Weak enemy tank AI

[MikeyD]

One small thing, are you running the v1.03 Beta? I believe 'cower' was tweaked a bit so Russian will stand and fight a bit more. If you're a CDV guy you'll just have to wait for the patch.

T-34/76 firing on a Tiger at ANY angle and ANY range is a very iffy proposition. The appearance of the invulnerable Tiger sent Russian tank designers into a panic. You've just got to accept the fact that T-34/76s vs Tigers is not going to be a fair fight no matter how you slice it. If the AI is refusing to fire because there's 0% chance of a kill, forcing your unit to fire will still mean its 0% chance of kill.

[Eric Alkema]

Factoring the enemy facing into the Tac-AI decision to run away seems like a good idea to me. Just another factor in the "perceived threat level."

[JasonC]

"sent Russian tank designers into a panic"

I really don't care if the tank designers panic, it is the freaking tank drivers that are annoying the heck out of me. I'd prefer if the game designers panicked.

Shooting is better than not shooting. Plenty of Tigers ran onto the field at Kursk, and 2 weeks in only a handful were left running. Hail fire is the major reason why. If there is a round in the tube, shoot. No penetration chance is not no damage chance. Plenty of tanks that could not be penetrated were still disabled, on all sides (Matidas and Char Bs, KVs, Tigers).

When the player does absolutely everything right, and all the Tac AI has to do is pulling the freaking trigger, or at the utmost continue an ordered move for less than 30 seconds and pull the freaking trigger, and instead the Tac AI wrenches control out of the intelligent player's hands to do something completely moronic and unhistorical, it has nothing to do with WW II tank designers and everything to do with present day game designers.

Tank cower is broken. Fix it please.

As a general rule, the Tac AI should ignore the player's orders to substitute its own, on the fly orders only if the SOP it then follows is right at least 9 times out of 10. The present tank reversing behavior is instead wrong 9 times out of 10.

Incidentally, infantry going to "sneak" toward the "nearest" cover when shot in open ground is another example of a broken Tac AI SOP. It is OK if the nearest cover is within 20m, perhaps. (Advance might work with cover within 50m). Otherwise hitting the deck completely is a far better SOP.

[walpurgis nacht]

Originally posted by JasonC:

Ahistorical horsefeathers and side splitting hysterical. Tank cower is broken. Fix it please.

You are not doing it right. I setup a small example, an ideal t34s to kill a tiger situation. 4 t34s, late model with tungsten (although I can do this same example without tungsten and still win every time), all regular experience, against 1 regular experience tiger. September '44. I ran it 6 times. Dead tiger every time. No t34 cowering . . not once. With t34 losses being either 1 each time, or none. If you want to try it with me TCP/IP. Let me know . . . so we can stop this tank "cowering" is broken nonsense. Just post a time to meet. If you can do it in the next 28 hours I can stop whatever I'm doing and run it with you. This will take less than 2 min of your time once we are connected, as t34s are already in position.

Generally speaking, when my tanks cower, it is under circumstance where they will likely not win, and so, they are often "saved" by what you call "cowering".

[ May 10, 2003, 01:20 AM: Message edited by: Walpurgis Night ]

[Apache]

I have encountered similar problems and related ones. As a for instance, lone Stug parked up in a village dodging buildings. Large number of light tanks, supported by smg infantry on carriers given a rapid advance order breaking from woods to get to the cover provided from hills and the outskirts of the village on their side. Result, vehicles break from the woods as ordered (probably 20-25 in total) Stug + an ATG opens up and whole lot immediately panic and start spinng round getting pretty much anhialated. Dumb!!! There is no way the Stug ATG could have been targeting more than 20% of the vehicles (which advanced on a broad front with 'Fast move' orders) and if they'd continued (as I'm sure they would in reality) I'm sure 50-60% at least would have made the cover, allowing them to adopt better firing and troop deployment positions.

While it's doubtless to much to expect the AI to do a 'how many of us are there' calculation (thus determining the lower chances of getting KOd) the facing routine suggested would certainly have allowed the bulk of the vehicles to get through.

I tend to agree with the suggesting that it needs fixing.

