DeepStack

The 1st #AI to beat pros at heads-up no-limit hold'em #poker - #DeepStackAI uses heuristic search techniques, #DeepLearning and regret minimization

[DeepStack @DeepStackAI]

⚡️ “DeepStack: AI beats poker pros for 1st time”

https://t.co/Yhdl7nezU0

[DeepStack @DeepStackAI]

Looking to go head-to-GPU and toe-to-transistor against #DeepStackAI, the 1st #AI to beat human pros at heads-up no limit #poker? Join us tomorrow at @TWoSEdm's Dark Matters: https://t.co/J8j6gZfVEx #Yeg

[Faculty of Science, University of Alberta @ualbertaScience]

Learn about ASTech Awards Finalist Michael Bowling's work with @DeepStackAI https://t.co/m8LVozlKjI #UAlberta #AI https://t.co/0PVfeGMMOu

[DeepStack @DeepStackAI]

@IFPMatchPoker @thinkingpoker @PokercastAdam @tchanpoker @NateMeyvis @bparispoker @MikeMcDonald89 @DutchBoyd 12/ Here's a summary table. The (X +/- Y) columns show estimated win probabilities, and 95% confidence, based on the W/L predictions.

[DeepStack @DeepStackAI]

@IFPMatchPoker @thinkingpoker @PokercastAdam @tchanpoker 11/ #AIVAT on DeepStack vs @NateMeyvis @ThinkingPoker @bparispoker @tchanpoker @MikeMcDonald89 @TaylorvonK @DutchBoyd @PokercastAdam:

[DeepStack @DeepStackAI]

@IFPMatchPoker @thinkingpoker 10/ Watch for an upcoming interview with @PokercastAdam and @tchanpoker on #AIVAT and these results too!

[DeepStack @DeepStackAI]

@IFPMatchPoker 9/ Last week, we were guests on @ThinkingPoker to describe #AIVAT and its analysis of our freezeouts: https://t.co/Sx4gGRVY0k

[DeepStack @DeepStackAI]

@IFPMatchPoker 8/ While each freezeout always ends in a win or a loss, #AIVAT gives us win probabilities for winning a freezeout.

[DeepStack @DeepStackAI]

@IFPMatchPoker 7/ #AIVAT also works on freezeout tournaments, like the Twitch matches we played earlier this year: https://t.co/YIYhD6lF8v

[DeepStack @DeepStackAI]

6/ And described by @IFPMatchPoker using an example hand between Martin Sturc and DeepStack: https://t.co/oyZRTv11Ap

[DeepStack @DeepStackAI]

5/ #AIVAT is provably fair, described in mathematical detail in a research paper: https://t.co/xboozkSta8

[DeepStack @DeepStackAI]

4/ 3,000 hands evaluated with #AIVAT is statistically equivalent to almost 100,000 hands of poker!

[DeepStack @DeepStackAI]

3/ It can then subtract these terms from the outcome, producing a "luck-adjusted" estimate with much less variance than the actual outcome.

[DeepStack @DeepStackAI]

2/ #AIVAT analyzes hands in finished matches and estimates how much of a player's winnings came from lucky or unlucky cards/actions.

[DeepStack @DeepStackAI]

1/ #DeepStackAI uses a new evaluation technique called #AIVAT for separating luck from skill in games like #poker

[Demis Hassabis @demishassabis]

Spent a really enjoyable day brainstorming AI ideas with Rich, Mike, and Patrick. So exciting to welcome them and their team to #DeepMind ! https://t.co/MBUZte9XdB

[Google DeepMind @GoogleDeepMind]

We're opening a research office in #Edmonton, Canada, in collaboration with @UAlberta, led by RL pioneer Rich Sutton https://t.co/QQhuYngdkv

[Andrew Brokos @thinkingpoker]

I'm heads up vs a very nice Dutchman who does not seem to have much HU experience. I, on the other hand, have trained against @DeepStackAI.

[DeepStack @DeepStackAI]

All six of our #twitch #poker matches are now on our YouTube page: https://t.co/VwqVSly7yw

[DeepStack @DeepStackAI]

@andyjeffries 3+ player games are a tough challenge both in theory and in practice, but after 2p that's the next challenge problem for the field, yep!

[DeepStack @DeepStackAI]

@andyjeffries No. DeepStack only saw the showdown hands, while the humans saw both the showdown hands and DeepStack's folded hands.