Jump to content

Panther Shot Trap Still Not Trapping


Recommended Posts

Sgt Joch - asked and answered - the empirical test data is that it happened one out of eight times in their penetration tests of whether the US 76mm could defeat the Panther front. A hypothesis test p value of getting 1 out of 8 trials as a success in a geometric distribution from a hypothesis of a prior probability of only 2 in 2243 is 10 to the minus 31. - we can reject that hypothesis.

so you are saying the probability of Panzer shot trap occuring is 1 in 8 or 12.5% of every round fired?

again, where is the empirical test data that BFC is wrong. Otherwise you are just guessing.

Link to comment
Share on other sites

so you are saying the probability of Panzer shot trap occuring is 1 in 8 or 12.5% of every round fired?

again, where is the empirical test data that BFC is wrong. Otherwise you are just guessing.

No - he is saying that US experiments with a 76mm gun yielded 1 shot trap penetration in 8 shots. And that therefore the actual, real probability should be at least margunally consistent with that (subject to the proviso that the US army tests sufficiently mimic the properties of hits in combat situations).

Link to comment
Share on other sites

Statistical analysis from one test shot would be...presumptuous.

Now, I'm not saying it DIDN'T happen. It just didn't happen a lot. That's vague, isn't it?

The top picture, with the circled hits from the test, do not show the penetration. The lower picture has the turret slewed further to the right. It -seems- to be a different vehicle than the other picture.

FWIW, the hit which the mantlet deflected has gone through the junction of the roof and side armor. I'd say that's harder to penetrate than just the roof. FWIW. However, that picture does not state the range and weapon used. The hole is on the order of 3". That encompasses almost every allied tank and AA weapon which could penetrate the Panther.

If the shell has a high length to diameter ratio, it would be very prone to shattering rather than ricocheting.

Shell type, weapon, range, all matter. (As we all know.) A picture or two (well found!) does not give any kind of correlation.

If I flip a coin once, I cannot say that it will always land on heads. (Simplistic example with many flaws. Go with it.) One penetration in 8 may have been from a test specifically staged to maximize the shot trap weakness. Shatter gap, for example, is a non-intuitive characteristic of certain shells at certain speeds. Similarly, this may be less straightforward than assumed at first blush. I would not try to extrapolate any trends from one data point.

(No Jentz at hand atm, but I thought the 40mm mod to the roof thickness also corresponded with a turret roof increase, as well. A response to greater weight of allied artillery. Just from memory, so I may well be wrong on that.)

(And the mirthful part of me notices that the picture with the deflection and penetration clearly shows zimmerit present. Coincidence? I think not... ;) )

Ken

Link to comment
Share on other sites

Yeah, 8 shots is WAY too small of a sample. We also have to take into consideration these were non-combt aimed shots from one specific type of gun with one specific type of ammo. That no doubt has an effect on the outcome, though what that effect is I couldn't say. Just that the circumstances are totally different and therefore it's unlikely the outcomes would be totally the same.

I'm going to have to side with physics over a single US Army testrange test result.

Steve

Link to comment
Share on other sites

No - he is saying that US experiments with a 76mm gun yielded 1 shot trap penetration in 8 shots. And that therefore the actual, real probability should be at least margunally consistent with that (subject to the proviso that the US army tests sufficiently mimic the properties of hits in combat situations).

I was being sarcastic, but that seems to be a lost art around here.

a 8 shot sample is useless, the only thing it shows is that the probability is greater than zero.

In the Isigny Test, out of 80+ hits only one ricocheted into the hull.

http://wargaming.info/1998/us-army-1944-firing-test-no3/

so what does that mean, that the probability is less than 1.5%?

so probability greater than zero, but less than 1.5%. Is'nt that what we have now?

Aren't statistics wonderful? ;)

Link to comment
Share on other sites

I recall the first time I did shot trap testing I got a roof penetration (and catastrophic explosion) the first or second shot. If I had ended the test after eight shots I would have presumed a 1-in-8 probability. But the more I tested the lower the probability went. Remember, if you've got a 1-in-100 probability that doesn't mean it'll take 100 tries before you see the result, you might might get luck the first time. Or you might fail 300 times then get three in a row! Or it might never happen at all. Statistical probability is not certainty.

