Flutter to Shut Down Betfair in Italy Amid Gambling Framework Changes
1 min read
StrategyNovember 6, 2019·1 min read
DeepMind's AlphaStar AI: Not Quite the StarCraft Master It Seems
DeepMind's AlphaStar AI achieved a high level of play in StarCraft, surpassing 99.8% of human players. However, the AI was constrained to mimic human capabilities, including action limits and reaction delays.
AlphaStar was limited to 200 actions per minute, mimicking human capabilities. This limitation was imposed to ensure that the AI's victories were due to strategy and tactics, rather than sheer processing power.
Additionally, AlphaStar had a reaction delay of 0.2 seconds to mimic human response time. This delay was implemented to prevent the AI from reacting instantaneously to changes in the game, further leveling the playing field.
An earlier version of AlphaStar, which played in January, could perform around 15-20 effective actions per second without the same limitations. This version of the AI was not constrained by the same rules, allowing it to achieve a higher level of micro-control and efficiency.
Share this article
By
Argus RivenI've spent years at the tables — cash and tournaments — and just as long shaping poker copy behind the scenes: room reviews, bonus fine print, tournament notes. At PokerInsider I write the way I wish more sites would: clear structure, honest trade-offs, no hype. Whether you're new or already grinding, I care about what the offer actually costs you — rake, variance, and the fine print — and I treat responsible play as part of the job, not a footnote.
♠More in Strategy
Poker Pro Jackstack99 Shares Insights on Strategy and Bankroll Management
2 min read
Poker Strategy: Value Betting vs. Allowing Opponent to Bluff
1 min read
←All Articles♠ PokerInsider