Ok, so I did some testing today with an updated version of my simulator, now including a crude approximation of MG DPS. This allows the sample size with 2,000 iterations to be quite a bit bigger that what would be feasible in-game.
So far the numbers in the OP are in the same ballpark of what i got, just the relative order is different.
The short version:
Using the setup in the OP (same formation, no cover, 25 m distance to target, Sherman w/ AP rounds, combined base DPS for Hull and Coax MG of 9.47) and the respective defensive stats for each unit plugged from serealia's site I get the following T2K values:
Vet 0 VSL Grens 0.91 TS 0% DR EHP: 439.6 TTK: 36.2 +/- 5.4 s Vet 0 Riflemen 0.97 TS 0% DR EHP: 412.4 TTK: 33.8 +/- 4.9 s Vet 3 VSL Grens 0.91 TS 20% DR EHP: 527.5 TTK: 41.9 +/- 7.0 s Vet 3 Riflemen 0.63 TS 0% DR EHP: 634.9 TTK: 47.7 +/- 7.9 s
Actual probability distribution in the spoiler
Riflemen seem to die a tiny bit faster than VSL Grens, which they should considering their higher base RA. Vet obviously gives quite a bit more survivability but, in contrast to what I would have expected, the bonus RA of the Vet 3 Rifles seems to outperform the Gren's 20% DR. Quite surprising, as scatter and, hence, tank main gun damage is barely even affected by RA / target size.
And the reason for this leads us to...
The long version:
Looking at both the average damage dealt and models killed per shot over 2,000 repetitions, it's quite apparent that things are pretty even when the first shell lands - which it should, because at this point the damage output relies solely on the main gun and the differences in RA between the squads don't matter much (to illustrate, the chance of scoring a natural hit ranges between 3.3% (0.97 TSz) and 2.2% (0.63 TSz)). However, after that it gets interesting:
Now the average damage received increases inversely proportional to the EHP, suggesting the contribution of the Sherman's MGs becomes quite significant from now on. The extent of this contribution can be seen in the next two graphs, which summarize the percentages of the total damage and total kills, respectively, caused by the MG alone.
Obviously the MGs are responsible for the majority of the damage dealt and kills scored, with only about 20-30% coming from the Sherman's main gun. That as well should not be a big surprise, since the scatter profile of the AP shells isn't great to begin with. More importantly, due to the model spacing in the formation (average 5.9 m, minimum 3.6 m, maximum 10 m) it is literally impossible to catch more than two models in the AoE at the same time.
As a result, the superior RA of Vet 3 Rifles outweighs the 20% DR of Vet 3 VSL Grens by quite a margin, even though it does next to nothing against explosive damage at all.
To come back to the OPs conclusion: On paper I agree it looks like the survivability of VSL Grens is quite inferior to that of Vet 3 Riflemen, but I'd also say that you couldn't possibly have chosen a more favorable (or more deceptive?) example to prove your point:
The chosen squad formation to test this, together with the poor AI of the Sherman's AP shells means the result is dominated by the damage contribution of the MGs and, hence, the impact of RA on overall toughness (although I understand it is difficult to test other, more clumped formations in-game). For a more realistic test, I'd recommend to switch to a tank with better main gun AI, such as the PzIV or HE Sherman, where the added DR should have much greater effect. Also, choosing a more clumped formation, maybe even in light cover, should paint a more accurate picture of what to expect in a real game scenario where MGs are even less effective.
Wow you went tota; data scientist on this. Hmm... what did you use for 2000 itrs btw. This seems cool so was kinda curious.
Relic Entertainment
cblanco ★ 
보드카 중대 
VonManteuffel 
Heartless Jäger 
