Monday, December 26, 2011

Study: Cheating in Online Games Spreads Like A Virus

"Cheaters are unethical individuals who can model the position of individuals in large-scale non-hierarchical communities that abuse the shared social space. In online social networks, they can model the abuse of available, legal tools, such as intensive use of communication tools for political activism. Taken to the extreme, such behavior leads to the tragedy of the commons: all players become cheaters and then abandon the game; corruption escalates and chaos takes place; and communication is buried in noise."

Last week a research team led by Jeremy Blackburn of the University of South Florida published what they claim is the biggest study on cheating in on-line games done to date.  The study looked at cheating by players using Steam.  Not only does Steam have something called the Valve Anti-Cheat System that marks users as cheaters, but has APIs that the researchers could use to crawl through the available information on which accounts were flagged as having cheated on an on-line game.  The researchers collected data for more than 12 million Steam users and found over 720,000 accounts that had been flagged for cheating in at least one on-line game.

The big "news" coming out of the study is that players who are friends with players who cheat (like macro miners and other botters in Eve Online) are more likely to become cheaters than those who do not have friends in a game who cheat.  However, from an Eve perspective, I believe that the geographic information is just as noteworthy.  The study looked at the accounts from 10 countries, (US, Brazil, Russia, Germany, France, UK, Poland, Canada, Australia, Sweden, Denmark and Norway) and compared them, looking to see if geographic location accounted for the likelihood of a player cheating.  The results may surprise some Eve players.

Source: Cheaters in Steam.pdf (p. 9)

While many claim that the Russians are the biggest botters, at least in the Steam community that is not the case.  According to the research players from the Nordic countries of Denmark, Norway and Sweden are the most likely to engage in cheating.  The research team came to this conclusion about geography:
"Cheater distribution does not follow geographical, realworld [sic] population density. The fact that some regions have higher percentages of cheaters to the player population may suggest that cheating behavior is inspired by the tighter geo-social clustering speci c [sic] to a geo-social culture. Such cheating-prone communities can be the target of more scrutiny or are the result of higher tolerance to cheating behavior, both in the legislature and in the gaming population." (p. 14)
The research done might have some consequences for both gamers and social networks in general.

"Our study has consequences for gaming in particular, but also for other online social networks with unethical members. In the case of gaming, individual servers can evaluate the cheating risk of a new player by looking at a combination of attributes inferred from the player's pro le that include the fraction of VAC-banned friends. Our preliminary investigations in this direction show that traditional machine learning algorithms (such as logistical regression, naive Bayes, and decision trees) can classify players as cheaters or noncheaters with accuracy between 65% and 74%. More work in this direction is left for the future." (p. 14)
As much as I've applauded CCP Screeg's War on Bots this past spring, I'm not ready to start instituting screening procedures like this for games yet.  I guess I'm too much like CONCORD and prefer to punish wrongdoers after they commit the violation.

No comments:

Post a Comment