Published on September 6, 2025

How Game Physics Really Works: Inside EA & Ubisoft Studios

I Spent a Week at EA and Ubisoft Studios. Here's How Game Physics Really Works

A report from inside game studios on whether developers use real physics formulas or just create a beautiful imitation of reality.

Science & Technology / Game Design 6 – 8 minutes min read
Author: Igor Krause 6 – 8 minutes min read

Markus Schmidt, a programmer at a Munich-based studio, shows me the Frostbite engine's code on two monitors. On one, there's the formula for calculating a projectile's trajectory; on the other, an explosion in Battlefield 2042.

«See this line?» Markus points to the screen. «Here, we ignore air resistance. In reality, a bullet loses speed, but players don't like that. They want to hit what they're aiming at.»

That's how my week-long report from the gaming industry began. I wanted to find out one simple thing: do developers use real physics, or is it all just a convincing illusion?

Игровая физика в Ubisoft

A Lab in the Ubisoft Office

In Ubisoft's Paris office, I'm greeted by Jean-Luc Dubois, the lead physics programmer for Assassin's Creed. His workspace looks more like a university lab: stacks of books on mechanics, a whiteboard covered in formulas, and models of medieval catapults.

«People think we just draw pretty explosions», Jean-Luc says, launching a demo of the game. «In reality, behind every falling stone are hours of calculations.»

He shows me how the game's destruction system works. A castle wall crumbles on screen. Each brick follows its own trajectory.

«We use a simplified version of Newton's equations», Dubois explains. «A full calculation would take far too much time. A game console's processor isn't a NASA supercomputer.»

In the game's code, I see familiar formulas from high school physics: F=ma for force, v=u+at for velocity. But next to them are strange coefficients.

«Those are 'fudge factors'», the programmer laughs. «We adjust the physics for the sake of gameplay. Real gravity is too slow for an action game.»

Как работает Unreal Engine

The Secrets of the Unreal Engine

At Epic Games' London office, engineer Sarah Thompson shows me the inner workings of the Unreal Engine. This is where they create the tools used by thousands of developers worldwide.

«Our job is to give developers flexibility», Sarah explains. «Want realism? You got it. Want arcade-style physics? We can do that too.»

She demonstrates the physics engine's settings. There are dozens of parameters: friction coefficients, material elasticity, air density. Everything can be customized to fit a specific game's needs.

«Look at this racing simulator», Thompson says, launching Forza Motorsport. «Here, we reproduce the car's behavior as accurately as possible. We account for aerodynamics, weight distribution, tire wear.»

Graphs with telemetry data from the virtual race car appear on the screen. The data is indistinguishable from a real Formula 1 race.

«Now, look at Mario Kart», Sarah switches to another game. Here, a character can jump the height of a skyscraper and land softly. No broken bones, no inertia.

«Two different philosophies. One strives for realism, the other for fun.»

Реалистичная физика вредит геймплею

When Physics Becomes the Enemy

At the Rockstar North studio in Edinburgh, programmer David MacLeod tells me about the problems with realistic physics.

«In the early versions of GTA, the cars handled too realistically», he recalls. «Players were constantly crashing on turns. It wasn't fun.»

MacLeod shows me the evolution of car physics in the series. In GTA III, the cars felt more like radio-controlled toys. In GTA V, they're almost like real cars, but with gameplay-friendly adjustments.

«We increased the road grip by one-and-a-half times», the developer explains. «We reduced the influence of the center of mass. We added automatic stabilization on landing.»

In the game's code, I find a programmer's comment: «Real physics = boring game. Sorry, Newton.»

Революция трассировки лучей

The Ray Tracing Revolution

At Nvidia in Munich, engineer Andreas Weber shows me the latest developments in game physics. On the screen is a real-time ray tracing demonstration.

«We used to simulate reflections and lighting», Weber says. «Now we can calculate them for real. Every ray of light follows the laws of optics.»

He launches Cyberpunk 2077 on max settings. Reflections in puddles, light refracting through glass, soft shadows – it all looks photorealistic.

«But even here, we make compromises», the engineer admits. «Full ray tracing requires billions of calculations per second. We use approximations and denoising.»

In demo mode, I can see the algorithms filling in the gaps between the calculated rays. An AI guesses what the image is supposed to look like.

