Dev blog #359 / IL-2 Sturmovik: Great Battles
Dev blog #359
Dev blog #359

Dear Friends,

At the end of last week, IL-2 series producer Daniel Tuseyev and studio director Albert Zhiltsov took part in a studio podcast. What was planned as an informal conversation on various (and not necessarily game-related) topics turned into a long discussion of the series. Since the conversation was in Russian, we've prepared a summary of the most interesting parts of it. Unfortunately, it's not easy to summarize more than two hours of conversation without missing something interesting, which is why our Dev Blog today is so long - simply click on the topics that interest you.

Readiness of the new project

If we talk about the content, it’s between a third and a half: the map is only approaching a half-done state because most of the time in the map development is taken up by debugging new technologies, and the sound and graphic effects are also about half done, the airplanes are less than a half. In general, there is still a lot to do. But if we talk about technologies, we have already finished about half of them — in principle, we have all the technological base, the basis for the new project, created and are finalizing the details. This applies not only to the game engine but also to the physical model because we change many things in the physical model too.

Approach to the damage model in the new project

In Great Battles the principle of the damage model was inherited from Rise of Flight — for it to work fast, when a bullet hits, say, a part of the wing and depending on the angle of the hit there is a probability that a part of the airframe has been hit. If you know the construction of the wing, then if you hit exactly from behind you will hit the spar, and if you hit exactly from above, the spar surface is small and the rest is just the skin, so the probability of hitting the spar is low. This is a fairly accurate model in terms of combat statistics because the caliber is considered — a 0.303 caliber bullet may not hit anything important, but a 20 mm shell is nearly guaranteed to do so. These probability coefficients were not invented from nowhere, they are quite finely tuned.
But time does not stand still, the genre is developing, and it is time to consider a specific hit location. Moreover, we have added the hit decals and now you can see exactly where the bullet hit, which raises questions — why did it hit my aircraft visibly in the spar, and it did not break? Now we model all the parts of the airframe and the control system as separate objects that can be hit, and as a result, the randomness of the events in the new project will be lower than it was in BoS (as in BoS it was lower than in Rise of Flight and Rise of Flight lower than in the old IL-2 1946). The level of detail naturally increases.
It's the same with the engine. The engine in BoS is detailed highly already — the cylinder block is a separate element, oil tank too, etc., all parts like cylinders, crankshafts, and fluid lines exist, but they don’t have small separate objects that can be hit — hitting a certain element of the engine is determined by probability as well (a certain cylinder is damaged, the generator, etc.), In the new project, they all will become separate elements — if the hit occurred specifically in the place where the turbocharger is, it will be damaged.
From the point of view of altering the combat situation, it will not affect it in any radical way. Why is it so? Not everyone knows how complex aircraft engines were and are, but almost everyone knows what an automobile engine looks like. Open the hood of a car, and look how compact the devices are packed there, including a large amount of wiring. And now imagine that a small fragmentation grenade explodes there — this is an excellent analog of an HE shell of an aircraft gun. Can you imagine a place under the hood where an explosion of this imaginary grenade will allow the engine to continue to work? This is a nearly unbelievable event; it will damage something important — either the wiring, fuel tubes or something else. So, the outcome would be pretty much the same in both models, statistical and highly detailed.
The combat survivability of multi-engine airplanes will probably increase — smaller elements like generators will be less likely affected by distant explosions. By the way, the detail of the current model is already impressive — the engine cylinders are modeled individually, that is, for example, the compression in the fifth cylinder may drop. This will directly affect the operation of the engine, but the player does not know about the damage to the fifth cylinder — this information does not get through to him. Perhaps we will tell the player more about what is happening to his airplane. If we just started displaying this wealth of information, for example, in the current technochat, many players would probably think that we have made a more complex game, they would be impressed by the amount of this information, even though we would not change the modeling of the engine at all. It's sometimes frustrating for us as developers that people think things are so much simpler than they are. Why is that? The thing is that the player, just like a real pilot, has only the information he gets from the instruments or smoke from under the hood which may mean that something is not good.
So, the new project will have more detail, but, likely, this won’t fundamentally change the outcome of a dogfight compared to the current model.

