"When we started we knew what we wanted to accomplish [matching people] using machine learning, but we didn't know how to get there. Metal Toad figured it out for us."
- Russell Welch III, RPGMatch President
Despite the growth of TTRPG players and tools for playing online, finding a group that matches your play style, game preference, availability is still essentially in the digital Stone Age relying on message boards or groups on other networks.
RPGMatch's goal is to build a social network, with the aim of matching players, to help them find the perfect group for their TTRPG.

Business Problem: Matching Players
Playing TTRPG goes beyond just playing time and games. There are play styles, safety tools, and play environments. Complicating things further, many games last more than one session and can span months or years. To solve this, RPGMatch built a questionnaire with over XX questions. As players answered the questions and engineers looked at the data, It became clear that there were too many data points to program an algorithm that is accurate and robust.
RPGMatch realized they needed a better technical solution for matching players on their social network.
Technology Solution: Machine Learning to match players
RPGMatch approached Metal Toad with the need to use their user data and match users together into potential groups for playing TTRPGs with.
Metal Toad set up a webhook for API Gateway, to get new surveys as they were filled out. The data was processed and stored in Dynamodb, where it could easily be analyzed and Metal Toad data scientists could decide what features were needed.
Once the preliminary data review was done. Metal Toad used SageMaker to try different machine learning algorithms. After several tries, Principal Component Analysis provided the best matches with the current data.
With the matching algorithm ready to go. Metal Toad integrated it into the data pipeline to make the whole process seamless.
Impact: Quick Matches
With a complete data pipeline with integrated ML, users can now be matched in seconds. Using ML instead of traditional programming has meant that RPGMatch has gotten matches quicker and they are already looking at how to improve matches by iterating on their questionnaire.