[walpurgis nacht]

Originally posted by Apache:

I have encountered similar problems and related ones. As a for instance, lone Stug parked up in a village dodging buildings. Large number of light tanks, supported by smg infantry on carriers given a rapid advance order breaking from woods to get to the cover provided from hills and the outskirts of the village on their side. Result, vehicles break from the woods as ordered (probably 20-25 in total) Stug + an ATG opens up and whole lot immediately panic and start spinng round getting pretty much anhialated. Dumb!!! There is no way the Stug ATG could have been targeting more than 20% of the vehicles (which advanced on a broad front with 'Fast move' orders) and if they'd continued (as I'm sure they would in reality) I'm sure 50-60% at least would have made the cover, allowing them to adopt better firing and troop deployment positions.

While it's doubtless to much to expect the AI to do a 'how many of us are there' calculation (thus determining the lower chances of getting KOd) the facing routine suggested would certainly have allowed the bulk of the vehicles to get through.

I tend to agree with the suggesting that it needs fixing.

What was the distance from your light AFVs to the StuG?

I don't think the "cowering" is broken any more than, following your logic, infantry advances that are halted by enemy fire that sneak away to random cover, are broken. Seems like the same thing conceptually to me.

Both you and Jason have pointed out that when tanks do cower, they start turning and randomly making them selves more vulnerable. What is happening, is the "cowering" AFV makes a decision to retreat to the nearest place where the threat will not have LOS to it. If that is right next to it on the left, then it will have to turn first to reverse there. That AI will not always choose the best way to "retreat" . . . just as advancing infantry that decide to retreat, don't always "sneak" to the best cover.

[Apache]

So you think it quite reasonable that a couple of companies of light tanks, if ambushed by an ATG when moving rapidly to cover in front of them, will merely halt and rotate in circles or waste time going repeatedly back and forth a few feet rather than carry on with their high speed dash for the cover they were making for in the first place :confused: I'm not aware of too many infantry units able to move at at 15-30 mph in order to get to more suitable defensive positions?

The problem as I see it is that the Tac AI is overriding sensible orders when there is no need to do so. I accept the tanks being shot at, and those in the vicinity may well take some other drastic (if not perhaps a little foolish) action but I see little sense in vehicles 100m to their left and right adopting a 'We'll stop here and just wait so you can pick us of too when you've finished targeting the others'. That to me is somewhat illogical.

[ May 10, 2003, 07:44 AM: Message edited by: Apache ]

[SteveS]

I just made the following mistake in a PBEM, but the situation does, I think, illustrate a small problem with the tank Tac-AI.

I've given a veteran Valentine IX a HUNT command to move it forwards into a small valley in the middle of the map (I should have done a FAST MOVE). 20 seconds into the movie a Tiger appears on a far ridgeline, 880 metres away. The VIX spots the Tiger, a yellow line appears between the two, and the Tac-AI stops the forward motion and puts the tank into reverse back up the slope towards what it thinks is a safe position.

40 seconds later the Tiger has fired 3 shots at the VIX and missed. The VIX has at no time targeted the Tiger and tried to fire back. At the beginning of the next turn I attempt to redress my mistake by cancelling the reverse order it has (there's no way it is going to make it back up the slope as it would take over a minute to reach safety that way) and give it a fast move towards an out of LOS position in the valley floor. Checking the targeting stats I see the VIX is 883m from the target with a hit percentage of 2% and a kill of none.

The next movie starts. The VIX immediately targets the Tiger, raises its gun, but is blown up within 3 seconds.

Surely any real life veteran tank crew would have tried to fire back in the first 40 seconds, even if there was very little chance of hitting. I don't see the logic in preceiving a threat, going into reverse, but not firing back when under fire because of what...they want to preserve their ammo ?

Of course it was my mistake that led to the death of the tank, but in this particular situation the AI compounded the problem.

Fantastic game though

[Ron]

No, I am not seeing either what Jason and Apache are describing. Are you guys using the 1.03 patch?

I will describe a TCP game last night as an example. The date was September 42 and my opponent had two vet Tigers(Early) to my 9 T-34s(1942/no Radios), 3 Greens throughout the platoons, the rest Regs, no APCR. I had a relatively very open 800m distance to close in order to get to the Tiger's rear/flank. Using covered arcs, Fast move and what cover there was, I managed to get 4 T-34s into the Tiger's rear while they were 'distracted' to the front. Not once did the TacAI override my Fast move orders, even though the T-34s were in LOS of the Tigers practically the whole time. Even after the 4 T-34s in the rear of the Tigers ended their movement orders (range 50 - 100m), only one of the T-34s decided to 'cower' and seek safety, the remainder stayed and fought. The one T-34 that did 'cower' didn't rotate and reverse, it moved fast towards the nearest cover. Result was two Abandoned Tigers versus 6 KO'd T-34s, seems realistic to me considering the technical disparity between the AFVs.

