How developers procedurally created a bunch of "junk" mobile games and earned 50 thousand dollars

Two indie developers talk about how they used automation, one Google Play account and the only game template about slot machines to create and publish more than a thousand applications.

Two game developers spoke at the Game Developers Conference this year, able to write a new chapter in the dystopian novel about the future: they told the story of how they made money, allowing robots to do all the work. In their case, this work was the procedural generation of games for smartphones.

Participation in a regular game jam led to the creation of a data processing machine, which as a result managed to earn a decent amount: 50 thousand dollars for two and a half years. Years later, with data (and money) in hand, the developers of this game-making machine, which created only “garbage” slot machines with a free-to-play model, used the GDC as a platform to give an alarm to an industry in which “ correct ”behavior is often limited to attention to players, fawning on publishers and collecting positive reviews from critics. In the case of these developers, violation of all these rules led to too successful results, causing them great discomfort.

Win the race to the bottom

Back in 2013: at that time, two video game developers have been trying for several years to succeed in the growing mobile games market. One of them, Alex Schwartz, helped create a quality Jack Lumber mobile game . (In a previous life, I even wrote a good review about it in the already non-existent magazine The Daily , which worked only on tablets.) The second, Ziba Scott, developed a nice mobile puzzle game Girls Like Robots .

Both games used the “pay once” model without microtransactions. Both received awards, fame, good reviews at exhibitions and gained publishers. Both failed to succeed.

Both developers estimated their modest income earned on the “right” path paved by traditional publishers. Then they looked at the iOS and Google Play stores and saw that “free games, clones and trash” dominated there.

In one of the daily conversations about the realities of business, this duet came to the conclusion: “We can achieve better results ... by going the worst way!”


One example of the appearance of automatically generated slot machines for Android devices. All this was complemented by a text-to-speech song about curling, which was played with every jackpot.


Thanks to the use of Selenium, the development duo was able to automate mouse clicks, data entry, and the selection of options needed to automatically download each game.


The first statistics.


Surprise: these games make money.


Some of the results obtained at the end of the experiment.

They joined forces during the Global Game Jam 2013. Schwartz and Scott decided to create something reminiscent of the "race to the bottom" that they observed on mobile platforms. Due to the limited time allotted for game jam, they decided to buy an Asset three-dimensional slot machine from the Unity Store for $ 15 (this is a store that allows game developers to pay modelers and animators for unlimited use of various two-dimensional and three-dimensional resources). They spent the remaining time of the jam on creating a system that will automatically generate all the design necessary for the simplest virtual slot machine to be unique enough to be published as its own smartphone application.

“Let's adapt them the way other slot machine developers do,” Schwartz suggested. “They create thematic machines. What is the minimum set of elements you need to change to get another slot machine? Change the name. Change one image on the drum, which may be related to our theme. Let's say we make a slot machine about dolphins: we’ll place a dolphin there [like the jackpot logo] with a special icon. We’ll make a background image of a dolphin with scrolling. ”

A list of words was created manually based on what seemed interesting to the development team; at the same time, the names should be fairly general and acceptable. The most daring of the names was “3D Sexy Librarian Slots” (“Three-dimensional automaton“ Sexy Librarian “”).

Having dealt with the graphics, the duo proceeded to the next step: creating their own music.

“The game should play a shitty little music, and the text-to-speech mechanism should sing the word“ dolphin ”in the monotone voice of Google Translate. It will be played every time the user wins. She must pronounce the name of our game to music, ”Schwartz suggested.

As a result, after pressing a single button, Unity script was able to combine these steps and automatically generate hundreds of “unique” slot machines. Schwartz and Scott report that their public domain automated image capture system ran into one problem: Google Image Search generated errors about exceeding the request frequency. “We had to use Bing,” Schwartz told us in a telephone interview. “Many aspects of his image search were less demanding. I'm not trying to offend anyone, but this search engine has a reputation for being second-rate. Therefore, for our project, he was in a sense a kindred spirit. ”

"Portal to a better world"

Having prepared a template for creating “one-armed bandits”, the team automated the process of transferring information to Google Play (at that time it was much easier to operate than the iOS store) and creating free advertising applications. The process was completed after writing one simple Selenium script .

Now developers could feed one keyword into a system of scripts that spent several hours assembling, and then watch in the “virtual monitor” how the system imitated human behavior: it put flags, checked each country, agreed to the terms of use and Fill in all the text fields, and then take the automatically generated slot machine and load it into the store.

