Retreating Monster Tanks revealed (monster pictures)

[Redwolf]

OK, I have it. Watch this.

This is a new quickbattle, importing the map I had the problem earlier

on. You will recognize the hill.

Here is one Pz IV, identified as a Pz IV lang(?) slightly in cover.

The ISU has a command to move to hulldown to exactly that spot.

%%

This is 23 seconds into the move. The ISU has almost completed the

move to hulldown. Note that the little status window at the bottom of

the screen says "Seek hull down".

%%

Next second, the move to hulldown has been completed. Note that the

status window says it is not moving, this is because it found the

hulldown position. I don't have the LOS tool in this movie, but for

me that is proof.

%%

Two seconds later, its recognizes that the Pz IV is in LOS and targets

it. The Pz IV does not target the ISU yet, as evident in the next

picture which is the same moment, just the mouse clicked on the Mk IV.

%%

Second 27, the Pz IV fires it coaxial MG at the ISU. Note the tracer

above the ISU's target line. The ISU buttons up the next moment. It is not shocked.

%%

Stil second 27, the ISU is buttoned and it reverses, with no other

shot fired, no main gun shot from either tank. The ISU-122 just

retreats after receiving the coxial MG fire.

Same moment, but focus on Mk IV. It does not have a red line to the

ISU yet. The ISU retreats before the Mk IV gets a red line.

Savegames are on:

http://65.96.131.208/tmp/isu-coward/

Note that this is hotseat and hence you get a different movie

evertime. I hope it is still useful.

[ November 27, 2002, 09:34 PM: Message edited by: redwolf ]

[Michael Dorosh]

Yeah, but it didn't crawl, right?

But seriously....did the TC know he wasn't staring down a Tiger? I mean - you the player knew, cause the tank was IDed as a Pz IV - but does that mean our TC knows?

[ November 27, 2002, 08:22 PM: Message edited by: Michael Dorosh ]

[Redwolf]

I just re-ran the last savegame.

This movie (I spare you the screenshots):

second 21: ISU spots the Mk IV (a little earlier). Mk IV has not yet targetted the ISU

Second 25: MK IV shoots MG, but still has no red line to ISU. You here the bullets on metal, but the ISU does not button up.

Second 26: ISU shoots, misses.

Second 29: MK IV fires. It still has no red line to the ISU, although it clearly shoots the main gun at it. The ISU begins reversing the same moment the MK IV shot gets off.

One thing is fishy here, and that is that the MK IV does not get a red line to the ISU although it is definitivly shooting at it. Is this intended, does extreme fog of war allow for not showing enemy target lines?

Here is the shot:

[ November 27, 2002, 08:45 PM: Message edited by: redwolf ]

[Redwolf]

Originally posted by Michael Dorosh:

Yeah, but it didn't crawl, right?

Sneak, Michel, sneak

But seriously....did the TC know he wasn't staring down a Tiger? I mean - you the player knew, cause the tank was IDed as a Pz IV - but does that mean our TC knows?

As far as I understand the CM game system there is no seperate identification for enemy units for each friendly unit. The partical identification the players sees is the same as what the units assume, all units which have seen the enemy unit in first place. Note that I am only talking about identification, not the spotting in the first place, this can be different for each unit.

[ November 27, 2002, 08:49 PM: Message edited by: redwolf ]

[Kanonier Reichmann]

Excellent example Redwolf. I think the telling thing here is that the ISU122 has identified the enemy as a PzIV and not a Tiger so fear of being overmatched should not apply. Secondly, the situation is almost a perfect ambush setup with flank shots onto a lowly PzIV yet the damn heavy Soviet AFV still backed away. That's the sort of occurences that were happenning in the game I was playing against the AI and why I thought it odd in the first place.

Regards

Jim R.

[Cull]

Still at it, eh? Made any progress with Steve yet?

That's a good example. Nice job.

I'll be lurking on the edge of my desk chair.....

[zukkov]

nice job rw. this whole self preservation thing i think has been taken to the extreme in cmbb. i believe it needs some tweaking. not only in this case where the isu definitely should not back off against a mkIV, but in other situations as well. i'm playing a pbem vs a friend and he just "charged" my lone kv1 with 6-8 mkIIIs. as soon as they crested the hill and spotted it, every one of them reversed back to the other side. this is just all to common i'm afraid and it just makes it impossible to overwhelm a small number of uber tanks with many..uh..not so uber tanks.

to conclude, i would like to join those that think bfc should relook at their auto-retreat program and tweak it a bit...

