T-34s running away....

[bigred]

I was playing an email game a while back and this is what happened....

I have five t-34 tanks sitting on a top of a hill (3 in the hull down position)waiting to see if/when some German armor comes rumbling along. Excellent visibility and sunshine.

Finally, one lone Panther rolls into the open at about 900 meters and....to my shock and amazement, four of my T34 imediately (without even taking a shot!)going into reverse and proceed back down the other side of the hill and "one" T34 takes one shot, misses and then immediately goes into reverse back down the other side of the hill.

Now....what in the heck is that???

Five T34's setting there like an ambush waiting for something to shoot at..."one" Panther casually strolls into view and all five run away??

Surely, there simply "must" be something wrong with the AI files or something else?

Help!

[Schrullenhaft]

Are these T34-85's or T34-76's ?

The AI is performing its "armor morale checks" on an individual basis. There is no morale benefit with greater numbers in CMBB (or any CM game for that matter) that affects the willingness to fight or retreat. Each T34 in this case is measuring its chances against a Panther and backing down. This is a defensive/protective mindset for the armor. I believe that the AI is measuring the capability to be penetrated by the enemy armor and using that as the primary basis to back down or not.

Hopefully CMx2 will have a more complicated routine to determine the willingness (or idiocy) to fight or run. I'm fairly certain with the work on relative-spotting, unit-cohesion, unit-awareness and memory, communication and more complex chain-of-command relationships that will be part of CMx2 should add a whole lot more to the decisions of units in CMx2.

[ParaBellum]

What kind of T-34s are they? If they are the ones armed with low-velocity 76mm guns then they'd better back off when facing a Panther at 900m. Or, at any range. And even if they are armed with 85mm guns taking a Panther head-on at that range is not the smartest thing either.

What your AI guys are doing is assessing the situation. They realize that at the given range they have only a very small chance of killing the Panther while the Panther has a VERY GOOD chance of killing them. That's why they retreat.

5 T-34s sitting on a hill in front of a Panther at 900m isn't really a good 'ambush'. Maneuver your tanks so that they hit the Panther's flanks and they won't retreat.

[JasonC]

Horsefeathers, the lot. He didn't say anything about the Panther facing them. And it doesn't make any difference. Test it. As for the "making assessments" story, put 5 Pz IV longs on the hill and stroll one IS-2 into LOS. The IS-2 will be the one that cowers - and it will also die.

[bigred]

Schrullenhaft/JasonC;

Aha!

Thank you both so very much for your informed and knowledgeable replies. I understand now that I could have had 500 T34's on that hill and quite possibly they all might have run away.

Hopefully in the future there will be some type of option available that gives one the possibility of increasing the unit's willingness to "stand and fight" as this hill was simply critical for this scenario.

As three or four of those T34's (I honestly don't recall for certain what version of T-34s they were) were under one command I "assumed" there would be some type of group calculation.

Of course, we all know what happens when one "assumes".

Once again, thank you both so very, very much for sharing your expertise.

[JasonC]

The right procedure would be for the tanks to check the threat's facing and/or targeting. When it saw the gun within 30 degrees of arc of itself, say, it could check morale or reverse. Also, they should only undergo this check if their own readiness reads "reloading" - that way they would first fire the round in the tube. That is all one would need for realistic teamwork tactics, hail fire against stronger enemies, etc.

Since we don't have them, the thing you need to know is that fast move conquers "tank cower" and nothing else really does. A shoot and scoot with a forward scoot position can also work - both moves are "fast". That gives a firing halt. (Which is important, because you won't hit anything beyond point blank while actually fast moving). Put the end waypoint behind cover, so it won't cower while there. Stationary, hunting, or reversing vehicles will cower. Perhaps move too, I haven't really tested that one (don't use it much).

[bigred]

JasonC;

" The right procedure would be for the tanks to check the threat's facing and/or targeting. When it saw the gun within 30 degrees of arc of itself, say, it could check morale or reverse.

Also, they should only undergo this check if their own readiness reads "reloading" - that way they would first fire the round in the tube. That is all one would need for realistic teamwork tactics, hail fire against stronger enemies, etc. .