Why you need model details

I myself (Albert) am a scale modeler and I was interested in the detail of models: when you make a model, you already know, for example, the differences between one modification of Airacobra from another. Then there was the Jane's series games, where there was a virtual museum. But here is what I think is important — there should be not just a museum, but a virtual hangar. Just a beautiful model does not give an idea of the size of a P-47, most people do not realize the real size of this machine, it is huge! But a museum, to be called a museum, requires very high detailing of models, so that they can be seen up close, so even in Great Battles, well, when you fly in the game itself the details are quite enough to perceive the battle, but when you start to look at it lovingly as in a museum, the detail level is not enough. But in the new project, it should be possible to look at rivets to inscriptions on technical hatches. Right in the game we want to make an aviation museum and not only an aviation museum, where you can examine the equipment, and read its description. Perhaps something to turn, to open the cockpit, that's my dream. At the end of the podcast, we will talk about the "dream simulator," which is a part of it. This museum is not just a thing, it’s a starting point, where you choose an airplane, look at it, and if you want to fly it, you choose one of the missions prepared for it.
Every next simulator is always progress for us, in the sense that it's a further approach to our ideal. We want the spirit of that era to be present in this museum. It's difficult both in terms of the graphical interface (we spent a very long time looking for a solution), and in terms of the actual scene that surrounds the airplane. We had to ask an architect, and a man with a relevant education is now helping us, he suggested some interesting things.

About the history of the series

What's important from what we've just said is some changes in our mindset, because the mid-2000s was a difficult time for simulation developers — the world around us was changing, it was getting very simplified, casual products were making monstrous amounts of money compared to what we were making, and we were in our 30s at the time — that's about the age when you ask yourself "what am I doing with my life?". We had to answer the question "Are we going to change to fit the world around us or are we going to try to change the world around us?" During the development of the Battle for Stalingrad the first decision I (Albert) made was to try to change the gameplay to fit the new times. Back then we were faced with the task of bringing the series back from a dead end.
By the time we came to 1C, the series was formally closed — that's it, there would be no more IL-2. That time I decided that we could expand the audience of flight simulators through changes within the game (modifications that unlocked as your pilot was gaining experience) which was most likely my mistake, although it's impossible to determine that for sure now, because you can't change the past and try again. Now we decided to accomplish this not by changing its gameplay, but by creating more user-friendly elements of it, without touching the hardcore detail and complexity of the simulator itself.

