in the foxholes are not abstracted. If the guy is shown to be out of the foxholes than he is out and the protection and cover that the foxholes provide is gone.

3. Target arc might help or might hurt. You have to zoom into the init in a foxhole to check where the guys actually are within the action spot.

4. Guns in general are a huge disappointment in regarding to the TacAI. Basically once you olace them you better hope they hit something that is worth more than their point cost. In CM1 you could move them effectively. In CM2 i find it super hard to actually move them. There is one guy who pushes the gun and the rest walk in a different speed. Once the gunner gets shot the gun basically becomes either abandoned or stuck in place.

[Glubokii Boy]

44 minutes ago, Larsen said:

1. The strategic AI is very hard to make in a game like CM than the TacAI. There are just too many choices of units that could be used in a different way. 

I agree completally. Beliving that we will get a fully 'self thinking' strategic AI in CM2 or even CM3 is nothing but wishful thinking imo. As you mentioned the complexity of a battlefield is simply to big for a gaming AI to handle. That is not what i'm suggesting. What i'm suggesting is that the scenario designer should be granted better scripting tools to be able to guide the AI in these complex decisions. Decisions that the AI is painfully incapable of handleling on its own. 

As i have mentioned many times before one of those scripting tools would be an increased number of AI groups. For anyone who would like to run a small experiment on this try something like this. A company sized AI attack.

version 1. Use 3 AI groups when designing this scenario.

version 2. Use 16 AI groups when designing the same scenario.

Question....In witch of these two scenarios are you able to design an AI attack that makes good use of the terrain, possition and use support weapons in a decent way, coordinate with armour support somewhat effectivly etc, etc...

Simply adding more AI groups will obviously not solve every problem that the AI is facing but it is a REALISTIC improvement to CM2 imo. Combine that with some additional scenario editor UI changes like reinforcements by triggers for example as well as a reworked AI artillery programing interface. Smaller changes like this would improve the strategic AI significantelly. Expecting some major changes to the AI performance during the CM2 lifespan is not realistic imo. That will have to wait for CM3 i'm sure.

Scenario editor UI changes ought to be doable though even during the CM2 timeframe.

 

[womble]

In the end, there will always be some level of abstraction and approximation. I'm with Codreneau on wanting to see that abstraction layer pushed further and further down rather than trying to go back towards the level used in CMx1.

The original question: 

19 hours ago, Larsen said:

Summary, although CM2 is supposed to be a better, more realistic simulation than CM1 I find that this TacAI behavior is inferior to the simplistic representation of CM1 engine.

What do you think?

I think it's more realistic, since it models at an individual level, and can accommodate extremities that "design for effect" systems like CMx1 was would never permit to happen. You don't get CMx1 pTruppen playing hide and go shoot in and out of forest undergrowth. You don't see the MoH/VC/Hero of the Soviet Union/KC/MOVM moments when your representation of a *squad* is 3 men.

[chuckdyke]

The primary weapon should have the LOS.

[KGBoy]

I've seen my men rearrange themselves once the fighting starts. One on the left flank crawled right across 3 men to find better cover on the right flank. I was impressed with that!

[chuckdyke]

52 minutes ago, KGBoy said:

I've seen my men rearrange themselves once the fighting starts.

On Hunt they look very impressive going through the back streets all of the squad going through a proper drill.