Link to comment
Share on other sites

so what does that mean, that the probability is less than 1.5%?

so probability greater than zero, but less than 1.5%. Is'nt that what we have now?

Aren't statistics wonderful? ;)

Statistics are wonderful. And they tell you that the lower limit isn't zero. A 1:1000 chance would (for example) give an extremely low probability of getting one penetration in eight attempts, and so can be ruled out with high confidence. (Likewise a 99% chance of getting a penetration can also be ruled out but that is less interesting). When you are looking at low probability events the difference between 0 and 0.001 is actually very large.

So the real world test does say something potentially useful. It rules out a large range of probability space (viewed in logarithmic terms). Possibly not a range that was ever under serious consideration, but if game tests are showing 0.001 probabilities ( when appropriate tests are done) then that would matter.

Link to comment
Share on other sites

(No Jentz at hand atm, but I thought the 40mm mod to the roof thickness also corresponded with a turret roof increase, as well. A response to greater weight of allied artillery. Just from memory, so I may well be wrong on that.)

You're thinking of the schmalturm, which did have thicker roof armor. It had thicker armor everywhere. But the turret roof remained 16mm thick on all production Panthers.

Link to comment
Share on other sites

I wonder: is this shot trap thing a "hack" specifically coded for the Panther, or has anybody ever seen a shot being deflected onto itself by any other tank.

Not exactly, but I have at least one instance of a "ricochet into opening" penetration on a Tiger I. It was on a test Tiger that was hit after the test crew lost their nerve and bailed.

Link to comment
Share on other sites

Do you have empirical test data? Otherwise you are just guessing.

Of course I am guessing. Just like Steve was guessing. It's also irrelevant since that is not the question I'm trying to answer.

In the Isigny Test, out of 80+ hits only one ricocheted into the hull.

http://wargaming.info/1998/us-army-1944-firing-test-no3/

so what does that mean, that the probability is less than 1.5%?

so probability greater than zero, but less than 1.5%. Is'nt that what we have now?

Well no, it's not what we have now. Were all 80+ hits at Isigny on the mantlets? Because I'm only counting hits on the mantlet. The percentage of total hits on the tank is some fraction of that. Apples to oranges.

Link to comment
Share on other sites

Of course I am guessing. Just like Steve was guessing. It's also irrelevant since that is not the question I'm trying to answer.

of course it's relevant old boy. There is no hard data on this as on so much else, so everyone is making an educated guess.

Your question #1: is the Panther Shot Trap modeled?

Answer from BFC: yes, it is.

Your question #2: is the Panther Shot Trap properly modeled?

Answer from BFC: yes, it is.

Your counter-argument to Q#2 is that your "educated guess" is more accurate than BFC's "educated guess", but you have no data to support your "guess".

so, back to square one. ;)

Link to comment
Share on other sites

Your counter-argument to Q#2 is that your "educated guess" is more accurate than BFC's "educated guess", but you have no data to support your "guess".

so, back to square one. ;)

I don't think that BFC is making an educated guess: Steve was pretty straightforward about this. He said, literally, that Charles was just letting the "physics" talk. I bothered enough to run the numbers, and calculate the probability of not getting to see one of those lower mantlet hits that enable the shot trap to happen. Even on JasonC scenario where one gets 5 to 10 hits on the Panther frontal aspect, the chances of not hitting the lower mantlet were as high as a 76%. It was also easy to see that in order to be guaranteed to see such a hit, one would need to have about 200 rounds on the Panther frontal armour.

Because in order to the shot trap happening, you need 1) to hit the lower mantlet armour and 2) have the projectile ricochet in such a direction and kinetic energy that penetrates the Panther armour (which is what Ken is trying to say and seems to be ignored).

Getting a simple probabilistic model of 1) is easy, we've done that in this thread. But 2) is matter of physics, which is probably modeled by a standard Physics equation that uses as parameters the incoming angle, incoming projectile velocity and the characteristics of the plate being hit.