If the TacAI always 'cowered' and didn't fight then I would agree there is a serious problem, but from what I have seen in the games I have played with 1.03 that isn't the case.

Ron

[aka_tom_w]

Jason C are you playing in the latest v1.03c?

(just curious)

Are you playing aScenario?

You can provide the name of it so I can play the Russians vs the Computer AI?

I would like to play test your example in the game against the AI

if possible?

-tom w

[ May 10, 2003, 12:33 PM: Message edited by: aka_tom_w ]

[walpurgis nacht]

Originally posted by Apache:

The problem as I see it is that the Tac AI is overriding sensible orders when there is no need to do so. I accept the tanks being shot at, and those in the vicinity may well take some other drastic (if not perhaps a little foolish) action but I see little sense in vehicles 100m to their left and right adopting a 'We'll stop here and just wait so you can pick us of too when you've finished targeting the others'. That to me is somewhat illogical.

Well i guess we have a different idea of what are "sensible orders". Running LFVs through the path of a StuG may seem sensible to you, but quite obviously not to the LFVS.

How far, specifically, was the "fire window" the Stug had on them? Also, which LFVs?

[aka_tom_w]

I am using the latest v1.03c

I agree COMPLETELY with Ron and his post and comments

I have not seen the behaviour Jason C is complaining about. (I play the Russian ALL the time, but I have not tried to ambush Tiger's with T34's in CMBB yet)

-tom w

Originally posted by Ron:

No, I am not seeing either what Jason and Apache are describing. Are you guys using the 1.03 patch?

I will describe a TCP game last night as an example. The date was September 42 and my opponent had two vet Tigers(Early) to my 9 T-34s(1942/no Radios), 3 Greens throughout the platoons, the rest Regs, no APCR. I had a relatively very open 800m distance to close in order to get to the Tiger's rear/flank. Using covered arcs, Fast move and what cover there was, I managed to get 4 T-34s into the Tiger's rear while they were 'distracted' to the front. Not once did the TacAI override my Fast move orders, even though the T-34s were in LOS of the Tigers practically the whole time. Even after the 4 T-34s in the rear of the Tigers ended their movement orders (range 50 - 100m), only one of the T-34s decided to 'cower' and seek safety, the remainder stayed and fought. The one T-34 that did 'cower' didn't rotate and reverse, it moved fast towards the nearest cover. Result was two Abandoned Tigers versus 6 KO'd T-34s, seems realistic to me considering the technical disparity between the AFVs.

If the TacAI always 'cowered' and didn't fight then I would agree there is a serious problem, but from what I have seen in the games I have played with 1.03 that isn't the case.

Ron

[ May 10, 2003, 12:21 PM: Message edited by: aka_tom_w ]

[aka_tom_w]

in this case I would say the GAME is modeling the reality that the VIX tank crew knew they where dead meat!

"Checking the targeting stats I see the VIX is 883m from the target with a hit percentage of 2% and a kill of none.

The next movie starts. The VIX immediately targets the Tiger, raises its gun, but is blown up within 3 seconds.

Surely any real life veteran tank crew would have tried to fire back in the first 40 seconds, even if there was very little chance of hitting. "

the TAC AI knows this too!

"stats I see the VIX is 883m from the target with a hit percentage of 2% and a kill of none"

KILL = NONE

means get out of there fast

they got caught in the open with no chance of killing that other tank.

I think it is fair to say they new that firing the shot in that 40 sec was both a waste of time and ammo.

BUT yes trying to reverse BACK up the slope was equally foolish BUT it is possible they paniced, BUT then again now in CMBB the game tells you when the crew panics.

-tom w

[ May 11, 2003, 08:27 AM: Message edited by: aka_tom_w ]

[mchlstrt]

I'm a mediocre player at best, but this is the one problem I noticed by myself before I read about it. The experiences I had with it put it on the level of the 'unhittable gun' bug (which I've since had completely stop my attack. I'm NOT in 1.0.3). I absolutely believe it is a very undesirable situation.

