Deepmind AI releases DeepNash: The Stratego Conqueror!

Board games have historically served as an evaluation of AI development because they give us a controlled setting to analyze how humans and robots formulate and carry out tactics. AI has beaten humans at Go, Chess, Poker, and Backgammon.

Lee Sedol, the South Korean professional Go player of 9 dan rank was defeated by Google's Deepmind A.I. software, AlphaGo, on March 19, 2016. This was a historic accomplishment for AI; AlphaGo won the five-game series 4-1.

Source

But that’s not all!😉

In July 2022, DeepMind unveiled DeepNash, a model-free multi-agent reinforcement learning system that can outperform humans at the board game Stratego.

The classic board game of Stratego, which is trickier than poker and more difficult than chess and go, has now been perfected.

What is Stratego?

Stratego is a game of hidden information which is more complex than chess, Go and poker.

One of the few well-known board games that artificial intelligence (AI) has not yet perfected is Stratego. It is a challenging game that demands both long-term strategic thinking and the ability to deal with incomplete information, much like poker.

The 1947-first-published board game Stratego has since undergone several changes and has long captivated the interest of AI researchers.

Note: The goal of the game is to capture your adversary's flag, a piece that each army possesses.

Game Setup

• Left: The rating of the work. The 10 (Marshal) loses in battles when attacked by a Spy, and Bombs always triumph unless they are taken prisoner by a Miner. Higher-ranking pieces always prevail.
• Middle: A possible starting lineup. Note how the Flag is securely tucked away in the background, flanked by helpless Bombs. The two pale blue sections' "lakes" are never entered.
• Right: Displaying footage of Blue's Spy catching Red's 10 in the game.

Rules: How to play Stratego

1. Choose which player will be in charge of the Red Army and which will be in charge of the Blue Army.
2. Always, the Red Army strikes first.
3. Front the blue player with the blue side of the game board, and the red player with the red side.
4. Each player places one of their 30 pieces in each of the first three rows in front of them on the board (10 across by 3 deep). Only the centre two rows are vacant.
5. Only the commander of each army should be able to see where his or her piece is situated because every piece's "S" side should face the opposing player.

Note: The initial placement of your army is crucial and can make the difference between winning and losing.

To see more, Click here

DeepNash: Stratego AI

Model-free deep reinforcement learning and game theory are the foundation of DeepNash's innovative methodology. Its play style converges to a Nash equilibrium, making it very difficult for an adversary to take advantage of it. DeepNash has worked so hard that she is currently ranked among the top three human experts on Gravon, the largest online Stratego platform in the world. Players in Stratego do not have direct access to their adversaries' pieces, making it a game of perfect information.

In Stratego, information must be earned. Usually, a piece belonging to an opponent isn't disclosed until it comes into contact with that player on the battlefield. This stands in stark contrast to games with perfect information, like chess or Go, where both players are aware of every piece's location and identity. Reasoning in Stratego must be done over a large number of sequential actions with no obvious insight into how each action contributes to the outcome.

Stratego is extremely challenging to solve since there are so many different game states that might occur (represented as "game tree complexity" compared to chess, Go, and poker). This is why it appealed to DeepMind and why it has posed an ongoing challenge to the AI community for many years.

In the real world, DeepNash can be a game-changer. Its methods could be generalized to help solve problems often characterized by imperfect knowledge and unpredictable scenarios, like optimizing for traffic management to reduce driving times & vehicle emissions.

To read more about DeepNash: Click Here