[Pvt. Ryan]

The Panzer IV id is not certain, as indicated by the "?" It may be something more dangerous. But I agree that tanks tend to back off too quickly and frequently.

[aka_tom_w]

does this need to be reviewed and perhaps tweaked before they release v1.02???

:confused:

just curious?

-tom w

[mPisi]

While I have not seen the retreating behavior in question, I have seen a tank firing its main gun without a red targeting line to its target as Redwolf says above. I can probably find the savegame if needed. I wondered if it was a FoW enhancement, but it does not seem consistent.

[Battlefront.com]

Alright, got some info now and I think I can explain things a bit more...

Redwolf

As far as I understand the CM game system there is no seperate identification for enemy units for each friendly unit.

Correct. The identification of a unit is "global" and is not differently looked at by each friendly unit. However, each unit in the game must spot an enemy before it can shoot or react to it. This includes units that are already spotted by other units, but to a lesser degree. In the case of an already spotted enemy unit, a friendly must only "notice" it instead of actually go through the whole spotting process. This is more extreme for vehicles than infantry because infantry have more "eyes". Plus, spotting on the move for a vehicle is tougher to do. Spotting on the move and buttoned up is really bad. So obviously this has a lot to do with getting shots off before the other guy.

BUT... now onto the more interesting aspect I uncovered....

KR,

I think the telling thing here is that the ISU122 has identified the enemy as a PzIV and not a Tiger so fear of being overmatched should not apply.

What amazes me is that nobody apparently took the time to look at the armor and armor penetration stats. If people had, they would know that Jim's statement above is incorrect

The PzIV's gun can kill a ISU-122 from the 500m +/- range Redwolf tested at. Perhaps not by a margin of overkill, but it is statistically very possible even without factoring in the 90% armor Quality rating. Factoring in the 90% Quatity rating for a ISU-122S (not used in this example) the PzIV also can kill it frontally. It can also BARELY kill the IS-2 family of vehicles. The Panther, on the other hand, can kill any of these at 500m no problem.

What does this mean? That for some reason people have been going on the assumption that the PzIV was not really a threat. Looking at the stats it *clearly* is. Perhaps not an overkill threat, but a penetration is a pentration, which the PzIV is capable of doing to an ISU-122 FRONTALLY, not to mention from the side.

Hopefully people will sit back in their chairs and rethink this particular matchup again. The ISU-122 is not as über as some apparently think it is.

zukkov,

i believe it needs some tweaking. not only in this case where the isu definitely should not back off against a mkIV,

Looking at the statistics, I don't know how you could say that with this example after looking at the data. The PzIV can kill it just as well at that range as a King Tiger.

but in other situations as well. i'm playing a pbem vs a friend and he just "charged" my lone kv1 with 6-8 mkIIIs. as soon as they crested the hill and spotted it, every one of them reversed back to the other side. this is just all to common i'm afraid and it just makes it impossible to overwhelm a small number of uber tanks with many..uh..not so uber tanks.

If this were the case all the time, then I would agree with you. But it clearly is not. I just played a game where I had a platoon of T-26s facing off against several nasties, including StuGs and PzIVEs. I lost 2 right away DURING a tank duel in which one of the T-26s got a lucky flank shot and killed one of the PzIVs. I subsequently lost the other 3 in subsequent engagements, but they did engage.

The problem with a thread like this (no offense to anybody) is that one person says:

"I saw something that I didn't like. Fix it."

Then other people say the same thing. Usually saying:

"Yeah, what he said. This happens *all* the time. Makes it impossible to do anything".

While this MIGHT be true to some extent, it is always warpped and distorted. This ISU example is not an über matchup, at least not at that range. In other example the "wimp" went up against THREE threats and defeated two, but had trouble getting the third one killed. In my example a REAL wimp knocked out the best tank in the German Army of the day, while others attempted to do the same (and died in the process).