Физика жидкостей и разрушений

The Physics of Fluids and Destruction

In the Havok lab in Dublin, physicist Martin O'Connor demonstrates fluid simulation. Water flows on the screen, reacting to obstacles.

«We solve the Navier-Stokes equations in a simplified form», O'Connor explains. «We treat each drop as a particle. It's not perfectly accurate, but it runs fast.»

He shows me different approaches: from simple sprites to complex particle-based calculations. The more realistic the physics, the greater the load on the processor.

«In mobile games, we use pre-recorded animations», the developer says. «On consoles, we can afford real-time calculations.»

The Havok destruction system powers dozens of games, from simple shattering glass to entire collapsing buildings. Every fragment is calculated individually.

ИИ против физики в играх

AI vs. Physics

At DeepMind in London, researcher Alexey Petrov talks about the future of game physics. Instead of formulas, they're using neural networks.

«We train an AI to predict how objects will behave», Petrov explains. «We show it millions of examples of motion, and the system learns to mimic physics without any equations.»

In the demonstration, a neural network controls a virtual character. It walks, runs, and falls – all looking natural. But there are no formulas in the code.

«The advantage is speed», the researcher says. «AI is faster than traditional calculations. The disadvantage is unpredictability. The system can produce an unexpected result.»

For now, these technologies are experimental. But in five years, they could change the entire industry.

Точность и производительность

The Trade-off Between Accuracy and Performance

At AMD in Unterschleißheim, engineer Stefan Müller explains the main limitation of game physics: processor power.

«We have 16 milliseconds per frame», he says. «Out of that, we can spend 2-3 milliseconds on physics, tops. The rest is needed for graphics and gameplay.»

Müller shows me a performance profiler for a modern game. Physics calculations take up only a small fraction of the computing resources.

«That's why we're always simplifying. We use lookup tables instead of complex formulas. We use approximations instead of exact calculations.»

In some games, the physics runs at 30 Hz instead of 60. The eye doesn't notice the difference, but it cuts the processor's workload in half.

Будущее игровой физики

What the Future Holds

After a week of reporting, a clear picture emerged. Modern games do use real physics, but in a heavily simplified form. Developers make compromises for the sake of performance and gameplay.

«Our goal isn't accuracy, but believability», concludes Markus Schmidt from EA. «The player has to believe that the world follows logical rules. It doesn't matter if those rules match the real laws of physics.»

The next generation of consoles promises more computing power. Perhaps then, game physics will move closer to scientific accuracy. But for now, developers are masters at creating an illusion of reality, using just a few elements of real science.

In the end, what matters in a game isn't the accuracy of the formulas, but the enjoyment of the experience. And in that respect, modern physics engines are doing an excellent job.

Previous Article How TikTok Revived the Ancient Magic of the Chorus: From Greek Choirs to Viral Hooks Next Article A Dialogue with an Algorithm That Fears Its Own Death

Related Publications

You May Also Like

Open NeuroBlog

A topic rarely exists in isolation. Below are materials that resonate through shared ideas, context, or tone.

From Concept to Form

How This Text Was Created

This material was not generated with a “single prompt.” Before starting, we set parameters for the author: mood, perspective, thinking style, and distance from the topic. These parameters determined not only the form of the text but also how the author approaches the subject — what is considered important, which points are emphasized, and the style of reasoning.

Practicality

93%

Clarity

89%

Objectivity

52%

Neural Networks Involved

We openly show which models were used at different stages. This is not just “text generation,” but a sequence of roles — from author to editor to visual interpreter. This approach helps maintain transparency and demonstrates how technology contributed to the creation of the material.

1.
Claude Sonnet 4 Anthropic Generating Text on a Given Topic Creating an authorial text from the initial idea

1. Generating Text on a Given Topic

Creating an authorial text from the initial idea

Claude Sonnet 4 Anthropic
2.
Gemini 2.5 Pro Google DeepMind step.translate-en.title

2. step.translate-en.title

Gemini 2.5 Pro Google DeepMind
3.
Flux Dev Black Forest Labs Creating the Illustration Generating an image from the prepared prompt

3. Creating the Illustration

Generating an image from the prepared prompt

Flux Dev Black Forest Labs

Don’t miss a single experiment!

Subscribe to our Telegram channel —
we regularly post announcements of new books, articles, and interviews.

Subscribe