Balancing gameplay of aRPG using player behavioral patterns

What the talk is about

Andrzej Stroinski shows how A-Square Softworks balances its cooperative action RPG Gatewalkers using player behavior data. The challenge is a game with over 200 items, many spells, skill-based combat, combos and up to four players, which spreadsheets and manual testing struggle to cover. The team applies process mining: event logs are turned into a graph of behavioral patterns that reveals overpowered and underpowered skills, using their own extended version of the Fuzzy Miner algorithm. After a few iterations of parameter changes, the graph became more even. A bot replaying the discovered patterns matched real player actions roughly 60–75% of the time. He also covers separating casual and hardcore players and where the method fits other genres. Useful for designers and balancers of RPGs, MOBAs and MMOs.