May 28, 2020 Reading Time: 10 minutes

On September 13th, 2005, Blizzard Entertainment released Patch 1.7 for its still-young MMO game World of Warcraft. Titled “Rise of the Blood God,” it featured a series of new challenges and missions for players. And chief among those was a temple called Zul’Gurub containing a particularly nasty boss meant to challenge higher level players: the serpent-like demon god Hakkar the Soulflayer.  

A description of the game lore is unnecessary (there’s plenty of information available if you’re interested), but for one detail: Hakkar was equipped with a particular nasty attack known as the Corrupted Blood spell. But whereas many attacks in games simply deal damage to players, Corrupted Blood had two fairly new tweaks. In addition to doing initial damage to the player, the spell was a debuff: it would continue to deal damage once every few seconds, slowly wearing down a player’s health. It was also designed to be “contagious,” spreading from attacker to attacker, consequently forcing players to spread out during their assault on Zul’Gurub. An “infected” player could spread it to another in close proximity, that player could spread it to another, and so on. 

Because the Corrupted Blood incident occurred shortly after the initial introduction of Patch 1.7, it’s likely that one of the very first 20-player parties attempting to take on Hakkar in Zul’Gurub was responsible for the event which followed.   

A Consequential Oversight

During the fight, some players must have noticed that the Corrupted Blood spell also infected pets (animal minions which accompany some characters into battle). And so to simplify the fight and reduce the distraction (and energy “cost”) of healing both pets and characters, they dismissed them. (Dismissing temporarily frees a pet, which can be called back later.) 

As players emerged victorious from the Hakkar encounter — as is frequently the case in many fantasy MMOs — their first destination would typically be the game’s major urban centers: specifically, markets. There, they would sell some of the loot they’d won, repair damaged weapons and gear, and engage in social activities. Unsurprisingly, markets are at any given time among the most densely populated areas of World of Warcraft (and, in fact, most MMOs). And having dismissed their pets during the Hakkar battle, upon their return to cities and marketplaces many players called them back.

One can be reasonably certain that the development teams at Blizzard Entertainment were familiar with the concept of contagion and intended for the Corrupted Blood spell and its contagion to be limited to the instance within which the Hakkar the Soulflayer mission took place. But unbeknownst to players and developers alike, while the Corrupted Blood disease was removed when players left Zul’Gurub…it stayed active on players’ dismissed pets.

The Initial Outbreak

As pets were summoned back to masters in the densely-crowded virtual marketplaces, players immediately began contracting Corrupted Blood. And because the difficulty of the Hakkar encounter was reflective of the skill and strength of the high-level characters that it was intended to attract, Corrupted Blood was a particularly dangerous spell: a serious annoyance for high-level characters, but devastating and often instantly lethal to lower-level players. Some players with enough health to not be instantly killed ran out of the marketplaces into other areas of the city; others — some infected, some not — quickly teleported elsewhere. Corrupted Blood was soon racing through numerous World of Warcraft cities and towns all across the virtual world, spreading and killing scores of characters representing tens or even hundreds of thousands of human players. Meanwhile, players in areas far-flung from marketplaces or cities teleported closer to observe the growing panic, increasing both the virtual body count and in some cases furthering its spread. 


Some players initially thought that what they saw occurring was a prank or an unannounced Blizzard marketing event. But for others, the rapidity and indiscriminate spread of the Corrupted Blood led to the conclusion that they were witnessing the unfolding of an unintended disaster.

As information about the Corrupted Blood outbreak spread, some took action. Alone or with other players — some in groups already bound together by guilds, others spontaneously organized — some high-level players shepherded low-level, more vulnerable players out of cities to avoid the rapidly expanding plague. Players without healing abilities “yelled” updates: some warning healthy players nearby to stay out of World of Warcraft cities, others giving locations of safe (self-isolation) zones.