Getting a probabilistic model of 2) is not doable without making an 'educated guess' at what physics model Battlefront is using. But we can make a guess: that the probability of 2) being resolved in such a way as to cause a penetration are less than 1, which means that the shot trap is much more unlikely than what the probabilistic model of 1) alone suggests.

My educated guess is that Vanir will eventually get to see it if he tries enough times.

Link to comment
Share on other sites

Your counter-argument to Q#2 is that your "educated guess" is more accurate than BFC's "educated guess", but you have no data to support your "guess".

so, back to square one. ;)

You might want to actually read the thread before posting. I did some tests. The results are posted. These tests were not intended to answer the question Steve and I were discussing. That is a tangentially related discussion.

Link to comment
Share on other sites

Because in order to the shot trap happening, you need 1) to hit the lower mantlet armour and 2) have the projectile ricochet in such a direction and kinetic energy that penetrates the Panther armour (which is what Ken is trying to say and seems to be ignored).

Not ignored. I've addressed it in multiple posts.

Link to comment
Share on other sites

Since people apparently do not understand limited size samples and their proper statistical use, I will explain precisely what my reasoning was above. To start with, my actual working hypothesis is that the chance of a hit on the shot trap areas of the Panther front, given a shot from dead ahead, is around 2.5%. I believe the chance of deflection from that point of impact that goes into and through the roof is a substantial portion of that chance, again on the order of 2% or so in the case of a US 76mm APC round. I think the data available is consistent with a chance of that event as low as 1% but that it is quite unlikely the true chance in that low. I think it is most likely that the true chance is 2 to 2.5%.

Why do I think those things? First, modeling - area analysis. There is no magical way for the shot trap area to repel incoming shots that travel the length of the front of the tank in a single millisecond. There will be center weighting of the probability of a shot hitting in a certain location, due to center of mass aiming and the fact that shot dispersion around the point of aim will be bell shaped probabilities, highest nearest the center of mass. The shot trap area is in that highest chance region, not outside of it - it is not a peripheral location.

What does the area reasoning tell me, then? It tells me that a reasonable *lower bound* for the chance of a given shot hitting the shot trap region is around 2.5%. The chance could be appreciably higher - it depends on how sharp the center weighting is. Always, with the proviso that we are talking about hits.

Next I look at the single report of 1 such penetration in 8 trials. I ask the following statistical question - can I reject a null hypothesis that this 1 in 8 occurred due to 8 random trials each with a 2.5% chance of success? The answer is that I cannot, because the p value of that combination - a geometric distribution with 0.975 probability and 8 trials yielding 1 positive outcome - is 0.07, which is too high to reject that hypothesis. That sample, small as it is, is perfectly compatible with the hypothesis I derived from *modeling*, from prior casual understanding of the event.

The sample is certainly too small to *establish* the true probability empirically, but it can be used to test hypotheses and assess how likely the observation would be on a posited prior probability of the event.

But I am then asked to consider the alternate possibility, that the game has the correct prior probability of the event. The game shows a prior probability of the event of 2 in 2243, which is a very large sample size, enough to know that the game's prior probability is nothing like 2.5%, but is instead more like 0.09% or 1 in over 1000.

I can then ask the same question of the 1 in 8 sample. Could it have arisen due to pure chance, in 8 random draws each with a 1 in 1000 chance of success? Metaphysically and purely from the sample, of course it could. But I can reject that hypothesis as having a p value too low to be believed to be due to random chance.

But I am not done. I started with casual reasoning, and the alternate hypothesis has to go through the same gauntlet. What magical process is supposedly preventing an exposed area of 2.5% of the surface of the tank from being hit 2.5% of the time? Or if it is being hit 2.5% of the time, where are we to suppose that the factor of more than 40 that stands between that hit location chance and the event chance, comes from? We are to independently suppose that our empirical result was a 1 in 40 outlier, not from even two hits to that location only one of which went in, but from exactly one such hit which did go in.

