Deadlock Dev Uses ChatGPT to Help With Matchmaking Code

Valve's upcoming MOBA-hero shooter, Deadlock, recently overhauled its matchmaking system, thanks to a surprising source: the AI chatbot ChatGPT. A Valve engineer, Fletcher Dunn, revealed on Twitter (now X) that ChatGPT helped him find the perfect algorithm.

ChatGPT's Role in Deadlock's Matchmaking Overhaul

Deadlock's previous matchmaking, based on a Matchmaking Rating (MMR) system, faced significant criticism from players. Reddit threads were filled with complaints about unevenly skilled teams and frustrating matches. Players reported consistently facing highly skilled opponents while their teammates lacked comparable experience. This prompted the Deadlock team to announce a complete rewrite of their matchmaking system.

Dunn's Twitter posts detail how ChatGPT, using a prompt specifying a one-sided preference scenario, suggested the Hungarian algorithm. This algorithm proved ideal for Deadlock's needs. Dunn has become a vocal advocate for ChatGPT, highlighting its usefulness in his workflow. He even keeps a dedicated browser tab open for it.

However, Dunn also acknowledges the complexities of relying on AI. He notes that using ChatGPT sometimes replaces human interaction, prompting debate among users regarding AI's role in replacing programmers.

The algorithm's function in a gaming context is to optimize team composition based on player preferences. In Deadlock's case, it focuses on one side's preferences (e.g., a player's hero selection) to find the most suitable teammates and opponents.

Despite the improvements, some players remain unhappy with the new matchmaking system, expressing their frustration on social media. Negative feedback continues, highlighting the ongoing challenge of balancing player satisfaction in online competitive games.

Here at Game8, we remain optimistic about Deadlock's potential. For a more in-depth look at our playtest experience and overall impressions, check out the link below!