In this blog, I will use World of Warcraft (WOW) which I am familiar with to share my ideas of the potential application of data mining in this area.
I think WOW is the most successful MMORPG in recent years. It went online in different regions around 2004-2006. I began to play it from 2005, now due to my busy life, I have stopped playing. But I am still very interested in it.
In 2009, Blizzard crafted 7,650 quests, 70,000 spells, 40,000 NPCs, 1.5 million assets, and 5.5 million lines of code; some 4,000 employees, 13,250 server blades, and 75,000 CPU cores keep MMORPG running. This number would increase these days. The picture below is screenshots of character summary page, we can see, each character has lot of variables to describe it, and each variable has lots of information. Hence, there is a lot of data for mining to improve this game.
I think mining could be performed in aspects below.
- Tracking Fraud. Usually, user want to cheat or fraud have a lot of characters in level, and the character usually in the "main-cities" of the world in WOW. By filter these features, character with high intention of cheating could be known, and track by Blizzard.
- Optimizing game specifications. This is a big problem, in my opinion which is not including debugging. In the game, characters belongs to each class and race, which make performances different. For example, the dps of warrior is preset, and shaman's is preset too. Blizzard want to make the game "balanced", which mean the game result is determine by players, not the preset specifications. By data mining, they could find result of fighting between this two characters, or using association rules to find which warrior win the fight or kill someone, which class usually paired with warriors. Then, the specifications of that class should be modified.
- Marketing miming. There is a virtual market is the game. Association rules could find which items are always bought together. Then, they may find which specification or which kind of items are most wanted. The new dungeons could offers new items.
These are my ideas in MMORPG data mining. I did search online, their is not enough resource talk about this topic. I guess the reason is no public data is available.
ref:
I read this article and immediately thought of a similar application to another blizzard game, Diablo 3. I small project team is working on an application that mines data in real time while playing. This application records items that you pick up and traits of these items. These traits consist of rarity, area picked up, time picked up, and amount picked up. Now with this data mining tool. a player can effectively play the game optimally as long as he plays enough to receive a relatively large comprehensive data set. I thought that this was a clever use of data mining and look forward to seeing further applications.
ReplyDeleteThis is a video of a developer speaking about it: http://www.youtube.com/watch?v=KUkUilkSN7c
ReplyDeleteA few years ago, a close friend and I played this obscure Korean free, golf MMORPG, called Shot-Online. I know, haha. Whatever, it was addicting.
ReplyDeleteAnyways, though at the time I had no grasp on big data concepts, we did realize that the company was utilizing their data to provide the players a better service.
The game was free, but made their money through an "Item Shop" within the game, where players, if they choose, could buy the game's credits with real money, and use this to buy items. Of course, players could avoid it all together and simply play for free (See: me).
As the game evolved in regards of adding more courses and items, we realized that everything seemed to have stemmed from current popular courses/items. (We had put a considerable amount of time into this game.)
Now with my knowledge of big data and data mining, everything is clear as day. They were taking all this data, from what type of courses people tended to like the most, to items used in certain situations and most frequently. Everything simply got better and better.
An interesting idea would be to gather overall gaming interest data and make a game based on this data, much like how House of Cards was created for Netflix.