"

Well said, my kind sir.

That's precisely what initially blew my mind the worst. To not even shoot when that lone Panther casually strolled into view (easily within all five's 30 degrees of arc) was....frustrating. Also all five had been just sitting there for at least five turns doing nothing so reloading certainly wasn't an issue.

Good point regarding the "shoot and scoot" scheme. As it's at present all that's available I'll definately use it. It certainly beats the alternative!

Thank you once more!

[Bigduke6]

The recommended Soviet armor tactic for dealing with Panthers or Tigers was: the targeted Soviet tank target backs outta there as fast as it can, the others wait for a target. Most likely thy will get a semi-flanking shot on the German. If they don't they wait wait until the Tiger/Panther advances, and then shoot.

The idea is that if you get several unhindered hits the kittie is going to back up or turn a flank, or maybe even you get running gear or a weak point or something. If the Soviet tanks are spread out a bit, then the Tiger/Panther has a good chance of showing a flank.

CMBB makes this tactic pretty difficult as the mere gaze of a Tiger/Panther makes a Soviet tank cower. I can understand backing out of a 1-1, but Soviet tankers ought to be progressively braver, the more fire their buddies are putting on a Tiger/Panther. But then, I think Soviet AT weapons were more powerful in real life than they are in CMBB.

OBTW You can make a pretty good case Soviet tanks in CMBB are being very rational by cowering - they're not facing real Panther/Tigers, but much more dangerous CMBB ones.

[Kingfish]

Originally posted by JasonC:

As for the "making assessments" story, put 5 Pz IV longs on the hill and stroll one IS-2 into LOS. The IS-2 will be the one that cowers - and it will also die.

So, what your saying is that the proper response by the IS-2 is to stand its ground?

I guess it means nothing that in the time the IS-2 gets a shot off, reloads, and shoots again those 5 Pz IV longs would have unloaded 20+ shots on him, each one capable of penetrating his armor.

[Berlichtingen]

Originally posted by bigred:

Surely, there simply "must" be something wrong with the AI files or something else?

Nope, the AI was smart. Even a T-34/85 is going to have problems taking a Panther out from the front at that range, and the Panther will dice-n-slice them

[dieseltaylor]

It would be nice in RL to know that you are faced by a solitary Panther. Well known that they always went around singly and that if you could not see the the others in no way meant that they were about to crest the hill or fire from some undiscovered position.

Obviously therefore sensible to duke it out and try and strike lucky before it terminates you. I have a great respect for the AI's intelligence of when to cower. It does mean suicidal orders that are actually actioned occur far less than some players would want.

[JasonC]

Kingfish - the IS-2 should (1) obey its orders (2) pick a target and fire at it if there is a round in the tube, from any halt (3) if it then finds the tube is empty and a gun is pointing at it (not "in LOS", within 30 degrees of aligned), then it can get out of there if you like, minimizing time spent and keeping forward facing.

Reversing immediately is usually the stupidest available response. They don't fire, if they did they wouldn't hit because they are moving, they don't break LOS because they pick a silly direction and spend half the time rotating in order to present a flank etc.

Why not make Russian tanks that blow up when you look at them? Would that strike the Signal magazine fans as more realistic? Calling it smart is ridiculous, my mother could drive them better. The present SOP is far more suicidal than shooting.

The designers just took some operational advice about avoiding enemy armor to hit where they ain't way too tactically and literally. It is a mistake. Correctable with the right SOP, which I have given repeatedly (fire whats in the tube, pay attention to facing etc).

It is also of a piece with other AI subroutines (infantry as well as tanks) that are far too focused on solving things by movement choices rather than using fire or coordination. It is simply written to always reach for a new waypoint first, and therefore gets wrong the solid half the time when that isn't the right answer. When it is the strat AI you can play a human instead. When it is the tac AI you can't correct it, so it is far more maddening.

[ April 13, 2005, 05:53 AM: Message edited by: JasonC ]

[MikeyD]

I wonder. What if the situation were moved-up to VERY late war and T-44s substituted for T34s? I bring up the T-44 because it and the IS-3 are about the only things in the game capable of reliably taking a Panther gun hit. If THEY cower we've got a problem!

