It is documented. Personnal experiences aren’t.
More data may be necessary, but the “documenting your search” is here, even if perfectible
1 Like
Also agreed, but my point still stands. If anything it makes the report/instance even worse IMO, especially against the tanks with upgraded composites, let alone ERA (T-55AM-1, T-62M-1, T-72A after the armor upgrade, T-64B, and beyond).
Wait I think we got confused here.
I’m not advocating for HEAT to pen the ERA, if the ERA is effective enough (and HEAT has no business at top tier anyway, but that’s an other topic).
I’m just saying that the HEAT spalling pattern is different than APFSDS, the cone is narrower. Therefore a HEAT warhead through the roof can miss a crew member but still hit some ammo :
APFSDS :
1 Like
Nope, I agree with you here entirely.
Finally some progress…
37 deaths, 5 charges yeeted, maybe 1 undetonated round from a hit video will show.
So 42 or 43 shots fired hitting ammo using 3BM46. 12 - 14% survival rate.
Video will be edited into this post when it’s finished.
3 Likes
I haven’t done stochastics seriously in a decade, cause it’s not something I take interest in, so pardon me for forgetting the specific word [jargon] for it.
You never corrected my math on stochastics cause I didn’t do math on percentages, I made a hypothesis on the math WT uses based on known experience.
If it’s true that means WT doesn’t use stochastics, and if not true then WT does use stochastics.
That’s all I said, we’ve always agreed on the math itself.
Apologies for my harsh replies.
It it possible for you to use DM53? The angle of spalling from 3BM series is much greater than DM on russian vehicles. To keep experiment consistent. Thx
No, because my friend only has T-80U, so he had to shoot my BVM.
Video progress:
Alright. I’ll wait for video. If results are different from my test, we can schedule for practice room, 30 shots to side and 30 to ammo for both bvm and 2a6.
What does the code say about single piece ammo like DM53? Same detonation chance? How is the spalling going to be induced?
It’s based on tank, not specific ammo.
So it’s 0.15 on Leopard 2s, and still 0.15 on T-series.
Ah, okay. So how do you plan on spalling the rack?
I did correct you math because there is no way your claimed methodology made sense. You should also know that outside of thought experiments 100% certainty can never be reached. You still claimed it.
If gaijin doesn’t use a probabilistic model, then it would be a deterministuc one. So each ammo rack would detonate exactly after the same number of hits… every time. That’s what it would mean to “not use stochastics” which also is a weird way to phrase it. You don’t use stochastics in that way, you just put in an RNG and stochastics simply apply. You don’t have to do anything. Gaijin doesn’t have to use anything. That’s not how it works.
And that’s my point it’s also the applications of occams razor. Ammo detonation is chance based, so my calculation apply whether gaijin does something or not. You act like gaijin has to do anything for my calculations to be correct, that they need to implement it that way… they don’t. It simply follows from using a chance based system. Your idea needs gaijin to implement special rules. But again occams razor, the less assumptions the better.
And as i stated before if the system isn’t chance based the ammo racks would explode after the same number of shots, every single time (assuming the shots are identical, if the damage is random per fragment again it stops being deterministic and my calculation applies again, of course there are nuances regarding the models being used, but your hypothesis is Impossible at that point).
1 Like
No, I never claimed that 100% can be reached in the math sense.
ALL I claimed was War Thunder may not use stochastics, that’s it.
There’s been a miscommunication between us.
Again you don’t use stochastics. Stochastics applies as soon as there is chance involved. No need to apply anything.
Are you saying gaijin uses no RNG?
Because as soon as there is RNG → boom stochastics applies. Because stochastics is the art of modelling random events.
I don’t know if they use pure RNG or not. Cause I haven’t found the code that handles det chance as a whole, and differing contrary experiences.
If someone finds the code it would be interesting. We know from how the devs have explained ammo detonation that damaged ammo is more likely to implode with repeated damage. This is why the double spall (two hit) is usually how the BVM goes down.
1 Like
I have trouble understanding you. What makes RNG pure? Again as soon as any RNG exists, you need to use stochastics to model the effects of that RNG. As said before gaijin won’t need to implement a thing other than the RNG itself. The calculations are just the modelling of said RNG.
Experiences are anecdotal and not very useful. You do those videos, but even those lack controls since nothing is stopping you tonrepeat your recording until you get the right results. You can proclaim your honesty as much as you want but, for it to be a real test it needs those controls and if you are what you claim you are, you should know that.
But a very simple video is a video where you shoot a random tank with another tank that doesn’t reliably detonate the ammo rack. Shoot from the same position as often as you need until the rack detonates.
Kill the same tank with the same gun a sufficient number of times. For an infinite(!!!) Population the sample size if a yesno question would be just 385 by the way (95% confidence, 5% Margin of error). Just to give you an idea why i said that thunderskill has a decent sample size, it doesn’t need much to have a great sample size. But since what we are looking for is even more simplistic then a yes and no question 50 times would be sufficient. And simply count the number of shots needed for each attempt until the ammo rack detonates. If it isn’t the exact same number of shots for each of the 50 attempts, then there is RNG.
It is also very unlikely that there is no RNG since Gaijin pride themselves on realism and RNG is simply realistic.
As you can see below: same gun, same distance, same rounds… different results.
Sure, conditional probabilty, easy to implement, easy to screw up if you are a layman. This might even be a point where gaijin screwed up, because their knowledge about probability isn’t any good as they have proven.
It also adds another source of inaccuracies since declassified damage assessments afaik usually don’t account for cumulative damage so gaijin has little real world evidence to go on.