On the problems in AI

Artificial intelligence of pilots — for me no question hits harder than this, because we did a tremendous job and this work was not appreciated by the audience at all. It was such a leap of faith — we were basically hitting the wall with our heads and I was sure that we were going to break through the wall, but we were smeared all over the wall and we still have a whole bunch of criticism of artificial intelligence. This is the biggest problem right now by the way — we need an artificial intelligence programmer who is familiar with navigation in three-dimensional physics-based space. If you exist, can write in C++, and want to solve problems that probably only the aerospace industry solves except for us, please write and come — we are very much waiting for you, we have a terrible shortage of personnel for this task.
Let's explain this problem a bit — there are two components to why this happened. First, I don't want to say that our artificial intelligence is bad — but it didn't match the expectations of the players. The problem is that if you take the closest competitors, they focus 95% of their attention on the player’s airplane, so the problems of the surrounding world are not so critical there. It's a serious indulgence, but they chose this way and it works for them.
In our case we emphasize more on the components of the game world, that is, we need the game world to be interesting, battles happen in it, and therefore the requirements for us are much higher in terms of artificial intelligence. It's pointless to complain about it, we bring it on ourselves, we could have done things differently — for example, we could have said that we are only doing an online project or only a museum where you can look at a nice airplane and open hatches. There are different options, but we chose what we chose, and in what we chose, artificial intelligence is a key module. IL-2 Sturmovik is a simulation of air combat, and that's something that players don't always want to understand.
It's not an airplane simulator, it's not a ground surface simulator, it's not a simulator of some other element. A lot of effort has gone into making this an air combat simulator, which is built from a combination of all these elements. And here comes the problem in the single-player game: you are fighting with artificial intelligence, this is the key element that determines the picture of this air combat. 
And here, because of an ambitious decision that was made many years ago, while working on Rise of Flight (my fault again), there is a very complicated feature that only our project has — our artificial intelligence flies the same plane as the player — the AI has no cheats at all. All this detailing that we did, aircraft vortex paths, etc. - oh boy how much it requires CPU performance when five Heinkel 111s are lined up on the runway one after another and blow air at each other.
We tried to make the AI comparable to a live player from PVP (where players fight each other) servers because we played on them all the time. There are hardly any pilots from the WWII era alive. There are printed manuals, but people fly differently. You can't simply simulate the fear that you are lost, that you missed something crucial, that it's 40 degrees Celsius in the cockpit of a La-5 in winter. In general, the behavior of people in PVP multiplayer was taken as a basis and the logic of decision-making was transferred from there to planes with AI.
Turns out that's not what people are expecting. When I damage a player on a PVP server, I will chase him to his airfield, I want him to be counted as my kill. And when the AI behaves like that, people complain that it acts illogically. There was such a historical overlap, that new requirements were imposed on the original basis of the AI, which led to problems in its debugging, there are errors that are quite difficult to eliminate. We tried to fix the chase to the airfield three or four times — we had to get into the basic system of switching AI states and build it on a new crystal clear, simple principle. And there's almost a Maslow pyramid of priorities. Different tasks are dumped on the AI — evasion from the ground, evasion from the threat, fulfillment of tasks according to the mission scenario, and fulfillment of commands given to it by the player, if it is a wingman AI. But the main difficulty is maneuvering in three-dimensional space — the AI can't just "move" to the desired point of space, as in the absolute majority of games, it has only the same flight controls as the player does, it has to plan a maneuver. So that it doesn't try to do all this at once and fail, these tasks should be strictly prioritized, a clear chain of tasks should be built, and then the AI behavior should become more understandable and predictable in theory — we want to do this too.
Our fanatical decision to make the AI exactly on the same FM that the player flies has a very obvious inevitable downside — each of them requires the same processing power that the player's plane requires and the player's computer counts the planes of all AIs in combat. Let's look at this situation carefully — when you are in a dogfight against a fighter, it is fundamental that it has the same FM as you because otherwise, it will be either too easy or impossible. In other projects or the old IL-2 Sturmovik, the AI did things that had nothing to do with reality. But in a situation where the player attacks a line of bombers, the task of these bombers is to stay in formation and shoot back, and in this case, an ultra-detailed FM is not needed, and a decent damage model is necessary. This does not mean, of course, that the bombers will start flying "on rails", as in the old IL-2, but the FM detail can be reduced in this case. In the new project, we have already solved this problem, launching about 90 airplanes in formation. Closer to the release we will tune this number (maybe everything will be limited by graphics and the number of polygons and textures) but from the physics point of view, it will be possible to have large formations. Finally, there will be a picture, which the series lacked — the sky full of bombers, and massive raids.

About the Pacific theater

We really wanted and still do want to make a Pacific project, real steps are being taken for this, and most likely it will be done (not in the next project, but after it) if we manage some technical challenges. Imagine an aircraft carrier or a cruiser, there are thousands of anti-aircraft guns and machine guns on it, and when it all starts firing real bullets and each of them is looking for targets, it will result in single-digit FPS right away. That is, you need, as in reality, a fire director, which selects the target, and then a lot of guns on it shoot at its general location.


That's also why we're completely redesigning radio communication. It will be more appropriate to the environment, a bit more lively, and better connected to AI commands. We'll try to make the radio exchange more interesting. The design document for the AI and radio exchange in the next project is 50 pages long.
By the way, in the Great Battles series, the radio conversations were recorded by enthusiasts, players — German players voiced German pilots, etc., for which we thank them very much. If we had just hired actors, there could have been mistakes in flight terminology or pronunciation.