Unfortunately for the historical pureist, I rather doubt there were any T-44/IS-3 fights with Panthers dring the war.

[JasonC]

In reality, a Panzer IV had trouble with a W armor Sherman at ranges less than a kilometer. In BB, IS-2s cower from them. Anybody who thinks that isn't a realism problem is whacked.

[undead reindeer cavalry]

it's one of the more absurd design flaws in CM, underlined by an armored car that has no objections to charging a column of Tigers. the whole "feature" is totally counterproductive as is.

it's frustrating that such a brilliant game has been partially ruined by silly design decisions.

[MikeyD]

"partially ruined"

Well... That's a tad harsh.

What it does is make the game frustrating for people brought up on shooter games, where the approved form of combat is to charge forward with both guns blazing.

I played a game of Stugs vs T34s last night against the AI and to my surprise the battle see-sawed back and forth until those 'cowardly' T34s had knocked out all my Stugs and were in the process of overunning my infantry. Not half bad for a partially ruined game!

[Xipe66]

I wonder. What if the situation were moved-up to VERY late war and T-44s substituted for T34s? I bring up the T-44 because it and the IS-3 are about the only things in the game capable of reliably taking a Panther gun hit. If THEY cower we've got a problem!

They still cower, maybe not as bad though (up to 60% will stay at least for a while, or until they notice being targetted).

Setting up armor cover arcs and experience seems to play a role too, but without extensive testing its hard to say how much.

At 1040 m the T-44:s have Kill: Rare while the Panther has Kill: Good.

This is in May 1945.

In the example at hand (July 1943), the T-34's have Kill: None, while the Panther has Kill: Very Good.

Despite the T-34/44's being hull down the Panther still also had better chance to hit, it also has higher ROF (1-on-1). In one turn it's reasonably that it will take out 2-4 T-34's - would you stick around for the odds this presents?

Taken together I think what should be blamed is the tac AI's inability to take damaging hail fire into consideration; with lots of luck 5 T-34's might get a disabling hit - in the example at hand it seems rather prudent not to gamble that way. I.e. they're wiser for the fact that they disobeyed that particular order - maybe if there had been 10 of them they should stick around, take some losses and hope for a gun hit or for demoralizing the Panther...

However in the example with the T-44's I think they should stay around if ordered so and in sufficient numbers (5-6+), the Kill: Rare has a greater chance of partial penetrations which might shock or route the Panther - or with some luck a lucky penetration that will finish it.

Unfortunately for the historical pureist, I rather doubt there were any T-44/IS-3 fights with Panthers dring the war.

According to battlefield.ru (I think it was) only three T-44's actively participated on the eastern front - no documentation exists of any IS-3's seeing action against the germans.

[ April 13, 2005, 08:00 AM: Message edited by: Xipe66 ]

[JasonC]

Sometimes the routine does not kick in and everything works. Writing AIs is hard, I sympathize with the designers on that point. But the rule of thumb needs to be, do not override an explicit order from the player unless you are confident that 90% of the time your alternative will be the right decision, the one he would want.

Too many of the existing routines treat being shot at or a lack of cover as clearly unintentional, something the player must have wanted to avoid, something to be remedied by this unit changing its waypoint and speed. This probably stems from the underlying AI routine being primarily a path finder, a routing algorithm. And from an idea that the AI ought to kick in during an ambush in the middle of a minute, before the player has had a chance to react to new info.

The problem is, there are simply far more situations in which commanders expose a unit entirely reasonably, because of factors at play far beyond this individual unit's location or waypoint. A tac AI is never going to properly take all of them into account, as players can. I mean things like knowing there is actually little risk from that enemy unit because overwatch will respond, or running won't help and only firing will, or everything has already been set up to get him from the side he isn't ready for, etc. The player is quite often doing it for a sensible reason. And the tac AI should not undo it for him.

