The video game industry is notoriously tight-lipped about its numbers, filled with members-only metrics like wishlists, pre-orders and player numbers, so it can be hard to know which games have actually proven popular and turned a profit.
Industry associations such as France's SELL periodically release sales figures based on data taken from game publishers.
Private research firms also make their own estimates, based for example on publisher or public data as well as their own user panels and rankings.
But publishers and studios often highlight indicators that are most favourable to their messaging, which vary depending on the type of game, experts told AFP ahead of the Gamescom trade show now underway in Cologne, Germany.
Here is a look at some of the proffered gauges, and how to read them.
Online stores operated by console makers as well as gaming platforms like Steam or Epic open "wishlists" for players to register their interest before a game is available to play.
Some publishers tout wishlist numbers when they pass an impressive threshold, usually at least a million.
But Rhys Elliott of Alinea Analytics, a consulting firm, told AFP the indicator usually had remarkably little to do with actual sales, particularly for big-budget releases known as AAA in industry parlance.
"Never, ever, ever kind of look at wishlists as a reflection of the sales," Elliott said.
"For some very lucky indie games or something like that with lower price points, maybe they can convert 40 percent of their wishlist at launch," he said.
"But for an AAA game, usually it's under 10 percent or even under five percent."
Pre-orders are a more reliable indicator, Elliot said, but he warned against simply converting sales volume into a definitive amount of revenue.
"Regional pricing and discounts mean that huge unit count can actually be a small amount of revenue or the other way around," he said. "A game might cost $70 in the US and then cost the equivalent of $30 in China."
And a million copies sold for $10 each will of course not generate the same revenue as a game sold for $70.
"Drilling down a bit deeper into, like, where the units, the players, the copies sold come from is super important," Elliot said.
A publisher reporting a game's number of players rather than its sales is often seeking to disguise a game's true performance, Elliot said, particularly if it comes bundled with a subscription service -- increasing the number of players without any necessary effect on sales.
"I'm not going to say 'sneaky', but you're obfuscating how well your game is doing when you do that," Elliott said.
Many users on social media also track the "concurrent users" figure provided by Steam, one of the few statistics available online.
"Concurrent is very helpful," said Joost van Dreunen, a professor at New York University's Stern School of Business and a specialist in video game economics.
"That's where the audience attention is right now. But it doesn't say anything directionally about where they came from or where they're going next," van Dreunen said.
Also, Steam's figures cover only users on its own platform, not those on other providers or consoles.
Playing time is the key metric to watch for free-to-play titles such as "Fortnite" and "Roblox", games that make add-ons available for purchase, von Dreunen said, with developers battling for players' attention.
"If it's free-to-play, you look for retention," he said.
Fan groups, YouTube previews and social media chatter are also increasingly important for knowing how much of a splash a game has made, or might make.
"In the last year or two, my attention has really gone towards, what's the organic groundswell?" von Dreunen said.
"That tends to be a strong indicator for success once the game gets to the store."
The number of views on YouTube and Twitch related to a title after a few weeks, and streamers sticking with particular games for a while, can also be among the indicators of success.
For Elliott at Alinea Analytics, however, such data can give only a general picture.
"It's better than nothing," he said, but "there's a lot of noise in that data".
kf/jlo/vbw/fz/js/rmb