In Great Battles they behave more like a decoration, in the new project they will influence the battle picture according to the specifics of their coverage, unlike a simple "detection ball" around themselves they will have a detection profile (it will be possible to hide behind a hill). It will be possible to get direction from radar to bombers. The radars will be still early models, not Doppler radars — this means that if an airplane is between an elevation and the radar and is flying exactly towards the radar it won’t be detectable on this radar.

The pilot as a separate entity

There is a chance to fulfill my childhood dream — in flight simulators, I have always been unhappy that the pilot and the aircraft are a single entity. I would like to be able to somehow inspect my airplane from the outside as if to go around it. If you jumped out with a parachute, you could wander around a bit, just for fun. Or even after landing to get back inside and fly away. Perhaps one day we will drive up to our plane in a jeep — come out of the briefing shack, drive to it, sit in it with the animation of what is happening, etc., while we of course hardly have time for that right now.

Graphics level

You can hear some players complain that our competitors are already more impressive to them in terms of graphics. Any of our attempts to explain that in IL-2 you get stable 60 frames per second in VR instead of a slideshow on a middle-range PC are met with the fact that the audience of flight simulators is not a mass audience and can afford expensive hardware.
The thing is that when we were designing the first Great Battles theater, Stalingrad, we were again, as mostly online pilots, focused on performance and so we sacrificed all the graphical beauties if they gave some big performance drop. We achieved that goal — it turned out to be a very smooth simulation, but modern reality says that people buy a pretty picture anyway. We are not abandoning the basic idea that a player should have a smooth running game, but we are increasing the system requirements, first of all, it concerns video memory, 6-8 GB VRAM will be needed. We moved to the PBR model, based on the physical propagation of light, and it implies twice as many textures as before, that's how it works. What does it give us? Very realistic surfaces, for example, matte paint (as on top of the P-47 hood to avoid glare), or on the contrary, oil drips, or visible scuffs — an aircraft just from the factory and after a few months of fighting look very different.

About the dirt

Among scale modelers there are two irreconcilable camps — some believe that models should depict a clean airplane, and others on the contrary try to imitate dirt. Each group has its arguments — some of them say, what kind of a fool would mess up an airplane to such dirty a state? Since it is an airplane, it's looked after and it should always look not like it was recovered from a swamp. There are a huge number of photos, especially Americans like to show them, where an aircraft shines like a diamond. And another group says — even after the first flight there are some chips, inevitable traces of exploitation. In the new project, we will have a rather late period, the airplanes are not new there, and here we are with our artists looking at a photo of a squadron, 20 airplanes in the picture, and we think — well, we can't do that, they look like old cars that have been open to the elements for many years.
I have a favorite photo — a mechanic is standing in front of some British plane and painting a number on its side, and in his hands, he has a stick wrapped in some rags, and you can almost see how paint drips from it. As a modeler it hurts me to look at such a thing — in scale modeling, we glue beautiful decals so they look impeccable, etc. Modelers try to convey the shine of aluminum polished or lacquered using various methods, but when you look at many photos — there is no place of aircraft skin intact or clear, the kind of shine of aluminum one expects to see.

Break (Albert)

Sorry, went to move the car — I went out for a smoke, and there is a huge excavator (we have a big building renovation here and there is no more parking lot for now, a pipe was under it which of course urgently needed the service of the water company) maneuvering less an inch away from my car. A driver's nightmare. I said to the excavator operator — perhaps my car interferes with your work, better move it? And he had a sense of humor — he replied “Don't worry, nothing will happen to the excavator.”