There was language in the 1.0.3 fix list that indicated BFC may have done something about this, but it's not clear, & I've never seen this issue addressed directly by any of the B-boys. If they're aware of the issue, have given it a look & taken their best shot at it then I'm more than happy. If they were not even aware that anybody thought this might be broke that would concern me a bunch.

Maybe someone with a good understanding of this issue could test it under 1.0.3. I encourage JasonC, if he's got the right version. I'll dl it & see if I can tell much, but there are a lot more qualified people than me around here for this.

strt

[Apache]

Walpurgis - It may come as a shock to you but if combat units were assigned a mission to attack a piece of terrain (which unfortunately was not always blessed with covered lanes of approach - it does happen you know), sometimes they had to get off their backsides and DO IT! I presume you think it sensible that an entire asault is held back merely because a Stug might take a couple of vehicles out en route? Get real.

I am describing what happened in a scenario, without the surrounding contextual information which is to a large extent irrelevant. I CONSIDER it unrealistic. OK!

[ May 11, 2003, 04:14 AM: Message edited by: Apache ]

[walpurgis nacht]

Originally posted by Apache:

Walpurgis - It may come as a shock to you but if combat units were assigned a mission to attack a piece of terrain (which unfortunately was not always blessed with covered lanes of approach - it does happen you know), sometimes they had to get off their backsides and DO IT! I presume you think it sensible that an entire asault is held back merely because a Stug might take a couple of vehicles out en route? Get real.

I am describing what happened in a scenario, without the surrounding contextual information which is to a large extent irrelevant. I CONSIDER it unrealistic. OK!

Apache, no reason to get worked up there buddy, I'm just trying to help. The reason I asked the specifics of the situation is because I have played 3-4 QBs a day since CMBB came out, quite literally, and I don't have these problems.

If the window of your rush was too large, than

they are being realistic. Just as infantry disobey your orders and drop when taking MG fire.

There is often a better way.

[ May 11, 2003, 04:42 AM: Message edited by: Walpurgis Night ]

[SteveS]

Originally posted by aka_tom_w:

in this case I would say the GAME is modeling the reality that the VIX tank crew knew they where dead meat!

"Checking the targeting stats I see the VIX is 883m from the target with a hit percentage of 2% and a kill of none.

The next movie starts. The VIX immediately targets the Tiger, raises its gun, but is blown up within 3 seconds.

Surely any real life veteran tank crew would have tried to fire back in the first 40 seconds, even if there was very little chance of hitting. "

the TAC AI knows this too!

"stats I see the VIX is 883m from the target with a hit percentage of 2% and a kill of none"

KILL = NONE

means get out of there fast

they got caught in the open with no chance of killing that other tank.

I think it is fair to say they new that firing the shot in that 40 sec was both a waste of time and ammo.

But Tom, what do they have to lose ? They're in an extreme life threatening situation and by not firing the chances of getting a lucky gun hit or whatever are now ZERO. Also, no human crew would accurately assess the probabilities the way the AI does.

There are obviously circumstances when you would not want a threatened tank to take low probability shots, but when the threat is extreme as in this case it should do so.

BUT yes trying to reverse BACK up the slope was equally foolish BUT it is possible they paniced, BUT then again now in CMBO the game tells you when the crew panics.

-tom w

They didn't panic and in fact the quickest route to safety would have been to switch to FAST MOVE and go forward, which I guess a veteran crew might have expected to do in real life.

[Ron]

Originally posted by Apache:

Walpurgis - It may come as a shock to you but if combat units were assigned a mission to attack a piece of terrain (which unfortunately was not always blessed with covered lanes of approach - it does happen you know), sometimes they had to get off their backsides and DO IT! I presume you think it sensible that an entire asault is held back merely because a Stug might take a couple of vehicles out en route? Get real.

I am describing what happened in a scenario, without the surrounding contextual information which is to a large extent irrelevant. I CONSIDER it unrealistic. OK!

You haven't said which version you are using. I don't know exactly what transpired in your game but it must be a 'rare' occurance for I have never experienced that either, whether patch 1.01 or 1.02 or 1.03.