One way to handle this would be to have movement types that invoke the routines while others do not. E.g. let hunt leave engagement to tac AI discretion. Or let move to contact decide when to sideways sneak for the nearest cover. While others do not - stationary doesn't, advance doesn't, etc.

Another way is to set additional If-thens before the existing routines kick in, to make their choices more sensible. Only check tank cower if the gun state reads "reloading", for instance. Only cover panic if there is 25% better cover (absolute exposure difference) within 50m.

Another way is to have hard override flags for firing - in addition to selecting a target a unit can be ordered to fire only, no movement. Which it will only violate if it actually panics, not because of one of the routines that thinks it is helping.

A player can set a waypoint in the intervening command phase. But if the correct "waypoint" is right where they are, all he can do is halt - and the tac AI gives them a new one when what was explicitly wanted was "none, do not move under any circumstances, just shoot".

[Michael Dorosh]

Originally posted by JasonC:

Another way is to set additional If-thens before the existing routines kick in, to make their choices more sensible. Only check tank cower if the gun state reads "reloading", for instance. Only cover panic if there is 25% better cover (absolute exposure difference) within 50m.

Given existing hardware abilities, and the suggested system requirements for CMX2, I would venture to hope that this will be incorporated in the next iteration.

[Xipe66]

Simply: the downfall is that tac AI doesn't take numbers into account. For smaller scenarios this doesn't play much of a role (rather the tac AI overrides suicidal decisions) - but for larger armor engagements it plays a pivotal role: even 20 T-34's in hull down position will all cower within two turns (80% on the first turn) against a single Panther - while if they all stayed they might suffer only a few losses (one or two, tops) before damaging or routing the Panther.

Originally posted by Michael Dorosh:

</font><blockquote>quote:</font><hr />Originally posted by JasonC:

Another way is to set additional If-thens before the existing routines kick in, to make their choices more sensible. Only check tank cower if the gun state reads "reloading", for instance. Only cover panic if there is 25% better cover (absolute exposure difference) within 50m.

Given existing hardware abilities, and the suggested system requirements for CMX2, I would venture to hope that this will be incorporated in the next iteration. </font>

[undead reindeer cavalry]

Originally posted by MikeyD:

"partially ruined"

Well... That's a tad harsh.

What it does is make the game frustrating for people brought up on shooter games, where the approved form of combat is to charge forward with both guns blazing.

on the contrary. as it is, Combat Mission supports mindless charges forward. for example this very "feature" discussed here manifests only if you try to have some fancy tactics - if you just charge mindlessly towards the target you won't experience this "feature".

i don't mean that CM would be partially ruined just by this flaw. it's all the small flaws together that ruin it partially. flaws like the one discussed here wouldn't be frustrating if it wasn't purely just a simple decision on the designer's part.

still, CM is the best there is.

[Lt. Beavis]

In the original example, it's absolutely wrong that the T34s didn't fire their chambered round. All of them should have at least taken a shot.

They had height advantage, and they were hull-down. There is no reason for them not to take their shots at the Panther, for they stood to obtain top deck penetrations. In addition, the higher elevation negates the front glacis slope of the Panther.

The Panther is the one at a disadvantage in this scenario, not the T34s.

[bigred]

Lt. Beavis;

That was precisely my impression.

The combination of being hull down, a much higher position, a five to one tank ratio and the advantage of being stationary waiting for a German target to come rumbling along....well, I was just amazed they wouldn't even take a shot.

[ April 13, 2005, 12:30 PM: Message edited by: bigred ]

[Jack Carr]

Originally posted by Michael Dorosh:

</font><blockquote>quote:</font><hr />Originally posted by JasonC:

Another way is to set additional If-thens before the existing routines kick in, to make their choices more sensible. Only check tank cower if the gun state reads "reloading", for instance. Only cover panic if there is 25% better cover (absolute exposure difference) within 50m.

Given existing hardware abilities, and the suggested system requirements for CMX2, I would venture to hope that this will be incorporated in the next iteration. </font>

[Kingfish]

Here's the funny thing about this thread - BFC could go in and change the code so that the T34s would stand their ground, and within a day this board would be flooded with "T34s staying and dying..."