As with anything AI related, it can do some dumb or inexplicable things. But for every one of those, there are probably dozens or even tens of dozens of examples where it did exactly what should have been done. What people sometimes forget is that we can not cater the system to "outliers" without risking screwing up the entire system. Then instead of people complaining about the every once and a while thing they don't like (many of which are not in fact bad, just perception of bad) there will be many more people complaining about far more frequent bad AI decisions.

In this case, the risk is making tanks too brave. So when that PzIV gets the drop on your ISU-122, the ISU-122 doesn't pull back and instead gets one right through the snout. Then the shouts demanding that we change the system yet again come up, probably in larger numbers too.

My point is that it is impossible to make the TacAI much smarter than it is. Tweaking won't likely do anything but make things worse. Overall. Don't forget that we have been testing this thing for almost a year and a half. It isn't like we haven't already tweaked the system as best we can.

Now... if there is some sort of bug... that's a different story. But thus far I haven't seen any reason to suspect that. I will download the files now and double check that.

Steve

[Battlefront.com]

mPisi,

While I have not seen the retreating behavior in question, I have seen a tank firing its main gun without a red targeting line to its target as Redwolf says above. I can probably find the savegame if needed. I wondered if it was a FoW enhancement, but it does not seem consistent.

No... I think this might actually be a bug, but one that is not new. In the dusty reaches of my brain I seem to recall a discussion about this with Charles a long time ago. IIRC it is possible for the line to not show up, or remain in place even out of LOS, under certain circumstances. But in neither case is the game itself affected. Just something to do with interface updates or somefink

Steve

[Battlefront.com]

Results of first test using Redwolf's scenario...

ISU-122 vs. PzIV

Both spot each other right away, both target each other, neither pull back. Both fire at same time. PzIV is killed, ISU-122 suffers a partial upper front hull penetration and one crew casualty.

Er... so much for the PzIV not being a threat, eh?

Seriously, this is just ONE test and therefore statistically means nothing. However, it is interesting to note that I saw NO retreating, a PzIV kill and a semi-lethal hit on the ISU-122.

Funny enough, I didn't move some of the other stuff he had in there and to my surprise found three SU-76s going up against two PzIVs at around 760m. After one turn nobody pulled back, one SU-76 was killed, no other hits scored.

I'll run it again and see what happens.

Steve

[Battlefront.com]

Two more tests:

Test One (ISU-122 v PzIV)

ISU rotates to face PzIV, PzIV rotates turret to shoot (this happened in other test, forgot to mention). ISU fired two shots, missed. PzIV fires two shots, missed. Neither backed away. PzIV fires 3rd shot and gets partial frontal penetration. Shocks crew, but no casualties this time. ISU managed to crack off 3rd shot before this and brewed up the PzIV quite nicely.

Test Two (ISU-122 v PzIV)

ISU rotates to face PzIV, PzIV rotates turret to shoot. ISU gets off first shot, kills PzIV quite dead. End of test

Test One (3xSU-76s v 2xPzIVs)

Skipping details... nobody reversed, both PzIVs dead.

Test Two (3xSU-76s v 2xPzIVs)

Oooo... this one was funny!! One SU-76 targeted by two PzIVs decided to move forward and to the left to get out of LOS. This works. While moving the other two SU-76s kill one PzIV and get a non lethal hit to the turret of the other one. That PzIV kills one SU-76. BUT WHAT IS THIS?? Our hero from ISU Test Two is not satisfied with one PzIV kill, so he rotates into action and fires at the PzIV. The PzIV then turns to face the threat. On his second shot the ISU decides to pull back, but not quick enough! The PzIV's third shell makes a partial penetration and takes out one crew member, making the vehicle Shocked. This stops the ISU in its tracks. The PzIV scores ANOTHER partial penetration and puts the vehicle into Panic mode on top of Shocked. Turn ended with the PzIV still being fired at by the SU-76s and not moving.

OK folks... see why I want files to play around with? Now, who can tell me what the TacAI did wrong in these four examples? Who here can tell me what it did right? I would like Redwolf to lead the discussion

Steve

[ November 28, 2002, 12:14 AM: Message edited by: Battlefront.com ]

[Battlefront.com]

Oh man, this is just too fun! I did another one...