(The choice of the “yell” command is particularly interesting, as among the three forms of communication, the “say” command has a low, local range, and the “general chat” is as wide as the entire zone. Using “say” would have been of no use other than to players in the very immediate vicinity, while communicating via general chat would have drawn players from every corner of the zone, immediately overwhelming the safe zone. But “yell,” with an in-game range of roughly 400 meters, allowed players to inform others of both danger and safe havens nearby without bringing a frenzied rush from all corners of the map.)

Some of the most powerful, high-level players with healing spells (priests and shamans) bravely attempted to resurrect or heal dead or dying lower-level players for long enough that they could either escape from the cities or arrive in what amounted to virtual first aid areas. While many players were healed of Corrupted Blood long enough to reach the comparably safer digital wilderness, other resurrected players quickly became re-infected and died again, sometimes spreading the disease a bit further in the process. And even some of the most powerful healers ultimately succumbed to Corrupted Blood, surrounded by the afflicted while attempting to administer aid.

The player-coordinated response to the pandemic, hours (and perhaps longer) before World of Warcraft’s game managers got involved — perhaps before they even understood the dimensions of the virtual disaster — was voluntary, emergent, and with small groups of survivors gathering in the distant wilds of World of Warcraft, arguably effective.


For four days (Sept 13th – Sept 16th, 2005) the plague spread. And at some point news of the curiosity — which certain World of Warcraft players were still certain was an intentional, global event — spread to the real world. Consequently, casual players logged in to watch the disease spreading “first hand;” some of them quickly getting infected, teleporting to other areas, and passing the Corrupted Blood to “healthy” players before expiring.

Survivors began communicating with friends, allies, and sometimes strangers in other regions of the game: relaying the locations of safe zones and attempting to piece together the source of the Corrupted Blood scourge. As they pieced together the pet-borne transmission element, a previously unknown piece of the puzzle fell into place: NPCs, non-player characters (thousands of whom “staff” markets and other important locations within the game) were spreading the disease as well, but did not bear any of the outward signs (a “splashing” effect). Thus NPCs were a particularly dangerous, if completely unintended, facet of the Corrupted Blood epidemic: asymptomatic carriers. And because NPCs cannot be killed, they stood as persistent carriers of Corrupted Blood. Having determined that, further announcements were made via general chat, travelling game-wide: despite their lack of outward appearance, players were warned, assume that all NPCs are infected with Corrupted Blood. 

And as always with a highly social game (and no fear of ultimate mortality), a number of griefers took advantage of the fear and confusion: purposely infecting themselves with Corrupted Blood and searching for pockets of uninfected players. 


Major cities were all but abandoned, with some groups of players hiding out in dungeons (ironically, now some of the safest places in the game) or migrating toward less populated areas. By one account, the initial plague broke out in Ironforge; some survivors fled to Lakeshire, and others to Crossroads; then to the pirate haven of Booty Bay or Gadgetzan; and finally (again, ironically) many individuals and groups hunkered down in the unsettled, desolate region known as the Plaguelands.

At roughly 7:30am EST on September 16, 2005, a player posted in a forum reporting a disease spreading through Orgrimmar; a video appeared several hours later. At roughly 1:15pm EST, a Blizzard staffer reported that the firm was “aware of the issue and working on it.” 

Because of the ability of players to resurrect and be resurrected, and thus the possibility for the disease to spread without end — as opposed to real world outcomes including herd immunity, vaccine development, or darker (and unlikely) eschatological outcomes — the only solution was for game managers to reboot the servers: removing the Corrupted Blood infections all across the game world and modifying the Corrupted Blood spell code; the latter to no longer affect pets, and ensure in any case that the virtual disease could not leave the Hakkar encounter area.

Ten hours later (at about 11:30pm EST on September 16th, 2005), rolling restarts of the servers were announced, beginning at 7am EST on September 17th. The initial hotfix, applied during the rolling restarts, was unsuccessful, but after a second attempt the plague was stopped. The generally accepted time at which the Corrupted Blood plague ended was at just before 4pm EST on September 17th, 2005 after half a week of tumult and hundreds of thousands, perhaps millions, of character “deaths.”