I setup a test scenario this morning with 1.03 to roughly simulate what you described. A lone Stug and ATG facing 20 Light Tanks and HTs who are Fast moving from cover to cover towards the German positions. I ran it 10 times and not once did the TacAI override my movement orders. They all attempted to complete my orders, no matter the losses mounting. Even when I cancelled their movement orders and gave them new orders, leaving them exposed in the open, they simply waited out the delay before proceeding on the new path.

The only time the TacAI took over was when they had no movement orders and were exposed to the enemy. So again I don't know the circumstances you experienced, but your complaint is unfounded. They will 'get off their backsides and DO IT'.

Ron

[Andreas]

Originally posted by Apache:

I presume you think it sensible that an entire asault is held back merely because a Stug might take a couple of vehicles out en route? Get real.

Erm - I have heard that the Soviets had standing orders to break off an attack when facing Stugs, and move the axis of attack to another point, i.e. off the CMBB battlefield. Can't give you a source, because I never asked for it myself. If Kip reads this, he maybe able to dig it out. We are talking about the 75L43/8 armed, 80mm frontal armour type variety here. Also have a look at the special orders published by Konev in December 1942 following the Mars desaster - John Waters has posted them in another thread a few days ago. Those dealt with the 50mm frontal armour 75L24 armed variety.

[Ron]

Originally posted by SteveS:

But Tom, what do they have to lose ? They're in an extreme life threatening situation and by not firing the chances of getting a lucky gun hit or whatever are now ZERO. Also, no human crew would accurately assess the probabilities the way the AI does.

There are obviously circumstances when you would not want a threatened tank to take low probability shots, but when the threat is extreme as in this case it should do so.

They didn't panic and in fact the quickest route to safety would have been to switch to FAST MOVE and go forward, which I guess a veteran crew might have expected to do in real life.

Yes you should have used the Fast movement order. From the manual - "Hunt - Tells the vehicle to look for armoured targets and AT guns. When these targets come into LOS, the vehicle will stop and engage if it has a decent chance of causing damage to them,...

Obviously in your example you had zero chance of doing any damage and the TacAI decided retreat was preferable to engaging and firing.

You have to realize the TacAI does not recognize elevation changes as cover, only 'true' cover like woods, buildings etc. It is a game limitation but there you go.

In your example you may have been better off letting the TacAI complete its Reverse order instead of intervening. By cancelling its order and giving a new one, your tank became stationary, even for a few seconds, and in consequence much easier to hit. A moving target is better for survivability.

Ron

[Vanir Ausf B]

Originally posted by Walpurgis Night:

I have played 3-4 QBs a day since CMBB came out, quite literally

Someone call the people at Guinness.

[JasonC]

Starting a dozen tanks in LOS of a pair of shooters and charging to 50m away (1) isn't realistic at all and (2) won't reveal the problem. Cower occurs at the instant LOS is first achieved. You see it when "breaking from cover", when an enemy crests a hill, etc. If you start the whole thing already in LOS, you will not see it. It is also probably more common on "hunt" orders.

It is precisely when you are realistically making full use of terrain that it is most debilitating. It penalizes skill - Walpurgis, sorry but your mass suicide-exchange rushes don't count as "skill", even if sometimes they are tactically worth it in the fights you are in.

When a single T-34 is driven so well that it should kill the Tiger without loss (by getting the drop on it, from a killing range and angle), without the shooter himself even being shot at, that counts as skill. As little as possible is even risked and nothing need actually be lost.

The "retreat" SOP is designed to retreat from supposedly "hopeless" match ups, but they are only "hopeless" if a smart player hasn't properly lined them up. There is nothing "hopeless" about shooting at a tank facing the wrong direction. The present cower routine does not take enemy facing into account, and that is what most needs changing. It should also fire the round in the tube regardless i.e. only kick in during reloading.

The particular Tiger incident I last reported was in an armor QB using 1.02, not a scenario. But I've seen it in dozens of cases, with T-34s closing around cover, SU-85s and SU-152s doing cresting drills, even IS-2s sometimes refusing to duel (I've seen an IS-2 cower from a Pz IV), both in QBs and in scenarios.

"Iron roadblock" as the Germans will show plenty of it, if you properly use the scattered tree areas and try to "hail fire" the KV, shelter the SPWs but still get them to run through the gaps the KV can see, etc. It is still possible to "convince" the Germans to use hail fire, but after dozens of cowers interfer with the attempt.

