Dragons, Data Science, and Game Design

Dragons, Data Science, and Game Design

I'm a tabletop game designer. I recently built machine learning models to help with playtesting. However, the more I used AI the more I realized how important the human side of data was.

From basic machine learning algorithms to complicated neural networks, the AI playtesting models were only ever as useful as the people building and running them made them.

So I wanted to take a step back from AI and take a look at the role of data scientists. I felt the best way to do this was to look at all the mistakes I made when first using data for game design (I made a ton) because without those human errors, the AI tools wouldn't have had a functional foundation

I definitely have a lot of room for growth as an author. Please feel free to leave any and all feedback! Hope that mistakes made in this article make the next one better!

Key insights:

Sample size matters (its not just something your statistics prof rambles about)

Stratify your data!

Data drift can hit in unexpected ways, so remember the business case and don't get lost in the data itself

I will update the visual cues section. I also wrote a tips and tricks document for playtester which might have had a bigger impact than new art, so want to mention that as well

In you're more interested in the pure AI side please check out: How to Train Your AI Dragon