The blog entry, How the digital era has changed video game launches, summarizes the white paper.
The white paper itself, Understanding the Modern Gamer, has the details and is based on an analysis of "…video game searches…for the top 20 selling games of 2010 and 2011."
Quoting various bits:
Game searches included title keywords, such as ‘Elder Scrolls V’, along with all relevant game title keyword variations, such as ‘Gears of War 3 trailer’, ‘Battlefield 3 website’, ‘Batman Arkham City review’, and more.
Search data analysis is representative of gamer behavior for 2 reasons: (1) millions of gamers use search engines, providing us with an unparalleled data set and, (2) gamers are incredibly savvy Internet users whose searches reveal an extraordinarily high level of intention.
Desktop engagement patterns closely align with game unit sales trends (.92 correlation)…
In 2011, desktop searches per gamer grew 20% y/y. Additionally, desktop search volume for marquee titles increased 29% y/y, outpacing the 8% growth y/y in console gamers online.
Our data reveals a .92 correlation (on a scale of -1 to 1,1 signifies perfect correlation) between clicks during the 10 month game launch cycle and game units sold during the first 4 months post-release…
More importantly, our data demonstrates that 84% of sales can be predicted by all clicks during the 10 month launch cycle. We used the regression coefficient from our analysis to create a predictive model and found that if a game accrues 250,000 clicks in the 10 months around launch, it will likely sell between 2 and 4 million units in the first four months after release.
Although user click data is a powerful predictor of game unit sales, we readily acknowledge that other factors – such as game quality, TV investment, online display investment, social buzz, and more – must be incorporated into our analysis to create a predictive model that is even more accurate and reliable.
There is a quantifiable link between what people search for and what they buy that enables us to predict game sales.
Note that the study may involve console games only.
At the time when a person conducts a search, that game has the person's attention. This is more true than for Xfire or Raptr which merely measure when a program is running, not what the owner of the computer is doing.
Xfire and Raptr are not measuring attention. As Wilhelm Arcturus reports, they will report a game in use when a computer is left unattended with a game launcher open. Even if the game itself is running, the player may be on a bio break, macroing, or using some other AFK-based exploit where the game is running without the player present.
Google Trends statistics reflect people physically present at their computers actively typing the name of the game into the search box. That's a significant level of engagement with a game, even if they aren't actually playing at the time of the search.
Google Trends does have some issues that may result in it undercounting mature games. As a game matures, everyone becomes familiar with it. Once it's old news the number of people searching for it by name to get general information will fall.
Well after release games may retain high numbers of engaged players conducting searches, but these searches focus on details of the game. Such searches include highly targeted game-specific terms but often not the name of the game as discoverable by us using Google Trends, although Google itself could tie such searches back to games using its data. (See also Tobold's commenter Oscar's comment about "functional" searches.)
Discovery of game-world information for mature games may also move off Google to game-specific database sites, build configurators, or forums, depending on the extent of the game community's ecosystem.
Despite these shortcomings, I think Google Trends is much better suited than Xfire or Raptr to measuring the popularity of a game, due to Google's huge demographic advantage.