TacAI system (was AI cheat in CM) (long)

[tank_41]

I decide to start a new topic on this, because as you read on, you

will see that I am bringing up a new topic although it is related

to my original post "AI cheat in CM".

First, let me say that after numerous testing, I have to change my

mind and agree with most people here that computer does not cheat

in TacAI. So, sorry for all the confusions I created.

However, the testing does reveal many drawbacks in CM's TacAI system

and these drawbacks are part of the reasons that led me initially

think computer is cheating, which I will explain in the end of this

post

Before I go on, I want to say one thing: Please dont flame. These

are merely some suggestions that could potentially make CM better.

Now, into the topic (BTW, I am gonna talk about tank only as

I dont test infantry that much)

I think that the TacAI system depends too much on the static part of

unit, and fails to bring in the dynamic flavor of the battlefield.

Here is what I mean by static part, for example, a stuart vs Tiger,

or a Sherman 75 vs Panther, or a PzIH vs a Pershin, etc. In all these

examples, the decision is pretty much clear most of the time (unless of course, your stuart catchs a Tiger

in its rear).

However, when dealing with dynamic of the battlefield, the TacAI system

needs lots of improvement. Here are three typical scenarios:

1 Shouldn't strong units (whiling facing weaker opponenets) also

consider retreating in some case?

On one of my experiments, I have 2 Shermans 75 ambusing along a road

side (with 1 on each side) 100m away from the road. I put an Elite

Tiger close enough, but outside of the LOS of the shermans at the

beginning of the turn. This setting makes sure that Tiger moves

into ambush point unprepared.

I tried this setting 5 times both from human and computer, and

never seen Tiger back off after get ambushed. It turns the turret

and trying to pick off the Shermans but the attack from both

left and right side side are just too much for it.

Someone may argue that TacAI may not have enough time to react, but

in this case, sherman's first shoots are either missing or bouncing off,

and should give Tiger pretty of time to react.

What I am trying to say here is that yes, Tiger is much stronger unit

than Shermans. But at the close range, with attack from both side,

and with a much slower turret, shouldn't Tiger consider to back off

first?

2 Shouldn't equal units also consider retreating sometimes?

I tried the setting I described yesterday, that is, having four PzIVHs

moving along the road, and 4 Sherman 76s ambusing from left side (about

250M away).

This setting is different than the previous one because you have eqaul

amount of tanks in both side with similar performance (PzIVH is slightly

better), and ambusing is from one side only.

In this case, all computer and human controlled units would turn to

their left and engage the Shermans. The result, 90% of the time,

all PzIVHs are wipped out. Sherman's loss is usally 1 or 2, and sometimes

none.

I tried this setting with PzIV on the ambushing side, and had the

similar result, ie, Shermans would turn turrets to enagage and get wiped out.

My point here is: Yes, PzIVH should not be afraid of Sherman. But after

the first and second PzIVHs were sent into flame, shouldn't the rest two

decide maybe retreating is the only alternative at the moment?

3 I haven't tried this but I have seen it before, and some other people

also posted this: Shoudn't a gun damaged unit take a more precautious

decision on the battlefield?

A Panther with its gun damaged is vulunrable to even a Stuart. But when was

the last time you see this Panther firing a smoke and retreat?

So, I think the TacAI system in CM is somehow pre-calculated like A retreat

when facing B, and not vice-versa, etc. I hope it could be improved to

a more dynamic one and make better judgement.

Finally, I think the reason I initially though computer was cheating is simple:

while playing Germans, with this pre-calculated TacAI, you would hardly see

any retreat movement from German Panzers. Even a PZIVH is not afraid of

the enemy (on a static sense) most of the time (The only exception is perhapes

Pershing?), why would a Tiger, Panther, and KingTiger? But from my part,

I have seen many situations where I think my tanks should take a more

evasive move (sort of like those "weaker" Ameicans tanks), but instead

they take a stand and fight course. So, that gives me a wrong impression.

[This message has been edited by tank_41 (edited 07-26-2000).]

[JoshK]

I have seen a Stug pop smoke and reverse when faced frontally w/ a single Sherman(75). One could argue that is this situation, the Stug would have the advantage, yet still chose to bug out.

[Preacher]

Question: What result do you expect when 4 PzIVs stumble into an ambush by 4 Sherman 76s? Similarly, what result do you expect when a Tiger drives in between 2 ambushing Shermans at close range?

I believe that to expect anything other than devastating losses for the "ambushee" is unrealistic - barring some rare circumstance.