In the months and years after the incident, World of Warcraft’s Corrupted Blood epidemic became a cause célèbre among certain scientists, including a handful of sociologists and epidemiologists. To be sure, World of Warcraft’s Corrupted Blood episode makes a fascinating story. But to borrow from a memorable line, insofar as using it as a sociological, epidemiological, phenomenological, or in any sense as a model of human action under severe conditions: I come to bury the example of The Corrupted Blood, not to praise it. 

The Modeling Obsession

There is a view that MMOs and other complex, large scale games may function as simulations of human behavior under extreme conditions, given that they involve the independent decisions of millions of players in a persistent space and thus effectively constitute a massive, computational petri dish. 

To be sure, social science factors are missing from virtually all of the prevalent epidemiological models in use: regressions which, fed by historical data, current statistics, and certain environmental assumptions, trace a curve over time. So too for agent-based models — models which populate a landscape with agents (each essentially an “individual” data structure with embedded characteristics and functions) and track their interactions over time — to predict infection, fatality, and other characteristics of spread of disease. Epidemiological models which employ compartmental and network techniques are similarly bereft of social science elements.

But examining the actions of game players and game designers or managers in the hope of gleaning anything more than expressly precise, superficial insights where behavior or policy making are concerned is self-deluding. In particular, casting game designers, developers, virtual community managers, and other decision makers of gaming firms as de facto governments, regulators, or central bankers is a woefully corrupt representation of the relationship between game players (consumers) and gaming firms (producers of for-profit ventures). As I have written previously,

[w]hile their actions may ultimately generate similar outcomes, central planners seek and wield power whereas the actions of commercial gaming interests are undertaken to compete with other online entertainment providers by delicately balancing opportunities for newer players with the need to continually challenge experienced players.

Actions undertaken by players in massive, multiplayer games may appear identical to those undertaken by living, breathing people, but draw from wholly different functions and motivations: game mechanics are at best a product of constrained choices in unique, virtual domains. Attempting to draw lasting, non-trivial conclusions to real-world problems from even the most complex games, wherein players face an entirely different set of opportunities, incentives, and epistemic conditions, is a fool’s errand; although one for a sophisticated fool, to be sure. 

Indeed: a paper airplane might briefly soar across a room the way a 747 does at 35,000 feet; but who would want to travel in the latter, if designed by the principles attending the former?

Corrupted Blood in the age of COVID-19

And predictably, since the start of the coronavirus crisis the assertion has been made that the 2005 Corrupted Blood incident could, or should, be considered a guide for addressing the spread of COVID-19. How? 

Few laymen, let alone physicians, sociologists, economists, or epidemiologists, would be surprised that contagious diseases spread. Or that facing such a calamity, individuals may act heroically, idiotically, or even maliciously. Or that mistakes will be made. 

Would flesh-and-blood human beings be willing to live by policies derived from epidemiological studies on fantasy games? For example: one could argue that the strongest, highest-level healing characters did more harm than good by healing low-level infected players of the Corrupted Blood ailment who, in all likelihood, were going to die anyway; indeed, some of the weaker, rescued players lived long enough only to infect additional players before expiring anyway. Is the “lesson” that in a pandemic some citizens are beyond help and medical resources shouldn’t be expended upon them? 

Or is the lesson that players in the game were somehow inadequately prepared for the disease outbreak, and thus that real-world governments should divert massive resources toward onshoring massive stockpiles of medicine, equipment and PPE in the face of unknown (and likely unknowable) future disease outbreaks? (AIER has written at length about the short-sighted, autarkic response to the spread of the novel coronavirus known as “medical-supply nationalism.”)