Schwartz and Scott connected the Playhaven mobile advertising network to the scripting system because their philosophy was not to take real money from users who would download their applications with crazy names. Then they simply “forgot” the project for two months. After working for some time on their real job, they decided to take a look at income and ad statistics. The developers were amazed: people downloaded their applications, and 27 percent of these people clicked on their ads, making it approximately $ 211 a day.

The team had a theory: “All of our key advertising words were related to the casino,” Schwartz told us. “We thought about the reasons for such success, and it dawned on us: the game looked just awful, but people for some reason downloaded it. Then they saw an advertisement for a much better slot machine or casino, and clicked on it ... well, of course! In a neighboring meadow, the grass is always greener! The quality of our games was so low that these advertisements probably seemed to users as a portal to a better world. ”

Nevertheless, the development duo reluctantly admits that the average rating of many applications was approximately equal to four stars, and the reviews turned out to be rather supportive. Scott remembered one review of the automatically generated 3D Bowling Slots application: “The user wrote that he was disappointed that the bowling theme was poorly disclosed in the slot machine.”

The true story of supervillains?

In the end, a good amount of money came out.


One of the developers favorite slot machines: “Slot machine of an inexperienced long-eared owl 3D”


The automation system took into account the restriction on downloading on Google Play (no more than 15 applications per day). It worked automatically long enough and was able to download about 1,500 applications.


A slot machine template that the team eventually abandoned.


That's how “popular” the warez of developers has become (the sixth page for “the best new card games and casinos”).

The engineering part of their essence wanted to test how far this project could lead. Therefore, they began to experiment with the finished template, adding new elements to it, for example, automation of slot machine descriptions.

Google Trends was a good idea, but they usually led to the creation of trademarked names or to the names of celebrities, and the developers were not interested in breaking the “automation” cycle in order to solve the problem of many requests to remove applications. Therefore, they decided to use a slightly updated template for application names: the word “3D”, plus the adjective, plus the name of the animal, place or country, plus the word “free”.

Examples of titles: 3D Tremendous Face Pain Slots, 3D Rough Elbow Slots, 3D Mild Dogwood Slots, 3D Viceroy Butterfly Slots and 3D Inexperienced Great Horned Owl Slots. (Over time, the developers even bought T-shirts to capture the memory of the last name on them.)

Schwartz and Scott also paid some pennies to a small Romanian studio to create a better slot machine, which they eventually abandoned. This was partly due to the fact that new ongoing work and ethical considerations overcame their admiration and willingness to continue the experiment.

“Someone told us that on this idea you can make money, or sell data to someone, or sell your company,” Scott says. “We were at a crossroads - our joke began to resemble the true story of a supervillain. Will we abandon all our creative aspirations in order to create an enterprise for actively making money on slot machines? Or will it remain a tiny part of our lives, taking 1/20 of the time? ”

Over time, the development team touched on the problem of working with Google Play. Applications began to be deleted for violating the updated terms of service, which gave Google more freedom in the destruction of obvious garbage. In addition, Google regularly began to update the web interface. It was enough to shift the only flag by a few pixels to break the entire Selenium roboclick script that the team developed to automatically download 15 applications per day (at that time, this was the restriction of Google Play for one developer account).

At some point, the Playhaven ad network made an explicit statement. “We are seeing chaotic data in your account,” Schwartz retells this letter. “We do not quite understand what is happening, and we are not interested in publishing ads in your slot machines anymore. But we need to clarify the situation: you do not violate our terms of use. It's just a little ... inconvenient for us. "

Then Playhaven added: “Your users are the worst. People coming from your apps don’t spend money. ” The team immediately switched to another ad provider, Chartboost. “They roughly knew what we were doing,” Schwartz says. “And that was a great partner.”

“Optimized to remove our content”

After many years, all one and a half thousand applications generated as part of this experiment are already dead. Schwartz and Scott believe that the whole scam was partly a joke, and partly an alarm.

“You can almost jokingly say: thanks to the fact that we created a noticeable goal in the form of low-quality junk applications, application stores optimized their algorithms in order to remove our content,” says Schwartz. “In practice, we can say that our company trained their algorithm so that it no longer allowed this. But we were the first to raise this system to such a level. ”

In his presentation at the GDC, Schwartz presented information mostly jokingly, but also warned about the human factor: he talked about problems that could arise if someone starts to realize the dream of automatically generated applications not for fun but for money.

“This project was too tempting,” Schwartz explained at GDC. “We turned our back on the mobile market, which disappointed us in 2013, and always perceived our project with humor ... In fact, I am not at all interested in slot machines. We began to move on. If this is some lesson for business, then it consists in the following: realize your crazy ideas, experiment. But even if they give a result ... be prepared to leave, because perhaps it was the attempt, and not the result, that made you happy. ”