Whether to ditch bombs

Often we are asked about some situation in combat. For example, planes with bombs do not drop them when they are attacked by fighters, or on the contrary, they drop them when from the player's point of view they should not have done so. There is no correct answer in principle, because here’s an example. I fly online, so here’s a story — here we are, flying Focke-Wulf fighters carrying bombs, we have a mission to bomb a certain place. It took us a few minutes just to get the six of us together. Then some guy arrives in a Yak-1b and starts attacking us since we have bombs, we are slow, etc. So what, should we all drop the bombs and start circling with this Yak, completely disregarding the task, or try to take advantage of the numerical superiority and engage him with bombs, or do something else? That's the dilemma, it has no unambiguous solution because so many factors determine even this simple at first glance situation.
Another example — Messerschmitt 110s carrying bombs to the target, and they are twin-engine heavy fighters, and they are attacked by LaGG-3s. The Messerschmitts used to drop all the bombs and we were told, what a stupid thing your AI does? My pilot’s career stops, I can't fly to the target, and all my wingmen drop bombs at the sight of fighters.
You can remember the bomber formations who were explicitly ordered not to break formation, which was emotionally hard — you see your wingmen burning and falling and you have to fly forward, especially on a bomb run when you have to fly straight and level. 
Or with the US squadrons, for instance, they had one flight of Mustangs carrying bombs, and the other flight covering them without bombs. And here you have a combat flight in the game — what to do? It turns out, the AI has to somehow analyze the complex situation as a whole — maybe there are friendlies without bombs in the vicinity and they will cover me? And how far away are they? That is, a seemingly simple situation "to do this or that" turns into a whole branch diagram of options. Also when humans play, they are capable of spontaneous behavior, unlike the AI.

How a decision is made to create a new airplane

Now let's answer with an example. An opportunity arose to make a map of Odessa and then the thought immediately arose — what if we complete it with aircraft? Like this. In Odessa, there are two episodes — the defense of Odessa in 1941 and its liberation later in the war. For the later episode, we are thinking of Yak-3 and La-7, right now we are in final negotiations with the contractor who will make the models, and I think they will be successful. In the same way, the early episode of Odessa will include I-153, which is now in the works. The most characteristic airplanes are selected, and there will be others.
Sometimes there are spontaneous solutions, like the Airacobra, which is Albert's favorite plane. Or like the glider, how did we decide to make it? Because our partner said that it is their dream to make a glider. Of course, we carefully analyze the possibility of returning the invested funds, because you, our customers, tend to buy certain ones and the financial side of the issue is important. The Ta 152 will be of interest because it is a rather unusual machine from the late war period and people will want to fly it in multiplayer games. But not everything depends on potential success in sales — its importance to the theater in question is also, well, important since it is necessary to recreate a historically correct picture of the battle. Collector Planes, on the other hand, can be added just because they are interesting and popular. In general, the development of one aircraft takes up to a year and sometimes more, especially in the case of bombers.

About support for 3D monitors and other technologies

The question here is "why 3D monitor and 3D TV technologies have not been further developed?". It's a difficult question, our leading programmer is a big fan of the very idea of 3D glasses and monitors, and he developed the VR mode. 3D glasses have advantages — for example, your face does not sweat in a rubber mask of VR HMD, but, unfortunately, this tech somehow stalled and lost support from manufacturers, and then VR appeared. In general, there are a lot of various niche targeted devices, but we can't spend our efforts and resources developing (and most importantly then supporting for many years) something unpopular if there are few users of such technologies.

About the realism of the game

I like realism and do not replay unsuccessful sorties. And here was such a moment — I, already twice a Hero of the Soviet Union, was returning in an I-16 near Moscow. I had already crossed the front line and already saw the waving of handkerchiefs of the enthusiastic crowd in my imagination, but when the career is set to complex you can meet enemy and friendly flights on the whole route, and I forgot about it and was sure that I was already safe. Then blam — black screen, twice Hero of the Soviet Union shot down by a fighter on a free hunt.


The game's interface since Rise of Flight was built on Scaleform technology developed by a famous and huge Autodesk company. Who would have thought it would be shut down? Even if we wanted to continue to use it in a new project, we couldn't do it, you can't buy a license anymore, and so we had to create an interface from scratch. It's for the best because it's already hard to find people for that old technology. The interface is an important part of the game — half of it happens in the interface, the career mode, etc.

Transfer of Great Battles airplanes to a new project

For example, the airplanes of Flying Circus were transferred from our previous project Rise of Flight, so that is possible, but it is a matter of time and cost because their visual models will have to be redone.


We had a very important structural change in the engineering department: we have a group of veterans who are involved in the new technology development and modernization. For a very long time, we had one Andrei Petrovich there, a very talented person, but he was alone. Now this has changed.


