Poker / Indie dev/ ポーカー収支管理アプリ@poker_brm_jp を作っています / MTT学習アプリも作ってます
MTT学習アプリはこちらから➡️
Joined April 2024
- Tweets587
- Following500
- Followers271
- Likes999
Preflopのソルバーに限った話だと
各ノードのPostflopのCFRを解いてその価値をpreflopのCFRに伝えて解くという2段階の仕組みなんだけど、これらの噛み合わせ方やうまく抽象化するのが難しかった
zeno solver本当にクラッシャーすぎる
ソルバー自体は作れる人は多そうだけどAI solveを実現するには大量のcpuと時間が必要だから、それがGTO Wizardの優位性だと思ってたんだけどそこもクリアしてしまうとは
GTO Wizard焦ってんちゃうん
自分が作ろうとしてたものに近しいものが出てきてしまった..😭
計算速度速すぎるから、まともに計算せずに機械学習使って予測してると思うんだけど教師データどう用意したんやろ
zenosolver.com/
ruby666 retweeted
Today we are revealing how GTO Wizard AI computes ICM for large tournament fields, and publishing our proprietary algorithm, free, for everyone.
Traditional ICM calculators top out at a few dozen players. In 2024, we showed that ICM for thousands of players can be computed in seconds, a problem academia still considered intractable.
We are sharing it because we believe in contributing to research, and because we want others to build on it, in poker and well beyond it. The same math applies to ranking problems in many other industries.
Behind GTO Wizard is a team of researchers and scientists who built the state-of-the-art poker AI and work closely with @MIT and other leading universities around the world, from guest lectures to joint research.
This problem was cracked by the brilliant Wataru Inariba (@b_inary). Wataru, thank you for your outstanding work.
Read the full paper on arXiv 👇 arxiv.org/abs/2610.06028
あとで読む
Today we are revealing how GTO Wizard AI computes ICM for large tournament fields, and publishing our proprietary algorithm, free, for everyone.
Traditional ICM calculators top out at a few dozen players. In 2024, we showed that ICM for thousands of players can be computed in seconds, a problem academia still considered intractable.
We are sharing it because we believe in contributing to research, and because we want others to build on it, in poker and well beyond it. The same math applies to ranking problems in many other industries.
Behind GTO Wizard is a team of researchers and scientists who built the state-of-the-art poker AI and work closely with @MIT and other leading universities around the world, from guest lectures to joint research.
This problem was cracked by the brilliant Wataru Inariba (@b_inary). Wataru, thank you for your outstanding work.
Read the full paper on arXiv 👇 arxiv.org/abs/2610.06028