ISU-122 again scores first shot kill on his prey. One of the other PzIVs dies quickly and again one SU-76 (same one, only one PzIV shooting at it) pulls ahead out of LOS after exchanging two shots. Other two SU-76s kill PzIV (4 crew casualties!) before it does squat, then hit it twice more before they realize it is KO'd. The surviving PzIV starts to exchange fire with the SUs. Then the ISU fires a shot and misses, so thee PzIV rotates turret and... wait for it... kills the ISU quite dead on the first shot. It then takes two hits from the SUs (hull front, ricochet and gun hit no damage) and fires back a couple of times.

OK folks... explain again how the ISU has nothing to fear from a PzIV, that thin skinned vehicles won't stand up to the enemy, and why it is we should change the code to work differently?

Steve

[Battlefront.com]

HAHA!!! Couldn't resist yet another one! This is the best one yet. OK, I'll skip to the punch line...

After 60 second

Germans - 3xPanzer Mk 4 Auf H "ünter" tanks

German end status - all alive

Soviets - 1xISU 122 "über" SPG, 3xSU-76 SPGs

Soviet end status -all three vehicles KO'd

One SU-76 managed to score a partial penetration, but no damage. No other hits than that. Oh, and nobody scooted around.

I could go on with this for hours, but I have other things to do.

Steve

[Silvio Manuel]

Originally posted by zukkov:

nice job rw. this whole self preservation thing i think has been taken to the extreme in cmbb. i believe it needs some tweaking. not only in this case where the isu definitely should not back off against a mkIV, but in other situations as well. i'm playing a pbem vs a friend and he just "charged" my lone kv1 with 6-8 mkIIIs. as soon as they crested the hill and spotted it, every one of them reversed back to the other side. this is just all to common i'm afraid and it just makes it impossible to overwhelm a small number of uber tanks with many..uh..not so uber tanks.

to conclude, i would like to join those that think bfc should relook at their auto-retreat program and tweak it a bit...

Sounds like a job for your offmap spotters... bring on the smoke and let your wimpy MkIII's get very close so that they can get the kill b4 balking at the sight of the mighty Chewbacc, I mean, KV-1.

[Schrullenhaft]

I think everyone should take note that Steve is actually playing the game and having fun.

[Michael Emrys]

Originally posted by redwolf:

Second 29: MK IV fires. It still has no red line to the ISU, although it clearly shoots the main gun at it. The ISU begins reversing the same moment the MK IV shot gets off.

One thing is fishy here, and that is that the MK IV does not get a red line to the ISU although it is definitivly shooting at it. Is this intended, does extreme fog of war allow for not showing enemy target lines?

Something I've spotted a couple of times in a recent game is that if a unit is being targetted—shows a yellow line—and then targets one of the units targetting it, it won't always show a red line. Similarly, the other unit won't always show a yellow line. My guess was that the draw code in the engine gives precedence to one unit and its targetting lines and won't draw another on top of it if they would exactly coincide. I don't know if that is what is happening here, but this situation jogged my recollection.

Michael

[Dschugaschwili]

Originally posted by Battlefront.com:

[...]

Germans - 3xPanzer Mk 4 Auf H "ünter" tanks

[...]

Steve

Actually, it's "unter", not "ünter". Don't go overboard with the "ü"s.

Dschugaschwili

[DWG]

Looking at the screenshots that redwolf posted, it looks like the Pz IV is in a better position, in that he is sited in or around trees whilst the SU-122 is cresting a higher hill and is therefore silhoetted against the sky, making for an easier shot for the Pz IV. I don't know whether the AI factors this in, making vehicles realize that they are clearly visible against the skyline and therefore pulls them back. Just an observation.

[Kanonier Reichmann]

I think you're being a bit harsh there with my statement Steve. You said:

BUT... now onto the more interesting aspect I uncovered....

KR,

--------------------------------------------------------------------------------

I think the telling thing here is that the ISU122 has identified the enemy as a PzIV and not a Tiger so fear of being overmatched should not apply.

--------------------------------------------------------------------------------

What amazes me is that nobody apparently took the time to look at the armor and armor penetration stats. If people had, they would know that Jim's statement above is incorrect

The PzIV's gun can kill a ISU-122 from the 500m +/- range Redwolf tested at. Perhaps not by a margin of overkill, but it is statistically very possible even without factoring in the 90% armor Quality rating. Factoring in the 90% Quatity rating for a ISU-122S (not used in this example) the PzIV also can kill it frontally. It can also BARELY kill the IS-2 family of vehicles. The Panther, on the other hand, can kill any of these at 500m no problem.

