Featured Blog | This community-written post highlights the best of what the game industry has to offer. Read more like it on the Game Developer Blogs or learn how to Submit Your Own Blog Post
Speaker for the Data
Game researchers seek to understand gaming in multiple contexts using an objective data set. Without access to the big data sets that most game companies, platforms, and ISPs collect we are stymied.
In the past ten years the gaming industry has exploded from 200 million active gamers to over 1.5 billion worldwide. And these gamers are playing on multiple devices, all hours of the day, in different languages, across the globe. Like wildlife biologists looking for patterns in the ebb and flow of migrating water buffalo across the African savannah, the question that we ask ourselves as game researchers is, “what can looking at a large group of gamers tell us that looking at just one doesn’t?” When a wildlife biologist ventures into the wild they are not looking to observe the lone water buffalo in order to establish patterns of behaviour, they are looking for the group and what the group can tell them about water buffalo in general. By nature gamers are physically separated by language and continents while virtually they construct armies united in providing an unlimited data set that goes nowhere except to their respective platforms and Internet providers. This is where big data comes in and researchers worldwide fall to our knees to plea for more of it.
Big data is the popular term for the exponential growth, availability and use of structured and unstructured information in the cloud. Big data is classified by three dimensional imperatives, all of which research scientists are interested in: volume, variety, and velocity.
The volume of information comes from everywhere; data from dating websites, Twitter feeds, Facebook photo uploads, GPS directions, texting, weather balloons, and even barcode data is being collected and stored. The main reason for this is because storage costs are decreasing rapidly and the storage of this data is largely automated. However, the sheer volume of this big data, petrabytes by the week, have caused other issues to emerge; mainly, the big question for big data is how to make sense of it and create relevant value from the morass.
Variety of data flowing into the cloud is diverse. From traditional text documents, transaction data, meter-collected data, video, and audio to representational data stores created by end users containing game data from multiple quests.
Big data is being produced fast, torrents of data in real-time are being stored and we need an equally fast way to organize and analyse this data. The velocity at which big data is being acquired is currently driven by an increasing need to make this data useful. The reaction time between collecting and analysing data is a challenge that smart metering and tagging cannot handle alone.
Looking at these three dimensions the real issue is not that we are collecting large amounts of data, because everyone is. What do we do with this data is the big question and what actions separate reactive organizations from proactive organizations. The hopeful vision is that we will be able to coral and harness the relevant data and use it to make decisions. This process of data mining is currently specialized and expensive in both time and people hours.
Data mining began and remains strongly consumer focused, enabling companies to determine relationships among internal factors such as price and product positioning and external factors such as customer demographics and competition. Moving from the traditional supermarket model, data mining has evolved into a link between analytical systems and transactional data. Data mining software analyses relationships and patterns in stored data based on open-ended user queries. Within the last ten years several types of analytical software have become available: statistical, machine learning, and neural networks. Using this analytical software we can seek out relationships among the data in order to turn large amounts of records into usable chunks. The type of pattern the user identifies classifies relationships in data mining. The simplest type of data mining is classes where stored data is used to locate data in predetermined groups. For example, a game company could mine customer purchase data to determine when players buy credit and what they typically do with the credit. This information could be used to increase traffic by having daily deals. Clusters are data items grouped according to logical relationships or player preferences. For example, data can be mined to identify market segments or player affinities. Associations can be mined between two seemingly unlike actions. For example, the increase in signing up for online games during the holidays can be mapped by location. Finally, sequential patterns can be mined to anticipate behaviour patterns and trends. For example, a game researcher could predict the likelihood of certain objects being bartered based on the player’s joining of a guild or changing affiliations in World of Warcraft.
Mining big data for game research consists of the same five steps that other data mining employs: extract, transform and load data on the data warehouse system, store and manage the data in a multidimensional database, provide access, analyse the data, and present the data in a useful format. Online games, ISPs, game communities, and traditional game companies are all ready doing the first three of these steps; it is the access that is the missing link. Big data has the potential to show researchers the evolution of gaming from the earliest point of data collection to the current trends in gaming. By understanding our past we can inform the future steps in games research. But we need the key to the door in order to get started.
There is a multitude of ways that big data can help game researchers make better game experiences. One key example is the new model that Georgia Tech researchers have developed that can predict players’ in-game performance and provide a corresponding challenge they can beat, leading to master of new skills. Using collaborative filtering through data-driven model algorithms this model outperforms other current techniques specifically because it is informed by the player’s own actions over time. Given that this is one example of tracking one player’s actions over time what could we learn if we were able to crowdsource millions of player actions in something like EVE Online? Using the big data that the game collects from each player game researchers would be able to profile new players based on like profiles from players in their region, age range, gender, experience level, education, and preferences to predict a course for them to be guided along as they learn the game. Another possible scenario revolves around cheating or ‘gaming the system’ in games. Specific actions by individuals are hard to trace in large games but with the use of big data patterns emerge that are much easier to see, associations and sequential patterns bear out the qualities and habits of cheaters as they progress through the system.
Like the wildlife biologist scouring the Serengeti for herds of water buffalo, a game researcher is not only interested in the lone gamer but in the hoards that allow us to examine the cultural peculiarities and creative patterns of gamers worldwide. When conducting an ecological scan of gaming we have to look from multiple perspectives of platforms, genres, and above all culture. Currently this research is done through surveys, observation, and interviews which are subjective and up to the individual gamer to honestly participate. The use of objective big data from game companies and ISPs offer a larger and yet more nuanced view of gaming culture. We just can’t get at it.
Currently there is only one large data set freely available and that is from the World of Warcraft created originally by Blizzard. World of Warcraft offers one large data set on avatars from a three year period, 2006- 2009. However, pretty much all game companies track everything about their players. They use the data that is out there to track how people play the game, what they spend money on, and how their history will inform their future play. With the information that game companies regularly track game researchers could begin to answer questions about topics like non-player character AI, player interest modelling, and social network analysis. This is not to say that there are not partnerships out there between research institutions and game companies but they are the chosen ones with very little information available to the masses of game researchers eager to employ new models to this big data. Like the early computer gurus working out of their garages and the numerous start-ups that get their momentum in game collectives, game researchers may be working individually or in small groups where the only assets they have access to are their brains and laptops.
Freely available big data from the major game companies would change the way we think of gaming by providing a larger pool in which to gather, sort, and analyse patterns. Are you interested in the smurf mindset? Do you want to track the progress of grinders? What sorts of patterns do those who gank exhibit before and after the action? What are the social networks of female Australian players who play Petalz on every other Wednesday? What are the characteristics of a tank-and-spank in international groups versus national groups? At this point the game researcher can’t answer any of these questions in a large scale quantitative way. We need the data and we need it to be big.
So this is what it comes down to, a plea for big data, freely available and stripped of personal data. Game researchers do not care what the player’s name is nor do we want their credit card numbers or addresses, or phone numbers. We do want to be able to look at the groups of players and their actions over a long period of time both between and within platforms and game communities. Please?
Read more about:
Featured BlogsAbout the Author
You May Also Like