Riot aims to address roughly 3 billion annual player reports across League, VALORANT, and Wild Rift with automated detection and machine learning

The Player Dynamics team at Riot Games deals with an obscene number of in-game player reports, but the developers are planning to implement some new evaluation methods across their primary titles to reduce the amount of chat abuse and in-game toxicity, especially amongst repeat offenders.

In an Aug. 29 update on Player Dynamics, Riot said that in 2021, it averaged about 240 million reports across its game catalog, resulting in an annual total just under the 3 billion mark. By its own estimates, if every Riot employee spent 365 days a year only reviewing reports, each person would need to review “about six reports per minute to keep up.”

Resolving reports isn’t a straightforward process, according to the Player Dynamics team, who exclusively focuses on improving player experiences and reducing harmful ones. Many players make reports for behaviors that aren’t deserving of penalties, and others will send “deliberately malicious” reports. Additionally, behaviors like intentional feeding and trolling are much harder to detect than clear text/voice chat abuse and AFK behavior.

But Riot’s existent methods and penalty systems are doing well. Riot reports that of all the players who received a report in 2021, less than 10 percent of them received another report within the calendar year.

With the data received over the past couple of years, the Player Dynamics team at Riot has been able to develop some new ways to evaluate player behavior. Automated voice evaluation is set to be added to VALORANT later this year, with each report it receives set to improve its ability to detect disruptive voice comms. Investments in machine learning and the addition of multi-language support should improve the automatic text evaluation, and the team is also expanding the zero-tolerance word list.

The Player Dynamics team is even putting more effort into targeting repeat offenders, chat-based offenders, feeders, AFKers, and throwers, while also exploring options that would allow them to punish offenders in real time.

Many of these solutions and features should be rolled out over the next year.