Interview with Jurij Rogach (model specialist)

Source: via wot-news


Hello everyone,

Jurij (Yuri) Rogach is the lead modeller for World of Tanks and recently, people from portal managed to make an interview with him. I’ll be summing up the information contained in this interview (just like I did it with Ola before), because translating literally is too slow.

- he comes from a small town in western Belarus, has two children
- he joined Wargaming in July 2009 to start working on World of Tanks as Lead Tech-artist and in 2010 he moved on the position of “3D outsourcing manager” (managing 3rd part modellers, working on tank textures etc)
- the amount of employees of WG is flowing, but it’s roughly 2000 people by now
- WG will be hiring new specialists, but will be moving to another office, as the current offices are too small

Regarding the process of making tank models and maps:

- map and tank teams are separated, but they are linked not only by the fact they are situated in the officess across the hall, but also personally
- a tank starts being developed in a historical department, where the history consultants prepare drawings and references
- the materials then go to to 3D modelling experts, who create textures and models for the vehicles
- during this stage (modelling, texturing), the models are several times rechecked by historical experts and model/texture experts, who write detailed comments on them
- after the models are made, the collision model (armor thicknesses) is developed, plus the interface (icons and such)
- after that, the model is exported into the game engine for testing
- at that point, the first vehicle balance works are initiated, as well as its optimization (LOD), the destroyed tank model is made
- after all that is done, the vehicle is compiled and again exported into the game engine for testing

- maps also begin by collecting various locality references (photos, videos)
- after the objects (hills and such) are collected, the map department creates a simplified map with all the objects and works on it for a very long time, improving its playability and terrain flow
- at the same time, vegetation, textures and other objects for the map are gathered
- when the textures and vegetation references are gathered, they are passed to the map graphics people, who are professional artists
- the process of creating the map visually is long, historians are sometimes being consulted on the realities of the era
- when all that is done, the finished map is sent to testing and tested especially for the playability (play comfort) and to check whether the two teams are balanced
- there were cases when the mapmakers made beautifully looking maps, the work taking months, only to be sent to archives (scrapped) because of the playability issues

- a lots of tank references were taken in Kubinka and on the Stalin Line complex, where military objects are available to see
- the modelling work is apparently done on Autodesk Maya software, along with applications written in Python-e and MEL.
- the export of models from Maya to the game engine works in three steps:
1) first, they export an animated skeleton using Python, that models the suspension and such
2) after that, using MEL, they create lists (directories) with needed changes in game application structures, they align key procedural points and call the BigWorld engine interface plugin
3) after that, the model itself is exported from Maya to Bigworld, with shaders and such being already made using Bigworld instruments
- apart from Bigworld, Wargaming also has an engine called Influx, that was used in previous WG games and is currently somewhat compatible with Bigworld
- the main advantage of Bigworld are the networking technologies, because apparently no other game transfers so much info during a battle, the graphic effects in this case are secondary, but Bigworld is and will be developed further.

There was a lot more and if you are interested, you can check the original link, I didn’t translate everything (some parts are obvious, some not important). Also, I am not a programmer or an IT guy, so I simply don’t understand the terminology the guy used, I hope it makes sense.

16 thoughts on “Interview with Jurij Rogach (model specialist)

  1. >> – the amount of employees of WG is flowing, but it’s roughly 2000 people by now

    Including all PR guys/gals in all offices across the world? Man, it really leaves three drunken programmers to work on the code…

  2. Good read on how tanks and maps are developed.

    But, regarding the process of making maps and testing the playability, I wonder how on earth maps like Dragonridge could make it past this test. Other maps have flaws to, minor or not so minor, but Dragonridge realy is in it’s own category.

    • Dragonridge was really fun map, I liked the chinese jungle-mountain theme but I admit it that they couldn’t balance it and gameplay was stellar with both teams going to one point of map for all out slug fest and carnage while sometimes few tanks went for hill pass but that almost never yield any advantage or disadvantage and that was it. When they added a valley it was actually worse.

      Still, I hope they’re going to fix it at some point and return it to rotation….

    • I had the “chance” to play it only four or five times before they removed it. What was the problem with it? I only recall that you had to be carefull with your tracks while going down the steps.

    • Dragon Ridge was very good map and in terms of WoT maps quite different one. Only players’ whines sent it to the toilet. I hope WG will revive it at some point. Maybe moving the bases to the diagonally opposite corners would make it more balanced and less prone to ‘all-noobs-to-the-valley’ prenomenon.

  3. “no other game transfers so much info during a battle”

    Could this possibly be changed to reduce the chances of packet loss. I’m not a networking engineer, but I do know the game sends information which does not need to be sent (for example: the shell type you were hit by, whom exactly you got hit by). These two pieces of information are never displayed in the game client (possibly for replays?) and are only accessible through mods. Whilst WG are happy with mods in their game, and have made it fairly mod-capable, they have expressed issues with many mods over and over (i.e. recently tank names on the minimap being clutter) so I doubt they’d send this data just for mods.

    Could reducing the size and quantity of the data packets between the game servers and game clients reduce the pressure on them enough to at least relieve the increasing packet-loss problems many users are having with the game (esp. since 8.8)

    • The type of shell is require so your client know which shell impact animation to use. All shell types have different visual and sound effects e.g.( HEAT has the “sizzling sound on impact and APCR has a blue tracer).

      The minimap (I’m guessing) probably works by taking the coordinates of the tanks on the 3D map and produces them on a 2D display. That extra information is probably so the minimap can differentiate two players driving the same tank.

    • Packet loss is packet loss. You either have it or not and it’s nothing a server can do to fix this, except sending even more data (duplications) so the client can sync up when there’s packet loss. Most networks drop packets from 2 major conditions: misconfiguration and traffic load. Misconfigurations can be resolved but traffic can only be solved by increasing the capacity (or lowering the usage). For WG is very easy to see that users have the problem by checking the network graphs but my bet is on local ISPs with congested pipes and incorrect/bad QoS. I doubt that by re-engineering the client-server protocol you can get more than 10-15% reduction which does not help in the end.
      What people can do is use routers with gaming/qos capabilities which might help with lowering the ping/packet loss.

  4. - during this stage (modelling, texturing), the models are several times rechecked by historical experts and model/texture experts, who write detailed comments on them
    That’s why SP model was completely wrong and you had to change it? Because “historical experts” checked on it so many times? Punny lads.