Search Unity

On March 23rd, we announced Unity Game Simulation at the Google for Games Developer Summit. Unity Game Simulation is a cloud service that enables game studios to complement their pre-launch playtests through the power of simulation. Supported by Google Cloud’s infrastructure, Unity Game Simulation helps you evaluate your game balance at scale. Unity Game Simulation is now available in beta. To get started for free, sign up here.

A highly engaging and captivating game is all about making sure the gameplay is balanced. Every game, whether mobile or PC, hyper-casual or open-world, online or offline, requires thorough balancing. Launching an imbalanced game could negatively impact the player’s perception of the game.

Spreadsheets are one solution that many game designers use to model elements of the game. However, these spreadsheets oversimplify and inaccurately represent real gameplay. Human playtesting is another solution for testing and balancing, especially for highly subjective elements of a game. However, playtesting can quickly become expensive and run into scaling limitations.

At Unity, we envisioned a solution that leverages thousands of gameplay simulations in the cloud to provide more accurate results and to allow playtesters to focus on the subjective aspects of the game.

“Even well-developed games fail when they are not properly balanced. Traditional methods to balance a game rely heavily on human playtesting and countless, manual redeploys. The entire process is costly, slow and inefficient,” said Danny Lange, VP of AI and Machine Learning, Unity Technologies. “A better way to balance your game is with Unity Game Simulation, where you can run thousands of gameplay simulations in the cloud that can provide more accurate results to leave playtesters to focus on subjective criteria like game feel and the fun factor.”

A better way to balance your game

We are excited to announce Unity Game Simulation, a solution that complements playtesting by using the scale of the cloud. Unity Game Simulation allows game developers and designers to test more and iterate often over changes in the game. Simulations happen on actual game code, generating accurate data, not just approximations. And since we are leveraging the power of the cloud, studios can speed up testing, achieve massive scale, and obtain results in a matter of hours or even minutes. All of this is available on-demand at just a fraction of the testing budget.

How Game Simulation works


SDK and Unity Editor interface

We provide an easy to install Unity package that includes all the tools necessary to set up your game for Unity Game Simulation. The package allows you to configure parameters, instrument metrics, and upload your build. Instrument what you want to measure in the game with metrics. For example, one metric could be how long it takes a player to play through the game.

Configure parameters in your game that you want to adjust for each simulation. For example, you may want to try various combinations of speed boosts and power-ups.

After instrumenting and configuring, upload your game build directly from the Unity Editor and begin creating simulations in the Unity Game Simulation dashboard.

Setting up game parameters for Unity Game Simulation in the Editor

Game Simulation Dashboard

To help you easily create and manage your simulations, we provide an intuitive dashboard. Here, you can specify the parameter combinations you’d like to simulate in order to achieve your desired metrics. The UI is accessible by any user in your organization.

The data from simulations is available as raw data for individual simulations and as summary statistics for aggregations. All of this is accessible from the dashboard.

Balancing thousands of weapon combinations for a top-down arcade shooter

Furyion Games, a Toronto game development studio, created Death Carnival, a fast-paced top-down shooter with extreme weapons and online multiplayer mayhem. One of the features of the game is the weapon socket system. Each weapon allows players to select a base weapon, ammo, weapon module, and temporary power-ups. Each of these combinations is made up of many variables that control the playstyle and efficiency of the weapon. 

Furyion used Unity Game Simulation to run thousands of simulations, helping them identify the right combinations of base weapon, ammo, weapon module, and power up for their desired survivability and level completion time. This allowed them to ensure the weapons were balanced for maximum enjoyment.

“Unity Game Simulation helped us find the right parameters to create the best experience for our players,” said Herbert Yung, Studio Director at Furyion Games. “By hand, the process would have taken many more human balancing hours and downtime between reconfigurations. Going forward, we’re excited to use Unity Game Simulation to help solve level design problems and find ideal gameplay settings.”

Correct rubberbanding settings for a racing game

iLLOGIKA, a sixty person Montreal game studio, created Rogue Racers, a racing game in which heroes compete in multiplayer races across the world in gorgeous 3D environments. In Rogue Racers, when a skilled player plays against an unskilled player, the skilled player should finish no more than five seconds ahead of the unskilled player in almost all cases. This concept is called rubberbanding. If the skilled player pulls too far ahead of the unskilled player, then the unskilled player might think the game is too hard. Conversely, if the race is too close, then the unskilled player might think the game is too easy. Finding the right settings to achieve this can be challenging.

To solve this, iLLOGIKA implemented a catch-up mechanism that grants the losing player a speed boost. The boost is determined by three parameters: a speed multiplier, a frequency, and a duration. However, these parameters generate thousands of potential combinations. After running through these combinations through simulations, iLLOGIKA was able to identify the ideal settings for keeping 95% of finishing times close to 5 seconds. 

“By leveraging Unity Game Simulation, our team was able to run different combinations of in-game scenarios to find imbalances early in the development process. We believe Unity Game Simulation will play a big part in improving player experience in the future as well as help shorten the time it takes to soft launch a new title successfully,” said Régis Geoffrion, Studio Director at iLLOGIKA.

Try now for free

Unity Game Simulation is currently available in beta. To get started for free, sign up here.

Have feature requests, want to discuss your use case, or have general feedback for us? Please reach out to us at or visit our Unity forum.

9 replies on “Optimize your game balance with Unity Game Simulation”

Thanks, from reading the comments I think I have a better picture of what this is.

We create a player in our game that can play it by itself, we sort of create the main player’s ‘AI’. Then we upload it, and Unity runs the game many times, and gives us the resulting data on how our main player’s AI did.

The marketing of this mentions that some developers use humans to test their game… which we then assume will be replaced by Unity Game Simulation. As though Unity has a human-like AI that will play your games for you many times, and learn how difficult it is. Kind of like Google’s AlphaGo AI. But they’re actually handling more of the running of the game, doing many iterations quickly.

How exactly is the simulation run? I’m assuming you do not have an AI to play any time of game a developer might upload. Also, what data is returned?
This blog post, and any other mention of this system seems to completely avoid talking about this at all…

I was asking me the question indeed.
So, the metric you receive from the simulation will still be biased by the way you implement your AI.
Can be handy, but will never replace human behavior in a game.

So you have to create a version of your game that plays by itself, taking the variables as input.
Then the service runs this self-playing game while recording some metrics.

That could come in handy, but can you please make very clear that your game will NOT be played by AI from unity that acts like a human player. That you have to build a game that can play itself without help from unity or “the cloud”.
This is not stated anywhere so people think you’ve built an AI that can play games like a human does.

Comments are closed.