Massive Thought

Massive Thought: Researching Bot Behavior

The word bot is something almost all MMO players have encountered, some might only know it by name, and some may encounter the actual bot daily while playing. A bot is a program controlling a character in the game, grinding for experience, money, resources and much more, it is a program playing instead of a player. This is seen as a problem by big parts of the gaming community, where players hate how they mess up the game, developers on how they scare players away and messes up their cash flow. But what solutions are there to finding bots? And is botting really something bad? Or is it just another way of playing the game?


To come to term with all the bots in MMOs, the most common solution is to have players report them and then have game administrators investigate and shut them down. This is of course very time consuming both for the players and the people working at the game. So Kuan-Ta Chen with others analyzes the traffic generated by both human players and by game bots, to try and find a solution on how to automatically identify bots. They do find trends in in traffic burstiness where the bots are moving in a min loop which can be identified, they also find different reaction times to network conditions, where a player adapts to network delay while a bot continues on like nothing has happened.

They suggest using these traffic identifiers to enable a program to search out the bots by sending out 10000 packets over the connection and therefore distinguishing player from bots, generating a 90% correct decision rate. Their study is done on Ragnarok Online but is generalizable to other games as well. While their result is very interesting and something which scholars in the field should focus their attention on, if bots are seen as a problem in need of urgent attention. The method they use to identify bots is going to be at 100% certainty before any gaming company even considers it, seeing how banning a player believed to be a bot, could be seen as a much larger problem than the actual bot.


Botting is done for various reasons, one of them being for accumulating gold which is then used for real-money trading (RMT). Even if this doesn’t necessarily mean that bots are being used, it could still be seen as a problem in need of a better solution. Finding RMT is mostly done in the same way as discovering botting, with the players help, but Hiroshi Itsuki with others examine a better and more efficient way of doing so. To do this they use the player log data, which is often used in research purposes. In the data they use the player statistics on: Total number of action records, the amount of minutes in which at least one action is taken, the total number of utterances in chat recorded, the total amount of virtual currency handled in a period. By using this data, they can extract possible RMT suspects and focus on investigating them and\or shutting them down. While this solution is both effective and takes in consideration of the players in a sense, it still leaves the question open on whether RMT is a problem or only a natural act for the players.


Stefan Mitterhofer with others take a clear stand that botting is wrong, since it has a bad impact on the “honest” players to continue playing the game, and how it threatens the subscription-based business model. While this may be true, it is still far from being wrong by using these arguments. It could also be that botting is a way for some people to play the game on their terms, or a way for none-hardcore players to be on par with hardcore players. If one buys the game, shouldn’t they be able to play the game on their terms unless it hurts someone in real life?

Putting this discussion aside for a moment, in Mitterhofers research they looked how bots can be detected via an automatic system on the server-side. The way of doing this is by analyzing the movement of players and find movement patters which repeat frequently, which in turn would indicate that a character is controlled by a script and not a human player. This is actually something which is already in use to some degree at some companies, but there is still no statistics on how useful it is or how efficient it is. Using similar methods, Marlieke van Kesteren with others also examine player movement patterns, but instead of using checkpoint in which the character moves, they focus on how bots and players move around in different angles with a classification rate of 100%, something which the game companies might want to focus their attention on further.


Another way of approaching an automatic bot detection service is by looking at the frequency of player actions. Ruck Tawonmas with other do this in their research, where they look at the publisher’s database rollback. Their bot-detection strategy is done in two stages, in the first stage a character will be defined as a bot if the frequency of its actives is much higher than those of known humans. In the second stage the rest of the characters are defined based on their action types. This strategy was used in Cabal Online and did confirm its effectiveness.


As you have read above, there are multitudes of ways to distinguish bots from human players, and most of the scholar community sees botting as problem, a disease in need of a remedy, where everyone has their own remedy. Still, these remedies are still not in use in any regular frequency by the companies, at least not that we know of. That they in periods come and take care of a lot of botters I well know, but if this “problem” was as serious as they and the scholars try to make it, this would be dealt with daily and botting would not really be possible. So it can be that botting is helping driving the economy of the games, help driving in players and not kicking them out as claimed. Botting itself could be what keeps a lot of MMORPGs alive today, so are they really so bad?


Research used in the article:

Chen, Kuan-Ta., Jiang, Jhih-Wei., Huang, Polly., Chu, Hao-Hua., Lei, Chin-Laung and Chen, Wen-Chin. (2009) “Identifying MMORPG Bots: A Traffic Analysis Approach”

Itsuki, Hiroshi., Takeuchi, Asuka., Fujita, Atsushi and Matsubara, Hitoshi. (2009) “Exploiting MMORPG Log Data toward Efficient RMT Player Detection”

Mitterhofer, Stefan., Platzer, Christian., Kruegel, Cristopher and Kirda, Engin. (2009) “Server-side Bot Detection in Massive Multiplayer Online Games”

Thawonmas, Ruck., Kashifuji, Yoshitaka and Chen, Kuan-Ta. (2008) “Detection of MMORPG Bots Based on Behavior Analysis”

van Kesteren, Marlieke., Langevoort, Jurriaan and Grootjen, Franc. (2009) “A step in the right direction: Botdetection in MMORPGs Using movement analysis”

Related: , , ,