But as it's happened, the single most successful online gaming venture to date, and one of the few that actually operates in the black, doesn't charge. It doesn't have to. And yet it makes money. And paradoxically, its very success poses a threat to most other online game operations -- and suggests that it's the old store-based retail sales approach, rather than the new online medium, that's winning the war for gamers' dollars.

The success is called Battle.net. A free service run by Blizzard Entertainment, the developers of StarCraft and Diablo, Battle.net allows StarCraft and Diablo purchasers to play online. StarCraft, not incidentally, was the number one bestselling PC game of 1998, according to the PC Data list -- and has sold 1.6 million copies to date.

Battle.net claims 2.3 million active users. The MSN Gaming Zone claims more -- 4 million -- but the numbers are not directly comparable; according to Battle.net, "active" users are those who have actually played a game on their service within the past 90 days. The MSN Zone's figure presumably includes every account ever created on their (free) service, including ones that have never been used since they were created.

A better measure, anyway, is the number of simultaneous users at peak periods: Battle.net boasts more than 50,000, while the Zone reaches a bit over 30,000. Battle.net is clearly, by most measures, the single most-used game service on the Internet.

Battle.net's integration with Blizzard's games is tight: If you want to play StarCraft online, for instance, you just fire up the game, click on "Multiplayer" and then "Battle.net," and there you are, in a waiting room, looking at a list of open games and chatting with other waiting players. Join a game, and when the necessary number of players have joined, you're playing. As simple as that.

But more than this is going on behind the scenes. When a game starts, Battle.net looks at the players' computers and chooses one of them to act as the "game server" -- it'll choose the one with the lowest latency (delay) and the fastest processor. And then Battle.net goes away. The players' machines talk to each other, over the Internet, for the duration of the game. If the game server loses its connection, somehow, one of the other players' machines takes over smoothly, so the game is never "lost." The players never connect back to Battle.net itself until the game is over -- at which point, Battle.net takes the results of the game and integrates them into its "leaderboard," which lists the top players and player rankings.

The clever thing here is that Battle.net's servers don't shoulder much of the burden. Battle.net provides chat rooms and player matching and rankings, and that's about it. All actual game processing occurs on the players' machines, and Battle.net doesn't handle any of that traffic. The result is that Blizzard can run the service very cheaply; at peak times, Battle.net takes about 10 servers to operate. By contrast, a game like Ultima Online, which demands continuous traffic between Origin's servers and players' machines, requires dozens of servers to satisfy peak demand.

To be sure, Battle.net's approach is nothing new: If you play Quake on TEN, say, it works the same way. In fact, this is how online play of most Net-capable CD-ROM games works. The "aggregators" -- companies like TEN and MPath and Heat.net -- were all founded as services to provide this kind of player-matching.

In other words, if you're a game publisher, and you want to offer online play, you have two options: You can make a deal with one of the aggregators, or you can, like Blizzard, implement your own service.

There are advantages to both approaches. If you make a deal with MPath, say, you don't have to pay to implement the server-side software and maintain the service. And that's not an inconsiderable concern; this is a technically demanding kind of operation.

But there are some strong advantages to doing it yourself, too. One is control over technology; when a problem arises, you can fix it yourself, rather than calling a third party, which has its own priorities and may choose to fix a problem with some other publisher's game first. Another is differentiation: All services on MPath, say, look pretty much alike, because when MPath implements a new feature, it spreads it across all the products it offers, not just yours, so there's nothing you can do to make your product stand out.

Perhaps most important, when you run your own service you have a direct connection to your customers. You see what they're doing and how they play; you hear their complaints and their praise; you learn how to serve them better; and you can reach them immediately when you've got a patch or a new game to offer. And of course, if the operation is profitable, you keep all the income yourself.

Blizzard isn't the only company to have done this; Sierra offers its games through Won.net, and Bungie through Bungie.net. But Battle.net is, far and away, the most successful such operation.

It helps if you have a game that sells 1.6 million copies as a basis, of course.

If Battle.net lets people play for free, how does it make money? The answer is simple: advertising. Players waiting for a game-start, or chatting on Battle.net, are exposed to ad impressions as they do. "We're serving millions of ad impressions every month -- 30 million last month," says Paul Sams, Blizzard's vice president of business development (last month being March 1999). Moreover, according to Sams, Blizzard supports higher rates for its ads than most other Internet sites because their click-through is substantially higher than average -- "around 4 percent in a good month, and at least 2.5 percent in a bad one."

Although Sams says Blizzard originally accepted that Battle.net would lose money, the service has operated in the black since the third quarter of 1997 (it was launched in December 1996) -- purely on the basis of advertising revenue. In other words, it has more than covered its operating expenses since then, without including a proportion of boxed game sales or any other revenue as attributable to Battle.net's existence.

Blizzard says it thinks sales of StarCraft were 10 to 15 percent higher than they would have been if the game hadn't offered free online play. Sams cautions that the number is a guess, not based on any hard research.

From where I sit, that number looks far, far too low. In the last 90 days, 2.3 million people have played on Battle.net. Seventy percent or more of them were playing StarCraft or StarCraft: Brood Wars, an expansion. That's 1.6 million separate IDs in play -- and only 1.6 million copies of StarCraft have been sold (and about a half million copies of Brood Wars, but almost every purchaser of Brood Wars will have previously bought StarCraft). Even assuming that some players have multiple IDs, and that some are playing with pirated copies of StarCraft, a very, very high proportion of StarCraft purchasers are clearly playing online. This should be no surprise; StarCraft is faster, more engaging and more challenging when played online than when played solo.

In a way, that's good news for other online game providers: 1.6 million people are being conditioned to enjoy online play. But in another way, it's bad news: Battle.net's success implies that the traditional retail channel for games sales is winning. So much for the long-awaited online gaming revolution.

Next page | If Battle.net wins, why does online gaming lose?