May 20, 2013
Accuracy explained - WG dev on 0.8.6
SS: Pavel Myreev aka "Zlobny" is one of the WoT developers. His nickname might be familiar to the European supertesters - he is leading the supertest program as far as I know. The translation - as always - is at some points redacted for better comprehension in English. Also, I am neither physicist, nor mathematician, so I hope I don't get the English mathematics terms wrong. If I do, feel free to correct me. First, a bit of background though.
There was another post (which the developer is referring to) - that "another post" is basically the translation of what was published in English - you know, that "leaked" post, that appeared on saturday night or something. I recommend to all of you to read it (if you haven't already), so you know what Zlobny is talking about.
Well, here goes:
Currently I am working on the British artillery.
First I want to say that there is no need to write a lot of questions - think first and ask later - I am not a community moderator or producer, I have to work.
1) Standard deviation limit - how it was and why
Limiting the aim spread to 1,3 standard deviations before 8.6 (SS: explained here) had, in my opinion, very bad influence on gameplay in general. The principles are generally known I think. When the management decided to change it, I started to think about possible variants of implementing this.
Unfortunately, the first variants were very raw and presumed an "accuracy nerf" of all the vehicles in the game. I was still inclining to the 3-sigma limit variant and not even meetings with Michale Zhivets (Storm) and Maksim Bladyko (chief server programmer for World of Tanks) convinced me otherwise. But in the morning before a very important meeting on this topic in Minsk, I went for a walk and as it happens, it wasn't in vain. It became clear to me that the 3-sigma limit has several significant weakenesses, apart from the increased spread distribution for all the vehicles in game:
- since the difference between 2-sigma and 3-sigma is only 2,14 percent, there will be mods made immediately depicting the aim circle as 2-sigma. After all - why wait for the fully accurate aim circle from 3-sigma and give the enemy more time to cope with you, when the difference between 3-sigma and 2-sigma is so small?
- the distribution will still focus on the center and the accuracy nerf, which I assumed to implement until that morning, won't suffice. If Storm went to the meeting with that solution, it would result in a fail (and bad karma for me).
So I called Storm in panic and told him that our solution sucks. But - as it often happens - the second I said it, another solution came to me by itself.
When considering shooting in our game, the difference between 1,3-sigma and 2-sigma is not all that great and - connected with a camo factor change it might not be even spotted by our players, because the effective battle distance will increase (more on that further on). The solution to fix it with one more sigma (between 2 and 3) came by itself from the previous meetings with Maksim and Misha (Storm).
But some players (SS: arty players) won't have their effective battle distance increased, so these players had to recieve a spread radius modifier. The value of this modifier was discussed by us and Dmitry Dragunov (aka Marschig), who represented the arty players' interests. After the supertest testing, I still think that 50 percent represent a rather small buff (!) of accuracy for arties. On the nerf of arty I will post further.
SS: at this point there is a table posted in the original post. What it means is percentual chance for a shell to drop inside the aim circle (within its borders, not ON the borders or outside them (which gets "normalized") to drop on the borders):
SS: the meaning of this table is - first column are the parts of the aiming circle, with the last line being its border. Other columns are: Old system (how it is now), 2-sigma and New system (0.8.6+). In other words, right now, you have a 19,36 percent chance of the shell going on the borders of your aim circle, in the new system, you will have only 0,27 percent chance of aiming for the aim circle edges. Another example: today, you have 48,43 percent chance of your shell going to the inner half of the circle, in 0.8.6 it will be 70,41 percent.
2) Vehicle camo changes
Our camo system with multiplying visual camo coefficient and camo net coefficient has one big disadvantage, as all the systems such as this do: increasing size of the vehicle very quickly decreases the effectiveness of such devices. The decision to replace the multipliers with addition (SS: as in by adding some value instead of multiplying with it) was taken already in 2010-2011 by me, Maksim Bladyko and Pasha Vasilev (server structure architect for World of Tanks), but back then there was no visual camo coefficient (5 percent), so we didn't do it. This topic was brought up repeatedly in 2012 by L.Zacharchenko, chief game-designer for World of Tanks. Gradually, a snowball turned into an avalanche, which in the end was the reason for plans for the complete camo system rework.
I alread wrote enough letters, but basically we did the following:
- reworked all the camo factors in the game, based on the reference tanks (T-4, T-34) (SS: here, Zlobny probably means Panzer IV, not sure) and on extreme cases
- reworked the influence of shooting on camouflage, based on how it was before, but under the new system, so the "I just shot" negative modifier will be subtracted in the equation, rather than multiplied as it was now
- reworked the camo factor of camo net and visual camo based on the type of the vehicle (not individual vehicle, as it was before)
- reworked the camo bonus for bushes. This is a big topic, so a short version only: currently, maximum bonus from the bushes can be in 0.8.6 obtained by 4 thick bushes, instead of 2 not-thick bushes (pre 0.8.6). Furthermore, the exploit of firing thru the "non-transparent bushes" (skilled players will know about this and others don't really need to know) will be fixed completely. It is also worth noting that the battle (engaging) distance will increase by 15-20 percent.
A collossal amount of work was done on this, there were many tests and I really like the final version.
3) Penetration change
A very painful topic for me, I did that only very reluctantly. Generally it turned out better than I expected, I won't write much - you all saw the numbers.
4) Arty nerf
We did that together with Dmitry Dragunov, he fought for every split percent because I fear that we didn't nerf it enough, but I am generally satisfied by the numbers we sent to the supertest.
I wanted to make arty into a more complicated class with less possibility to fight on very short distances without diminishing the influence of the arty on combat there, where it is needed. New vehicles fit in very well, but I had some issues with SU-122A and M44.
XP buff wasn't done by me, thus I won't comment on that.
SS: Well, that's it. As always, a summary of info from the comments, given by Zlobny and others.
- camo skill won't be reworked (they decided it's not needed: "camo skill is more about positioning the tank by the crew better, but even if you position better, a Maus is still a Maus")
- vehicle camo factors won't be disclosed, but from now on, they directly depend on vehicle size and class (can be estimated by sight)
- camo skill will remain useless on big vehicles
- it's possible American vehicles will be rebalanced in 8.7
......nah, cba atm, will go thru the 700 or so comments later...