State of Heavies, take two

I'll give my imput as well after using the pershing, is2 and Tiger.

The conclusion I have with them is that in game they are mostly one trick ponies. What I mean by that they really shine with vet and loosing one especially with vet, you cannot afford to waist time and resources into investing into another one.

That's the only problem I have.

Snip

You mean they are all-ins, not 1 tricks. They have many tricks, just dont lose it ever.

You mean they are all-ins, not 1 tricks. They have many tricks, just dont lose it ever.

I think some, like the IS-2 and the King Tiger, should vet up faster

Fully agree to the last part. In fact, if i were up to me I'd probably go for a slight price reduction (maybe around 10%), but that's it. Performance-wise there's no doubt that heavies are decent; they provide good AT paired with excellent AI (with the possible exception of fighting units in heavy cover) and great survivability. A buff to mobility would probably require quite a bit of a rework, as this would eliminate one of their key weaknesses (in particular the slow turret/hull rotation). At the very least the range boost the Tiger and IS-2 get at Vet 2 would have to go in this case. However, I personally see the extra range as kind of a defining trait and would rather not have it replaced by something else.

Not talking about buffing all heavies mobility to medium tank level, just a tad bit of acceleration buff. Maybe by 0.2 max. For example tiger 1 is almost always faces a TD which makes its life miserable. Slight mobility buff means it is more responsive when caught out with AT/TDs.

Also I think slight popcap buff is also welcome. All heavies are very pop cap heavy + they almost always require several squads of engies to keep them in the fight. So slight popcap adjustments (-1 or -2 pop cap) seem like a nice buff. Currently if you have good micro it is better to go for 2 medium tanks then 1 heavy tank. Price and pop cap is near identical, but firepower of 2 mediums is substantially higher.

thanks a lot!

do you mean something like the probability distribution of how often the 1st shot dealt x (0, 20, 40, etc.) damage or scored n kills?
unfortunately i don't have this kind of data at the moment due to how the results are calculated (i.e. it stores only the average, not the individual results of each iteration used to derive it which reduces calculation time by a lot), but it would certainly be possible to do this for the first shot.

i agree that this would be quite interesting, especially with respect to judging how reliable the damage output is between different tanks. maybe i can get some data on that...

Yes, that's exactly what I meant.
Especially with tanks that use a long reload like the IS2, this would be very interesting. Units fall out of meta because they are simply unreliably when you need to count on them. Especially if a heavy does not perform by pure chance in the first engagements a lot of the shock value is gone. Making them more reliable was also one reason for the rework. Looking at the AoE changes this has been achieved 100%, but being able to deal not enough damage within the first 10-15 seconds of the engagement can still be a breaking point.

Also out of interest: how do you allocate MG damage? Start at model one and switch to the next one ones it dies?

Also out of interest: how do you allocate MG damage? Start at model one and switch to the next one ones it dies?

yeah, pretty much. all mgs use the same target as the main gun and switch to the next in line as soon as the targeted model dies. this may not be 100% accurate but helps to keep things simple. damage is properly randomized though as a function of accuracy and number of bullets fired, also taking "over damage" and ready aim time upon target switching into account.

Yes, that's exactly what I meant.
Especially with tanks that use a long reload like the IS2, this would be very interesting. Units fall out of meta because they are simply unreliably when you need to count on them. Especially if a heavy does not perform by pure chance in the first engagements a lot of the shock value is gone. Making them more reliable was also one reason for the rework. Looking at the AoE changes this has been achieved 100%, but being able to deal not enough damage within the first 10-15 seconds of the engagement can still be a breaking point.

This is a very good point and something that isn't really well portrayed by the averages in the graphs, as you've mentioned. Actually, I'd even take it a step further and say we'd need not only the probability distribution for the 1st shot, but for the whole duration of the firefight to make an educated guess about how reliable the damage output/kill rate is.

I've actually had a bit of time to spare on collecting some more data yesterday, and the results are quite interesting. However, since this expands the problem into a 3D data set it's a bit hard to visualize properly, but for the three heavies in question it looks something like this:

Show Spoiler

In these graphs, the x- and y-axes represent the number of shots fired and damage dealt (divided into 40 damage intervals) or kills scored, respectively, while the height of the surface/columns shows the probability.

