Hey Guys, I just finished my number crunching on non-rocket artillery: https://www.reddit.com/r/CompanyOfHeroes/comments/g1ikmr/howitzers_and_you_big_guns_and_what_they_do/ Howitzers and You: Big Guns and What They Do
My conclusion came down to Sexton/Priest need a pop decrease (Sexton to 10-12, Priest to 13), Sexton needs its gardening Super-Charged shells to have normal AoE (2.5 Mid AOE, not 2), and ML/20 needs an extra round base barrage, currently Vet 1 just brings it up to par. Also, EITHER something special for the ML/20 another another round/an ability or leIF needs to lose Counter Barrage. Statistically, they are almost exactly the same gun (once ML/20 has 9 shots), just leIF has serious icing on top.
(just realized this post has become super long, sorry for that)
MMX and I have been working independently on theoretical AoE calculations as well.
To be honest some those numbers don't say too much, that is my conclusion from thinking about it and discussing with him for quite some time. MMX's work has the advantage of also simulating scatter and thereby more "in-game like" scenarios. But let's break it down point by point to be more precise:
First table: Overall AoE damage. Represents the AoE "volume".
This is the core issue since all other calculations rely on this. This is also what I tended to calculate when I set up my Excel sheet. But to be honest this does not say much by far. The issues here are:
- what are we shooting at? All of these units are mostly used against infantry. So if we want to compare against infantry, we need to cap all damage at 80, otherwise overkill damage will be factored in which falsifies the data.
- The damage is not normalized and hard to normalize. Damage in the inner core is WAY more valuable than on the edges, since it can insta-kill models or bring them down to a point where small arms can quickly finish them off. However, the outer third of the circle where the damage is abysemal often contributes a whole lot to the overall damage volume.
- Against somewhat decent players, you usually only shoot at 1-2 squads at most. So most of the (often outer circle) damage will not be applied.
- The damage in the inner core is also more like to be actually applied, of course depending on the scatter.
This leads to the next point:
Normalization vs scatter. This is by itself good. One minor issue you make is that you neglected the scatter_offset (haven't checked, but I assume those units also have that value; Please correct me here if I am wrong). From the current scatter model, this value basically leads to a larger scatter area by moving the scatter center point further back. However, scatter_offset might be the same for all units, so the introduced error is the same or at least similar.
While this value is better in terms of it shows which units are better able to really damage the intended area, it still gets fed "bad" data due to the above reasons.
Afterwards, you normalize vs time etc which is all fine apart from the already mentioned base data.
Now don't get me wrong, I don't want to shit talk your work here. I like this kind of approach and I think for generally comparing those units within the group itself it can show some differences. BUT:
We should keep in mind that nobody knows how this data actually translates into in-game performance. Does 1000 more damage volume help you? Well it at least does not hurt. If it's overkill damage than it does not help against infantry, just against buildings. Maybe it does not help at all because the damage comes from small damage values in a large outer area of the AoE circle. But since the cannon tends to hit well the outer areas are rarely applied, etc etc. Reliability is a huge thing. The B4 can kill a heavy tank, but people don't like it because your investment into it might only pay off every Xth game or so, while it single handedly wins you the next one because two shots are on target. Then there is shock value. The air time of the shell, which is probably one of the most important things. Unit density (mostly depending on the game mode played). Now obviously you did not claim that your numbers would factor all this in and your calculations are the end-all for defining the performance on this unit. But as I said, as much as I like and also trust in theoretical calculations, we should keep in mind that they have some shortcoming that do not allow us to really draw conclusions for the game itself if we do not correct for them.
Now the thing I think is probably be the "best" indicator is what MMX is currently doing: Making a simulation against a "real" squad. Calculate how long a unit takes to kill a whole squad. Obviously this still does not simulate moving of the squad etc, but it's at least something. If you want to put real work into this, I'd recommend getting in touch with MMX.