Aeon of Strife & co.

There is a very old variation on real-time strategy games. Teams of players controlling only one unit each are allied to computer-controlled armies continually fighting on a battlefield. The players must push towards the enemy AI's base until they destroy the main building, be it castle, dark tower of sorcery, magic crystal or chess king. As far as i can tell, this idea was around since Command and Conquer: Red Alert, with its 'sole survivor' modification. It became popular in Starcraft with a custom map called Aeon of Strife, which i never actually played.
It was also one of the two perennial favorites of Warcraft 3 players, along with tower defense. Dozens of maps appeared, most not making it past their second version. The most popular was Defense of the Ancients (DotA) whose popularity in fact served as sufficient proof of concept to justify founding companies to create marketable ripoffs.

Overall, the best version of this concept was Demigod, which unfortunately died due to player frustration due to lack of testing, servers and players due to Stardock's unfortunate habit of supporting good developers by not actually providing the advertising and other support they need.
The best proof of concept was a Warcraft 3 map called Eve of the Apocalypse. I still miss the old custom game list name EotA>DotA.
DotA itself was actually one of the worst. It owed its popularity partly to being one of the worst, partly to its playability. While other maps crashed, lagged or simply frustrated players out of the game with their glitches, DotA was always quickly patched to keep every version playable. Kudos. It was also dumbed down to only a shadow of what the game could be. Lack of kudos.

The odd part of marketing this concept is its relation to normal RTS games. Because the player only controls one unit, the first impression is that this is necessarily a simplification. In reality, it only removes the worst aspect of real-time strategy: the lack of strategy. Most RTS games are clickfests, rewarding micromanagement over any sort of actual thought. Players are even stupid enough to rank each other based on their 'APM' (actions per minute), the number of commands recorded by the game client from each particular player.

Having just one unit to control fixes this mostly. There are still micromanagement problems. The most common is giving players items they need to activate on the fly in combat in addition to their character's skills. Demigod failed in keeping its summoning system simple by giving players the option to micromanage, if not the hordes of AI-player creatures, at least the half-dozen minions summonable by 'general' player characters. Still, it can be a vast improvement. Demigod's best feature, just as EotA's, was the focus on improving the team's AI. Instead of spending money on items alone, players had to invest in improving the AI's capabilities with new units, siege weapons, bonus damage, all the usual upgrades. This put the strategy back into things. A team could completely dominate a game in terms of kills and still lose if the opposing team springs a few waves of catapults with damage buffs on them. Another option is giving players the ability to set waypoints for AI waves to attack, giving them some control while reducing the clickfest element from flipping through a dozen units every second to one order to the AI every two or five minutes.

Unfortunately, putting more emphasis on thought than on reaction time scares away the leet-kiddies, and players who would enjoy the concept in its strategy-heavy form don't expect complexity or subtlety from a game that seemingly simplifies the RTS form into a deathmatch, so it's difficult pinning down a target audience especially since no company investing in these games has the money for large-scale advertising. The result, of course, is that the dumbed-down versions (Heroes of Newerth, League of Legends) have gotten steadily more successful, while the much better attempts like Demigod are stillborn.