I don't see the problem. You're getting what you should get, IMHO, based on your description of the test.

Preacher

[This message has been edited by Preacher (edited 07-26-2000).]

[tank_41]

<BLOCKQUOTE>quote:</font><HR>Originally posted by Preacher:

Question: What result do you expect when 4 PzIVs stumble into an ambush by 4 Sherman 76s? Similarly, what result do you expect when a Tiger drives in between 2 ambushing Shermans at close range? <HR></BLOCKQUOTE>

Good question! Then what result do you expect

when a stuart stumble into an ambush

by 1 Tiger at close range?

However, why can Stuart choose to back-off

and retreat while Tiger has to be doomed.

BTW, I am using the example where the first

shoot from these ambushing units missed.

During the time of reloading, I think

Tiger should have plenty of time to decide

its course.

In terms of PzIVHs and Shermans shoot out,

if you tries it, you see they exchange rounds

2 to three times. Again, plenty of time to

decide to back off.

[This message has been edited by tank_41 (edited 07-26-2000).]

[Jarmo]

Good points.

The AI could use improvements, this being one.

In fact there are many, many things to improve in this game.

As good as it is.

As far as I know, improvements are being made all the time.

1.03 patch will fix many problems. Many will remain.

It's still the best wargame around. Until CM2.

I assume, many of the problems will have to wait until that.

[OsinO]

Jarmo said it best..period

[tank_41]

<BLOCKQUOTE>quote:</font><HR>Originally posted by Jarmo:

It's still the best wargame around. Until CM2.

I assume, many of the problems will have to wait until that.<HR></BLOCKQUOTE>

Jarmo, I completely agree on this. Although

I have posted many negative things on CM,

I personally emailed a thank you letter

to BTS to thank and congradulate them on

this piece of art

[Jarmo]

Hey, I've about 300 postings on me.

In (at least) 100 of them I go like a queen bitch about the lack

of this or that, or about why a feature works in a wrong way.

[Priest]

Hey Preacher is above me. How spiritual is that? Anyways I am not going to rip you at all this time Tank_41. Actually I am quite impressed with the level of your testing. As far as the gun damage goes I agree that there is some problems but that has already been covered and I think BTS knows about it.

These are my thoughts on the rest.

1. You are testing an ambush situation which means that confusion would reign and while I admit shooting back does not quite simulate this either, immediate retreat and and popping of smoke while the proper action may be just as hard to coordinate. There might be a lag time. Also experience probably comes into play. Again I am just extrapolating anf do agree with certain points you make. Also how easy is it to put a Tiger in reverse. Lastly if you are traine d from the begginning that your vehicle is the king of the battlefield it could be the game showing German "arrogance". Again a little weak I know but....

2.) I think the above also has some meaning in your second point. How many meters are we talking about? If it 200 or less then shooting may be (?) a better tactic. At that range allied guns are dangerous and also pretty accurate. If they miss the opening shot and it's equal I can see a commander not ordering a cut and run. Even though this may be a bad decision in retrospect I can see it happening in the heat of battle, and that is what Tac AI is supposed to represent.

Again I am kind of playing a partial devils advocate but good disccussionary topic.

[Sitting Duck]

tank_41 - excellent analysis and testing!

Do you (or anyone) have any ideas on how to improve this particular portion of the TacAI?

It would seem that there are at least 2 questions that would have to be answered:

1. How do you determine when a theoretically superior tank needs to make an evasive move? Stated another way, what situational factors should be considered and how should they be weighed?

2. What evasive move does it make?

It seems like question 2 is already answered satisfactorily, at least in the case of Ami tanks, or more generally, when the ambushed tank is theoretically inferior to the ambushing tank(s).

I guess the simplest situation is the gun damage case (scenario 3). In such a case, the Tank with gun damage should probably behave 'as if' it were infantry or half-track when faced with enemy armour, yes?

The more complicated situation would be Scenario 1 where the lone Tiger is ambushed on either side by 2 Sherm 75's.

In this case, the Tank vs. Tank logic might dictate that the Tiger stand and fight (Tiger>2xSherm75), but the situational factors might suggest an evasive action. Perhaps another way to say this is that a non-"arrogant" tank commander would take an evasive action of some sort. [Not sure I buy the German "arrogance" modeling suggested by Priest - but I am constantly amazed by the detail of this game!]

What logic would you (or anyone) suggest?

Maybe this could even be a preference item like Fog of War? I want aggressive Tanks when on the offensive! I want conservative tanks when on defense?

[This message has been edited by Sitting Duck (edited 07-26-2000).]