Consider the following comment from one academic paper. In Fall 2019 — with the initial wave of COVID-19 infections only a month or two away — the Berkeley Scientific Journal reported that

[d]uring the initial phases of the [Corrupted Blood] epidemic, players, and later Blizzard Entertainment itself, established quarantines for characters to wait out the infection. Many players simply refused to obey the quarantine.

Far more interesting than the idea of using a fantasy game as a model of the spread of a disease in the real world is a comment which seems to suggest a degree of incredulity at the notion that individuals, in any sense (and in particular, in a virtual, essentially riskless, entertainment venue) might not choose to stay put when and where ordered to.

People everywhere can be thankful, or perhaps hopeful, that this comment (from the same article) bears out and remains true:

There have been many attempts to create a virtual world dedicated to social science and epidemiology research, with mixed success. Most manufactured worlds cannot hope to match the success of commercial MMOs, and most commercial MMOs do not have the time or motivation to work with epidemiologists without incentive. 

If We Must Draw Lessons

The concept of using MMOs, incidentally or purposely, to plan epidemic (or any other disaster) policies confuses discoveries about gameplay with those regarding the actual, boundless choices, actions, and motivations of corporeal human beings under conditions of uncertainty and panic. 

No, the Corrupted Blood event did not “perfectly predict” the outbreak and spread of the novel coronavirus. No, the Corrupted Blood episode did not “teach us” how humans behave in outbreaks. And no, video games will certainly not “rid the world” of COVID-19 or any other virus. On the other hand, the Corrupted Blood outbreak in World of Warcraft was an excellent model, and provided highly relevant insights, as to how in Azeroth in 2005 fantasy game players would behave if a virulent, virtual plague suddenly appeared.

If we are bound to draw lessons — which may be another fool’s errand — there are two that come to mind: one specifically for game developers, and one more broadly. Both come from Hayek.  

Drawing from Hayek’s adage that “the curious task of economics is to demonstrate to men how little they really know about what they imagine they can design,” modelers of real-world phenomena are well-reminded that in any complex system comprising numerous layers and connections, small changes in inputs lead to massive, often incomprehensible changes in output. Thus with the addition of another hundred, thousand, or million lines of new code; among hundreds, thousands, and millions of new objects, systems, and rules of virtual interaction; on top of the addition of another hundred, thousand, or million new players, the complexity of a social activity – and therefore of errors or unforeseen outcomes – increases, and in so doing becomes increasingly difficult to predict. Which is to say: the real world, where individual human beings act, react and interact, is impossible to simulate. 

To the second point, there is a conversation — again, in the spirit of Hayek — to be had regarding the way in which World of Warcraft players, shortly after the Corrupted Blood epidemic began, gathered and utilized relevant, actionable local information, employing it to guide arguably effective small-scale responses. Responses which, by nearly every account, facilitated responses long before Blizzard employees knew something unexpected was happening within the game. 

But that conversation, and an apprehension of the factors at work there, is beyond models, computers, and social science supercolliders; it requires recognizing the existence and power of a social structure more complex than the human mind

It is a conversation regarding the unsimulatable, a good start to which begins here

Peter C. Earle

Peter C. Earle

Peter C. Earle, Ph.D, is a Senior Research Fellow who joined AIER in 2018. He holds a Ph.D in Economics from l’Universite d’Angers, an MA in Applied Economics from American University, an MBA (Finance), and a BS in Engineering from the United States Military Academy at West Point.

Prior to joining AIER, Dr. Earle spent over 20 years as a trader and analyst at a number of securities firms and hedge funds in the New York metropolitan area as well as engaging in extensive consulting within the cryptocurrency and gaming sectors. His research focuses on financial markets, monetary policy, macroeconomic forecasting, and problems in economic measurement. He has been quoted by the Wall Street Journal, the Financial Times, Barron’s, Bloomberg, Reuters, CNBC, Grant’s Interest Rate Observer, NPR, and in numerous other media outlets and publications.

Get notified of new articles from Peter C. Earle and AIER.