What does this mean? That for some reason people have been going on the assumption that the PzIV was not really a threat. Looking at the stats it *clearly* is. Perhaps not an overkill threat, but a penetration is a pentration, which the PzIV is capable of doing to an ISU-122 FRONTALLY, not to mention from the side.

When I was talking about being overmatched I wasn't trying to say the Pz IV couldn't kill the ISU122 but I sure as hell know which AFV I would prefer to be in in the circumstances presented by Redwolf's test. The screenshots presented by Redwolf clearly shows the ISU122 has the "jump" on the Pz IV yet it's the Soviet AFV that chickens out first rather than the "caught out" Pz IV. I personally think this is a bit strange myself especially considering the odds that the Pz IV was up against, trying to slug it out with an enemy AFV it can marginally penetrate while the IS122 should pretty much blow the Pz IV sky high if it hits. Shouldn't it have been the Pz IV that backed up in this example?

One other thing. I have a sneaking suspiscion that Redwolfs example may well have seen the ISU122 reverse even if it had caught out a Stug III F/G where the Pz IV was. Based on my experience with the IS 2's, reversing behaviour seemed to occur even if you catch an enemy Assault Gun with its proverbial pants down with a flank shot. It seems the computer code acknowledges that the gun on the Stug III can kill it even if it will take a month of Sundays for the enemy Assault Gun to be able to traverse first to get a shot off. If you can confirm this is not the case & the game engine allows for catching out enemy "fixed gun" AFV's then I'll accept that.

Unfortunately I have no proof so I acknowledge that it's hardly going to convince anyone that the code should be changed but I'd love to hear from anyone that may have experienced similar behaviour, especially when ambushing enemy Assault Guns.

Regards

Jim R.

[ November 28, 2002, 05:39 AM: Message edited by: Kanonier Reichmann ]

[Monty]

The only time i saw my Russian tanks retreat was when they faced ID't Tigers ( 3 ), so i could understand that.

I see regularly infantery support tanks or SP tanks retreat but i would be surprised if they did not.

Otherwise i see very brave tank crews

btw in RW's pictures i see the Russian crew was regular.

I think experience is also important in crew behavior.....

Monty

[ November 28, 2002, 06:57 AM: Message edited by: Monty ]

[Redwolf]

Originally posted by Battlefront.com:

What amazes me is that nobody apparently took the time to look at the armor and armor penetration stats. If people had, they would know that Jim's statement above is incorrect

Steve, dammit, I can't believe this. Search for the term "excellent" in any of my posting in the other thread to see what kind of armor kill stats I gave. I never claimed that the Pz IV is no threat. I was always sure to point out that the ISU is more dangerous to the Mk IV than vice versa.

It is totally obvious that the ISU should not retreat in this situation. In my first movie it didn't get a shot off at all, in the second it shot while retreating, getting lower hitprobablity from being on the move.

[Bruno Weiss]

I don't unnerstand why some folks are experiencing this problem, as I haven't as yet. Only on one occasion, and it was because of what I'm going to point out.

The ISU is clearly out in the open. I can't see from the camera angle, but it does not appear to be hull down, just dangling up there on the ridge in the open. Now, the MkIV, is behind trees, at a lower elevation. Meaning, in the world of AI LOS computations, the ISU is potentially at threat from a shot to it's frontal underside, or at least out in the open, where the MkIV is more covered. I'll wager if the ISU were nicely hull down, and barely peeping over that crest, then there wouldn't be a problem. I think the fact that it took the ISU some time before it spotted the MkIV, only proves my point. It does not have as good of an LOS angle (percentage of kill chance, etc), as the MkIV.

The one time I had an experience like this, was when I ran an armored unit out in the open against a 20mm armored car, and it back away. A turn or so later, I edged the same unit through scattered trees, gave it a good hull down position, and it never flinched and took out the same target with one shot.

I believe the CMBB AI is more concerned than in CMBO, with individual unit survival, and places that calculation as a priority over attack calculations.