DeepMind’s AlphaStar ranked above 99.8% of officially ranked human StarCraft II players in a 2019 evaluation. That figure describes its position on the rankings—not the percentage of matches it won. AlphaStar reached Grandmaster level with all three playable races: Protoss, Terran and Zerg.
What does “above 99.8%” mean?
The Nature paper reports that AlphaStar was “rated at Grandmaster level for all three StarCraft races and above 99.8% of officially ranked human players.” In other words, its rating placed it higher than more than 99.8% of the officially ranked human-player population used for the comparison. It does not mean AlphaStar won 99.8% of its matches. The 2019 Nature paper does not define this statistic as a match win rate.
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
StarCraft II: Wings of Liberty | $13.12 | Buy on Amazon |
| 2 |
|
Starcraft II: Legacy of the Void - Standard Edition | $42.94 | Buy on Amazon |
| 3 |
|
Starcraft II PC | $21.45 | Buy on Amazon |
DeepMind’s October 30, 2019 announcement described the comparison as “above 99.8% of active players on Battle.net.” The paper’s more specific wording is “officially ranked human players”; neither statement means every person who has ever played StarCraft II. DeepMind’s announcement also says AlphaStar achieved Grandmaster level for all three races.
How was AlphaStar evaluated?
AlphaStar played the full game online against human players. DeepMind said it used the official Battle.net server and the same maps and conditions as human players. The result therefore came from competition in StarCraft II’s live online environment, rather than from a demonstration limited to a simplified version of the game.
Recommended Free Tools
#1 Best Overall
- Fast-paced, hard-hitting, tightly balanced competitive real-time strategy gameplay that recaptures and improves on the original game
- Three completely distinct races: Protoss, Terran, and Zerg
- Units and gameplay mechanics distinguish each race
- 3D-graphics engine with support for visual effects and massive unit and army sizes
- Full multiplayer support, with competitive features and matchmaking utilities available through Battle.net
The reported evaluation covered Protoss, Terran and Zerg. Grandmaster is the game’s highest competitive league, so reaching that level across all three races was a distinct part of the achievement, alongside AlphaStar’s relative ranking among officially ranked humans.
How did AlphaStar learn to play?
AlphaStar’s result came from multi-agent learning, not just a single strategy trained in isolation. DeepMind’s announcement describes an initial supervised-learning stage followed by training in a fully automated league. The Nature paper explains that the approach used both human and agent games, with a diverse population of agents continually adapting strategies and counter-strategies. Those agents were represented by deep neural networks.
Rank #2
- This is a standalone product. It does not require any other version of StarCraft II to play
- Internet Connection Required
- Battle.net registration and Battle.net Desktop Application required
A league of competing approaches can expose an agent to a wider range of tactics than repeatedly training against one fixed opponent. In StarCraft II, where opponents can choose different races and strategies, adapting to counters is central to competitive play. The sources establish this broad training approach, but the headline statistic alone does not reveal how many matches were played or the uncertainty around AlphaStar’s rating.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.When did this happen, and what does it establish?
DeepMind announced the result on October 30, 2019, and the Nature paper was published in 2019. “Now” in the original headline refers to that historical milestone; it is not an announcement of a new 2026 system.
Rank #3
- Videogame Software
The finding establishes a high competitive standing in StarCraft II under the reported evaluation. It is not evidence that AlphaStar achieved human-level intelligence generally, or that the result transfers to other games or real-world tasks. The claim is specific to the game, player population and evaluation described by DeepMind and the paper’s authors.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