How do we have time to do both a new project and produce a lot of stuff for Great Battles? Because we have created cooperation, we have partners with whom we produce new things, including the Odessa theater. In principle, both maps and aircraft will be produced by external developers, except the flight model, which we are developing internally — it is so complex and inside the code, that we can't find engineers who would be capable of it and who we won’t offer to work for us. Right now we are using about 10 percent of our internal resources to develop new Great Battles content, but thanks to a lot of external cooperation we are developing more, for instance, the Odessa module. We are expecting it in 2025, but we would like it sooner, of course.


Sometimes people get "burned out" because they read nasty things on the forum, and insults — a community manager is usually a sincere person, he sees how the team works, but he gets offended — he and we work every day, we give up a lot, and then someone writes something like "you are useless assholes", etc. Kurt in particular, our previous community manager, had a hard time with this and always said "How can the world be such a shitty place, how can people not see all this effort, why they don't appreciate all this work?". Anyway, not everyone is capable of working with people regularly. But Luke is a tough guy, hang in there, bro! It's not easy to make money from an airplane simulator in these challenging times of tiktok mentality.

Our team history

We have been through the third phase of our lives as a team. The first stage was a "group of friends", a stage of incredible enthusiasm when we made "War in the Skies", which turned into Rise of Flight. We were young, we were 20-something years old, and we understood absolutely nothing about how the world works. We were ready to eat instant noodles, and borrow money from each other, realizing that what we were doing was not a mass product and we would not make millions — it was a dream. We traveled this path strictly on enthusiasm, working at night.
Then came the stage of "maturity", of growing up, when we made Battle of Stalingrad, Battle of Moscow, and beyond, when we became more mature and began to question whether we were fools. It was a time when a lot of arcade projects on similar themes were coming out, where nobody bothered with real physics, and in terms of labor, they required much less investment and solving the problems related to this. We added unlocking modifications for experience points to make the game more interesting for newcomers — the player took it as an offense. This was done because we felt like we didn't understand the world, we needed to adapt.
Now we are in the third stage, the "grey beard". On the contrary, we have concluded that we do not need to bend to the changing world and that we already have enough value, willpower, and respect for ourselves that we will do what we believe in. Let me (Albert) be a naive dreamer, but I believe that the audience can appreciate, that in this smoothie-plastic world, those who manage to keep their direction and create an unusual value product have a better chance of getting a loyal and financially secure audience (we are not trying to get money from children). Of course, it's a huge risk because the labor and financial cost of assembling a team of unique specialists as I said is significant. Literally, they can work (and many did) in aerospace companies. There's at least this point — if you want to make enough money to buy a supercar, this is not the genre. To make this kind of money you have to put the lid on your desires unless you desire a life of crime. If you want to create something truly valuable to people, you can't cash in a lot of money from them. On the other hand, the Great Battles series is in good health — there are no super profits, but the series pays for itself and we earn our living, which is a very serious achievement in this genre, my friends.

About our vacancies

Here’s another question, "What is the minimum knowledge a candidate should have as a map artist, for example?"
A map artist can be a very good texture artist, may have skills in technical modeling, and maybe a surveyor, that is, can come from a variety of directions. Fill in the form on our site and you will be contacted in detail, given a test task, and in this case, you’ll get a choice in what you want to do as a test, because different skills are needed. Your desire to learn something is also important — we understand that you can't learn our technologies anywhere, we invented them, and there is no textbook on them. You will have to learn a huge amount of very focused information that you are unlikely to apply anywhere else but working for us. How does an airplane simulator build the Rhine Valley, for example?  Not the way the locations are built in popular universal game engines, where a location is a thousand feet or a couple of miles large. It's a complex system that allows you to simulate maps hundreds of miles wide. So if you're just from the gaming industry, it's unlikely you've worked with this before. And this and other systems are gradually improving, and changing.

To finish our podcast — if you want to learn more about the history of aviation, why air battles happened the way they did, and why people wrote in memoirs as they did and not in another way — our project is great for this. See you in the skies!