[manchildstein (ii)]

along these same lines, last night i had an spw 251/10 with 37mm and 7 rounds of tungsten attack a 1943 t/34 from the rear... the spw rushed ahead and fired a couple rounds of tungsten as it went... interestingly, when it got within about 120 meters and the end of its move orders... and when it still had 5 rounds of (t) left, it began firing regular ap as the t34 turned around... i was thinking if the spw had just fired its t rounds then things might have been different... as it was, the t34 spun its turret around and waxed the spw...

what first clued me into the halftrack using standard ap were the ricochet sounds as two rounds bounced - one first off of the rear, then the second off of the side turret of the t34... the spw had the highlight and sure enough, when i glanced down it had switched to firing ap... arg

one behavior i have not seen modelled in the game is some sort of 'desperation fire'... sometimes i'd like to see an overmatched vehicle stand fast, and not only continue to fire its main gun, but to fire all of its machineguns at the threat... just like some drivers and crews would panick and retreat, i believe others would simply attempt to empty all of their weapons at the threat instead... trying desperately to get some bullets into vision slits or down the gun barrel of the threat...

anyway, my spw not 'dipping into' its last 5 rounds of t ammo when it most needed them... that didn't make any sense to me... now this was with 1.02 so maybe things will have changed with 1.03...

[walpurgis nacht]

Originally posted by JasonC:

Starting a dozen tanks in LOS of a pair of shooters and charging to 50m away (1) isn't realistic at all and (2) won't reveal the problem. Cower occurs at the instant LOS is first achieved. You see it when "breaking from cover", when an enemy crests a hill, etc. If you start the whole thing already in LOS, you will not see it. It is also probably more common on "hunt" orders.

It is precisely when you are realistically making full use of terrain that it is most debilitating. It penalizes skill - Walpurgis, sorry but your mass suicide-exchange rushes don't count as "skill", even if sometimes they are tactically worth it in the fights you are in.

Why do you assume in my example that the t34s initially have LOS? . . . . they do not. And at the moment of coming into LOS, they do not cower, as you suggest.

What's not realistic about a tank charge on the flanks? The American Shermans took on tigers by shoot and scoot from the front (often losing their armor) to distract the tiger, while other armor rushed in on the flanks for a kill.

Getting your t34s into position without getting killed often requires a great deal of "skill". Timing the flanking assualt correctly= skill. Grand strategic flanking manuevers . . .I'd say that's at the heart of what it means to use the t34 well . . . maximizing it's strengths.

And what about it inspires you to call it a "mass suicide-exchange rush"? As I said, 1 or no t34s are killed. The example I setup is not only realistic (in CMBB terms), it works almost all the time, with no cowering.

Of course, you're entitled to continue thinking your idea of skill is right and anyone who thinks differently is wrong. Just as, you're entitled to continue losing your t34s to tigers, while others succeed in killing them.

[ May 11, 2003, 01:54 PM: Message edited by: Walpurgis Night ]

[JasonC]

Walpurgis, I was reacting to the following narrative, which looking back may have been from Ron (and quoted later) not from you. He described a 9 T-34 vs. 2 Tiger attack over 800m of open ground -

"relatively very open 800m distance to close in order to get to the Tiger's rear/flank. Using covered arcs, Fast move and what cover there was, I managed to get 4 T-34s into the Tiger's rear while they were 'distracted' to the front... Result was two Abandoned Tigers versus 6 KO'd T-34s"

I don't call losing 6 T-34s to kill 2 Tigers in a 9 on 2 "skill", I call it a "suicidal exchange". It may sometimes be warranted by the tactical situation, if nothing better is possible, as I carefully allowed.

"Skill" is driving good enough that you get the shot from the range and angle you need without being shot - either not even shot at (best case), or not hit because exposures were brief, at range, on fast move, etc.

I now see you are saying basically the same, that the standard you are holding yourself to is use of cover in the approach good enough to get to the firing location without losing multiple tanks along the way.

That is the right standard. And it is where I've encounter the cower problem, involving as it does frequently leaving LOS and always entering it, often with just 1-2 AFVs. If you say you simply have never seen it, then we seem to be playing 2 different games.