Releasing Parsertime's Dataset to the Community

Today, I'm excited to announce that Parsertime's dataset of over 1,000 player-uploaded scrims, 430,000+ player kills, and nearly 300,000 rows of player stats is being released to the community. Parsertime has always been built with an emphasis on transparency and accessibility, with the source code having been open source since day 1. In that same spirit, we’re opening a portion of our anonymized gameplay dataset to the community.

Whether you’re a data science student exploring competitive analytics, an analyst studying gameplay trends, or an engineer experimenting with machine learning models, this dataset is meant to serve as a foundation for your work.

You can get started by downloading the dataset via GitHub or Google Drive.

Staying True to Our Privacy Promise

While we’re excited to share data, player privacy remains a core commitment. As outlined in previous updates, we’ve always taken data handling seriously—from improved error messages and privacy policy expansions in earlier releases to ongoing backend improvements that keep our services stable and secure.

To strike the balance between openness and privacy, all identifying information has been replaced with deterministic pseudonyms. For example, player and team names show as follows:

Player1 -> P_f081f9e5
Player2 -> P_a92c1b44
Team 1  -> T_3c10ae19

These pseudonyms are:

• Deterministic — The same real-world player or team always maps to the same pseudonym within this dataset.
• Non-reversible — Without access to our full production data and internal encryption keys, it is virtually impossible to identify a real player.
• Consistent across tables — Allowing accurate joins for multi-table analysis.

This ensures engineers and analysts can study performance trends, hero interactions, map flow, and macro gameplay patterns—without exposing any personally identifiable information.

What’s Included

This initial dataset includes anonymized versions of the game-event tables that power many of Parsertime’s analytics tools. You’ll find data for:

• Eliminations, kills, and assists
• Ultimate usage and charge timing
• Objective captures and map progression
• Hero swaps, spawns, and round transitions
• Player performance metrics at the per-map and per-round level

These are the same core building blocks we use internally to drive visualizations, dashboards, and performance summaries.

How to Use It

The dataset is available in two different formats to make analysis accessible to everyone:

• SQL dump: Ideal for users who want the full relational schema.
• CSV files: Perfect for data scientists using pandas, R, or spreadsheets.

The repository includes a README with details on how to use each dataset, with included instructions on how to get a Postgres instance up and running from the SQL dump. Additionally, you can view the ScrimTime schema here.

Whether you're analyzing ultimate economy trends or building a model to cluster teamfight outcomes, you’ll have access to everything you need to begin.

What’s Next?

Just as we’ve continued to evolve Parsertime's performance, stability, and feature set over time (from debugging tools to codebase refactors to new map support), this data initiative will continue to grow alongside the platform.

And of course, community input will shape where we go next. If there’s data you’d like to see added—or if you build something cool with the dataset—share it with us in our Discord or open a discussion on GitHub.

A Thank You to the Community

Parsertime has always been a community-driven project, shaped by your passion and your feedback. From day one, your support has fueled every release—from UI overhauls to accessibility improvements to behind-the-scenes engineering work that keeps the platform running smoothly.

Releasing this dataset is our way of saying thank you and continuing to support the analysts, coaches, players, students, and researchers who push the competitive scene forward.

We can’t wait to see what you build!