Talk Title: 1,500 Slot Machines Walk into a Bar: Adventures in Quantity Over Quality

Effective talk: Tricks for design

Year of Talk: 2019

Video Link: 1,500 Slot Machines Walk into a Bar: Adventures in Quantity Over Quality

Journey to the slot machines:

Setting the Scene:

Starting in the 2010s when the presenter released paid phone apps.
We weren't getting many sales at the time, while free games were getting a lot of downloads.
During this time, they realized a lot of free apps were just bad quality games and didn't have any substantial gameplay mechanics behind them.
So they reworked a lot of the casino games to have better gameplay mechanics.
With this, they started to make some money and, after the holidays, had earned a few thousand dollars.
The common narrative for making mobile games is that they should be decent to have a high enough retention of users.
With higher retentions, it lets more ads be played and gets more money.
So they decided to make hyper-specific bad games that targeted certain keywords and were of bad quality.
The bad games would cause players to get bored easily and want to switch.
Before they leave, they get hit by the perfect ad that incentivises them to click on it and switch over to a new game/app.
This lets let get an even higher clickthrough rate and improve the money they get.
Proceed to use a previously created theme generator, which would create slot machines related to a given theme or set of words.
A few dozen to a few hundred new slot machines with different themes would be made a day.
They proceeded to automate a process in which their website could continually generate ideas for slot themes, send them to Unity to create, and create a package that a Chrome extension could detect and publish on the Google Play.

Results:

A year in 750 slots with unique themes has been made.
235,067 downloads total.
More positive results than initially expected. An expected amount of disappointment.
After some time, the spam/AI emails started to come in requesting collaborations.

The next step was to determine how to make the whole process more streamlined
Started by improving the tags being used on the Play Store, as well as reworking the keywords and themes used to create new themes for the slots.
When going over results from Google Trends and getting words that wouldn't cause any legal problems, they would test out the remaining words.
A basic title system was made with the following: 3d + [adjactive] + [animal],[location],[noun],[country] + Slot Free.
The use of 3d was meant to help get the game on top of the list when listed alphabetically.
They hired a Harvard phd psychologist to help decipher the data they had collected.
Also used Amazon Mechanical Turk, which are people who do tasks to complex for automation, but that might be time-consuming.
Helped properly tag all of the games and themes.

Observations from the Harvard PHD:

Themes related to s** tended to perform well while people and objects did not, as they normally were polarizing.
Animals tended to return a mix of bags of results for what people liked or didn't like.
An area they hadn't properly tested yet was whether improving the quality affected the stats in any way. At one point, they started reworking the project, but it was never published.

Around this time, they had come to terms with the fact that their system was essentially a lot of scaffolding that could break at any moment, and they needed to pivot or get ready for a crash.
When they were planning on what to do next, the advertisement they had been working with decided to cancel the contract as it was inconvenient for them.
Google also started updating its TOS, targeting the type of apps that the team was making. Those are repetitive apps with little to no difference between them.
At this point, they stopped updating everything and let the system they had built slowly break as updates came out and their word banks ran out of words.

Did the team harm players?

Did the team harm the game developer?

The same game dev tool where used. And followed many of the same rules.
Also, help identify players for the advertisement company that weren't spenders.

Final stats:

GDCTalksToText