You say I have no evidence? You have much less evidence. I have a straighforward model that predicts the hit chance, is statistically compatible with the thin available report on both the overall event probability and the penetration if hit on the shot trap. You have neither, and instead a magical extra factor of 40 you pulled out of your backside, and an overall event p value way too low to fit the hypothesis that the only sample we both have arose from random chance.

I am being guided by what is actually known about the problem in both of the above steps. You are being guided by a desire to believe a very low event probability that has no causal explanation and no empirical evidence in its favor, whatever.

Link to comment
Share on other sites

I won't pretend to be able to contribute anything regarding what the hit probability should be. I do have one question. The answer may be somewhere in this thread, but I didn't notice it and considering the thread starts with links to other threads...too much to read. How is the angle of fire being determined? As far as I can see it simply states with two tanks on level ground. What is the relative height difference of the barrel of the tank firing and the expected location to cause the shot trap? Is that angle conducive to testing to produce the expected result? (in other words is it a level shot or is the angle such that it is affecting the test results?)

Just a thought, perhaps already accounted for but I figured worth asking at least.

Link to comment
Share on other sites

What is the relative height difference of the barrel of the tank firing and the expected location to cause the shot trap? Is that angle conducive to testing to produce the expected result? (in other words is it a level shot or is the angle such that it is affecting the test results?)

I had wondered the same thing but all I could find is that the Panther is about 10" higher than the Sherman overall - I couldn't find heights to the barrel. I guess you could put them side by side in-game and check the heights. I also assume over the distances being tested that the shell is following a flat trajectory and not arcing?

Link to comment
Share on other sites

Of course the shells arc in flight. But it is a very shallow arc.

What is the relative height difference of the barrel of the tank firing and the expected location to cause the shot trap? Is that angle conducive to testing to produce the expected result? (in other words is it a level shot or is the angle such that it is affecting the test results?).

The tanks are all at the same elevation.

As for gun barrel height, I'm already in the midst of testing that will hopefully give some food for thought along those lines. I have replaced the Shermans with Cromwell VIIs, which are significantly shorter. I have also moved the ill-fated Panthers into within 100 meters of the Cromwells. I have a suspicion -- as of yet unconfirmed -- that barrel height is playing a factor but not with regards to the shot trap.

Link to comment
Share on other sites

VaB, I know you've done a LOT of drudge work. How about replacing the Panther in your test with a version with the chin? That would be interesting.

I went to do your suggested test only to be reminded that there are no Panther models in CMBN with the mantlet chin :o

Link to comment
Share on other sites

Next I look at the single report of 1 such penetration in 8 trials. I ask the following statistical question - can I reject a null hypothesis that this 1 in 8 occurred due to 8 random trials each with a 2.5% chance of success? The answer is that I cannot, because the p value of that combination - a geometric distribution with 0.975 probability and 8 trials yielding 1 positive outcome - is 0.07, which is too high to reject that hypothesis. That sample, small as it is, is perfectly compatible with the hypothesis I derived from *modeling*, from prior casual understanding of the event.

And that's a fairly subjective assessment, Jason. If the null hypothesis is that "the probability of the shot trap happening is less than 0.025" I wouldn't reject it with a p-value over 0.01 (that's 99% certainty). Of course, the grounds on which p-value you select as the one rejecting or accepting the null hypothesis, depend on how certain you're of your claim.

Given that we don't know much about the second step in the "panther shot trap" stochastic process, I'd be very very conservative and go for the 0.01 rather than the 0.07 you're happy with :)

PS: It's nice to find someone to discuss statistics with, yet I'm surprised this happened in a forum about a WW2 war game.

Link to comment
Share on other sites

adgy.png

It is interesting that this ricochet would, in fact, be blocked by the mantlet "chin". The statement about hitting below the transverse centerline notwithstanding, I wonder if the shot trap only works if the shell hits the bottom 4" or so of the mantlet.

Edit: it also took out a chunk of the turret on its way through the hull roof.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Unfortunately, your content contains terms that we do not allow. Please edit your content to remove the highlighted words below.
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...