Chess is a two-player strategy board game played on a checkerboard with 64 squares, arranged in an 8×8 grid. Played by millions of people worldwide, chess is believed to be derived from the Indian game chaturanga sometime before the seventh century.
Chess has gained tremendous popularity, and it’s only grown in recent years. There’s no better time to analyze the role of artificial intelligence in improving the quality of chess.
A Brief History of Chess AI
- 1951: Alan Turing published the first program on paper theoretically capable of playing chess.
- 1989: Chess world champion Gary Kasparov defeated IBM’s Deep Thought in a chess match.
- 1996: Kasparov defeated IBM’s Deep Blue in another match.
- 1997: IBM’s Deep Blue becomes the first chess AI to defeat a grandmaster in a match.
- 2017: AlphaZero, a neural net-based system created by DeepMind, beats Stockfish, with 28 wins, 72 draws and zero losses in 100 chess matches.
- 2019: Leela Chess Zero (LCZero v0.21.1-nT40.T8.610) defeats Stockfish 19050918 in a 100-game match 53.5 to 46.5 for the Top Chess Engine Championship season 15 title.
- Present: Modern chess AI engines deploy deep learning to learn from thousands of matches. They regularly have estimated FIDE ratings — chess’ rating system — above 3400, far beyond the best human players. But these ratings are not official.
Artificial intelligence continues to advance at a rapid pace, offering more use cases for improving overall quality of life. It also promises to reshape numerous industries like gaming, and chess is no exception.
We’ll first go over a brief introduction to the history of AI in chess. Then, we’ll learn about the modern evolution of chess engines and the influence of AI in the universe of chess. Finally, we’ll conclude with why every developer should implement the programming of a similar chess engine.
Chess AI History
In 1951, Alan Turing was the first to publish a program, developed on paper, that was capable of playing a full game of chess. Consistent developments were made in the upcoming years. New chess games and chess engines were developed during this period. However, these AI chess engines had yet to achieve success on a higher level, possibly due to the lack of effective resources and tools.
In the 1980s, chess world champion Garry Kasparov reportedly claimed that AI chess engines would never reach a level where they could defeat top-level chess grandmasters. His statement would remain true for a few years, as he successfully defended his throne in 1996 against IBM’s Deep Blue in a match over six games with 4:2. He had also defeated Deep Blue’s predecessor, IBM’s Deep Thought, in a 1989 clash.
A year after losing to the world champion, Deep Blue came back to beat the world champion in a rematch. Kasparov was defeated by Deep Blue in the rematch with 2.5:3.5. Although there is a small controversy regarding the authority of this win, it’s mostly regarded in favor of the chess engine. In one of Kasparov’s interviews, he also agreed that he lost fair and square.
Chess engines have come a long way from the era of the 1950s to the current generation of chess games. In the next section, we’ll analyze the modern evolution of chess engines to gain a better picture of the accomplishments made by AI in chess.
Modern Evolution Of Chess AI
Artificial intelligence algorithms developed for human play utilize many different principles. While it’s complicated to discuss what algorithm each engine uses for its functionality and workings, we know that a few popular engines like AlphaZero make use of neural networks, deep learning and neural net-like automation. Leela Chess Zero utilizes an open-source implementation of AlphaZero, which learns chess through self-play games and deep reinforcement learning.
Today, modern chess engines are so well-developed that they won’t drop a single game to human players. Even the current reigning world champion failed to beat the best modern chess engine once in a span of 100 games. The world champion has a FIDE rating of over 2800 across all formats. The contest usually takes place in a classical time format.
The match was against Stockfish 9, which has a rating of 3438 (engine ratings are not the same as FIDE ratings, but the player pool for engines is much stronger than for humans, so theoretically a FIDE rating for Stockfish 9 would be even higher). The result of the two modern clashes of chess engines are as follows:
- 2017: AlphaZero, a neural net-based system, beats Stockfish, with 28 wins, 72 draws and zero losses in a 100-game match.
- 2019: Leela Chess Zero (LCZero v0.21.1-nT40.T8.610) defeats Stockfish 19050918 in a 100-game match 53.5 to 46.5 for the TCEC season 15 title.
These results, coupled with the rise of chess engines and neural networks and deep learning-based chess networks taking over the chess world by storm, are a huge sign of the potential for greater possibilities.
How Chess AI Has Influenced Chess
Artificial intelligence has influenced the way chess games are played at the top level. Most of the grandmasters and super grandmasters (rated at a FIDE above 2700) utilize modern AI chess engines to analyze their games, as well as the games of their competitors. There’s a complete turnaround in the way in which chess games are now played.
The basic opening theories and other analytical concepts are thoroughly analyzed. In classical formats of chess, you’ll typically see these high-level players make about 10 to 15 of their first moves from previously analyzed games or the top engine recommendations.
The quality of top-level games has also drastically improved because of the help of these engines. It’s almost impossible to rate or compare a modern player against a legendary player from past decades thanks to the enormous improvement these chess engines have provided.
Some have argued that chess engines have had a negative effect on the game because it’s more about theory than actual practice and play. Others argue that the influence of AI on chess has led to a drastic improvement in competition and further advancements are yet to be made to challenge modern players.
Ultimately, AI chess engines make these players much better both with opening theory and other tricks up their sleeves. There is still room for human error, and these errors can be exploited by players to gain advantages and create interesting games of attacking chess.
Advantages of Building Your Own Chess AI
No matter what level of programmer you are, you will have a great time exploring the numerous aspects of building your own chess engine. I firmly believe that programmers and developers should try to develop a game with Python and AI.
Tips for Beginners
If you’re a beginner, explore the artistic features while implementing the structure of the board and the pieces. You will learn the features of functions and classes to implement the structure of the board, as well as place the pieces in their respective positions. You can experiment with the two colors of the chessboard and the chess pieces. Finally, you will gain a cool graphical structure that you’ll design from scratch using your coding skills.
Tips for Intermediate Developers
If you are an intermediate AI developer, you can start implementing the functionalities of the pieces and their respective movements. Each piece has its own representation and notation for the particular motion. You need to implement these positions and also compute the capturing of each piece. All this work does not have to be perfect to develop the skills and get a working program.
Tips for Advanced Programmers
For advanced programmers and AI developers, study a bunch of games and chess engines that have been developed in recent years. Using the data available, you can construct deep learning neural networks for your chess engine. The chess engine can learn from games played since the 1800s. Chess has a huge data set, allowing developers to create some quality chess engines from scratch by themselves.
The Future of Chess AI
Artificial intelligence has undoubtedly changed the landscape of chess. While its impact has come with pros and cons, the excitement generated by AI is undeniable. AI has great potential in the field of chess, with developments in chess engines increasing rapidly. The modern world of chess may witness more impressive accomplishments than before, and I am excited to see where these improvements reach. I also hope all of you will try to implement a chessboard or a chess engine of your own as a fun project.
Frequently Asked Questions
What was the first chess AI program?
The first chess AI program was created in 1951 by Alan Turing, although the program was only a theoretical concept developed on paper.
When did a chess AI first defeat a world champion?
IBM’s Deep Blue became the first chess AI to defeat a human world champion when it bested Garry Kasparov over the course of six matches in 1997. This followed Kasparov defeating Deep Blue in 1996 and IBM’s Deep Thought in 1989.
What is AlphaZero, and how did it impact chess AI?
AlphaZero is a neural net-based system that leverages deep learning and neural networks. Created by Google’s DeepMind, AlphaZero demonstrated the power of self-learning when it dominated Stockfish in a 100-game match, with 28 wins, 72 draws and zero losses.
