EverQuest · PC · 1999 · Impact: Competitive
A fundamental AI and aggro mechanic limitation in EverQuest allowed players to deliberately "train" large groups of enemies through populated dungeon areas, wiping out other players' groups — producing a years-long social and competitive crisis that shaped MMO design philosophy for a decade.
EverQuest's enemy AI used a simple aggro-follow system: if a monster was attacking a player who then moved away, the monster would follow until either the player died, the monster was killed, or the monster leashed back to its spawn point. EverQuest had no real leash — the mechanic by which enemies give up pursuit after a certain distance from their origin — so in crowded dungeons like Lower Guk, the Plane of Hate and Sebilis, a player chased by a large group of enemies could run through areas populated by other players' groups without the enemies returning home, effectively "training" the enemy group into the populated area where it would begin attacking everyone present. Training was initially accidental — a player overwhelmed in combat would flee and inadvertently wipe nearby groups. But the mechanics made intentional training trivially easy to execute: a player who pulled enemies toward a rival guild's raid camp and then zoned out could destroy hours of progress at little risk to themselves. The social consequences were severe: server-wide reputations developed for notorious trainers, community blacklists were maintained, and with no game mechanic to prevent it, enforcement fell to social sanction and to game masters, who treated deliberate training as harassment when they caught it. Training remained viable, and occasionally exploited, for years. The EverQuest training crisis became required reading in early MMO design discussions, and later MMOs made leashing a standard feature.
EverQuest's aggro model was designed for individual encounter management: a monster that is attacked will chase the player until one of them dies, the monster loses track of the player, or the player leaves the zone. The system worked adequately in open outdoor zones where the terrain was forgiving. In dungeon environments — particularly multi-floor complex dungeons like Lower Guk — the system's assumptions failed. Dungeon corridors created movement paths that allowed a running player to thread through multiple occupied rooms, and the lack of leashing meant enemies from one room would follow into rooms they should not have reached.
The specific aggro priority system compounded the problem: when a running player passed through another group's combat, the enemies in pursuit would sometimes switch aggro to nearby players based on proximity or attack priority, fragmenting the pursuing group and distributing its aggro among innocent bystanders. A single train could cascade into a zone-wide wipe if the dungeon was densely populated and the leashing failures were severe.
EverQuest's training problem produced one of the most sophisticated player-built governance systems in early MMO history. Servers maintained community databases of known trainers — players with established histories of intentional griefing — and organised guilds enforced informal blacklists by refusing to group with or sell items to flagged players. Some servers developed formalised "train rules" that were posted in community spaces and treated as binding social contracts for dungeon use, with violations subject to coordinated social exclusion.
This player-driven governance demonstrated both the sophistication of MMORPG social systems and their limits: social sanction could contain the problem but not eliminate it, because new players continuously joined servers without knowledge of blacklists, and determined griefers could create new characters or transfer servers. The lesson the industry took was that technical solutions were necessary where social solutions were insufficient — a principle that shaped MMO design philosophy from World of Warcraft onward and that continues to inform how modern online games think about griefing mechanics.