Taking the Tiger for example, it's quite obvious that there is a rather huge variability in damage dealt for the 1st shot (anywhere between 40 and 300, with 120 -200 being the most likely), but dealing no damage at all has only a ~15.4% chance. This quickly shifts towards higher average damage and a narrower probability distribution the more shots are fired, and it is actually quite unlikely (less than 10% chance) to deal less than 200 damage after the 2nd shot.

With respect to model kills basically the same applies, although here the chance not to drop one or more models with the first hit is a staggering 76%. This goes down naturally as time progresses and after the 2nd shot (or ~ 11 s) there’s only a 26% probability to not having killed a single entity.

The differences between the heavies are actually relatively small. The IS-2 and Pershing are a bit less likely to completely miss (as in deal no damage whatsoever) with the 1st shot and the most probable results shift a bit to higher average damage numbers. Also interesting, the Pershing has a small chance (5%) to kill 3 models with its alpha strike, whereas the Tiger and IS-2 can’t kill more than 2.

As said before it’s a bit hard to read these graphs as you can’t rotate them to get a different perspective, so for those interested I’ve uploaded the source spreadsheet with all graphs and additional data for the preWBP heavies plus the Pz.IV, T-34/76 and HE-Sherman
here for download.

Show Spoiler

This is great!
How long do your simulations take to run? I assume the program will be quite busy for running through all these iterations.

I think the most important shots are shot 2-3. The first one you can't prevent, the second one only if you retreat instantly. Usually I would decide based on the first 2 shots if my squad needs to go home or if it is worth taking the risk for a snare or staying behind cover.
I assume the data will be slightly favouring the Allied tanks due to using 6 men, which allows them to do more damage than they could in game. As a benchmark it is very nice. May I ask how far the models are spaced apart in your simulation? I assume something between 0.5-0.6 meters? That's the only way to explain how the Pershing can get 3 model kills in one shot. We can also see the effect of the formation on the data when looking at the model kills: Killing 2 models is much more likely than killing only one, because the only way to kill one single model is at both ends of the line formation.

Based on your data I'd say that all tanks are very consistent in damage, but at least the Pershing seems to be inconsistent in model kills ranging anywhere between 0 to 6 for the first two shots. The probability of killing fewer models will also go up for smaller squad sizes. But given the otherwise high and consistent damage might not be an issue though.

[...]

yeah it's kind of slow, unfortunately. 2k iterations clock in at about 10 min on my (rather crappy) laptop, but unless you need to use excel in the meantime it can just run in the background.

as to your analysis, i'd say that is spot on. the squad size plays of course an important role for the amount of damage that can be dealt - the more models available the more absolute damage can be dished out, which can skew the results a bit. one way to circumvent this artificial overestimation, especially when comparing the performance against squads with different model count is to rely on the relative damage (i.e. percent of total squad hp). this would probably interesting if we'd want to compare the performance of, say, the Tiger and IS-2 against a typical mainline of their respective opposing factions (Tiger vs Cons; IS-2 vs Grens)

with respect to the squad spacing, the models in the sim setup are exactly 1 m apart from each other. however, each entity has also a circular hitbox with a radius of 0.5 m around it so, in theory, the Tiger should be able to OHK 3 models at once if it directly hits one in between two others. this is next to impossible via scatter, but since there's a ~1.5% chance from accuracy rolls for this to happen it should show up with the same probability in the data... but it doesn't.

Show Spoiler

i've recalculated the data for the Tiger once again after fixing the error and now the triple OHK shows up as expected with a 1.5% probability.

as far as reliability is concerned, i'd agree heavies are quite good in terms of damage output. though it has to be said that the damage and kill numbers from the 2nd shot onwards already include the contribution of the MGs. if these are excluded (i.e. at a range greater than 35 m) i'd expect the damage distribution to change quite substantially.

might be worth a look as well...

That's a lot of work and I agree for the most part and I just wanted to have a special thought for the Ace Tiger, an alien between heavies and super heavies, which struggle a bit in this patch.