Over the past two decades chess computers have absolutely exceeded human capabilities. However in late 2017 everything changed. The Deepmind team at Google created Alphazero. They gave it four hours and the basic rules of chess. The algorithm played millions of games against itself and then it was pitted against Stockfish – the strongest computer at the time. In a hundred games it won 28 drew 72 and lost none of them. However, not everybody was impressed. The issue was that Alphazero was using the entire google server system while Stockfish was playing on the equivalent of a laptop. That is why in late 2018 they played a rematch. in this post i’m going to give you one example and show you how powerful, creative and exciting Alphazero is to chess. The Google team released the sample size of 10 games from the initial match and this is one of them that i’m going to show you.
This game is one of the few that featured an opening possible after move 1.e4. Alpha zero did not like 1.e4 because it couldn’t find consistent advantages against the sicilian defence and 1.e5 positions. Alphazero simply could not create the imbalances it was able to create in 1.d4 positions. This game however After 1.d4 Stockfish played the move 1. e6 which allowed the transposition to the french defence after the move 2.e4. The position after what is called a classical french after 6 moves looked like this.

The most common and natural move in this position would be of course to take the pawn back with the knight. However Alphazero plays 7.Nb5. The point is to allow 7.Bb4 8.Bd2 Bc5 and create a very peculiar pawn structure. In many positions in the french defence black commits many pawns to light squares early. The consequence is that the bishop from c8 is very bad throughout the game being stuck behind the pawns. There are many moments when black needs to trade it off or break through the center.
Over the next couple of moves white suppresses black’s position and allows the move 15.Nxb2 after which it plays 16.Kxb2 which waives the rights to castle. If this move doesn’t look weird to you after 16.Bd7 it plays 17.Ke3! All of this looks completely ridiculous but black cannot do anything! Both of their bishops look awful and the only chance that black has of ever breaking out is playing f6 and if black ever takes on e5 white can recapture with Nxe5 activating their knight. To add insult to injury the white king is very safe on e3.

After maneuvering, improving placement of it’s pieces and trading a pair of rooks white plays a stunning sacrifice 30.Bxg6. After 30.Bxg5 31.Qxg5 fxg6 on the first glance it seems like white is just down a piece, but when we look deeper we can see that the bishop on b7 is practically worthless and alphazero continues its relentless attack with 32.f5, which is another pawn sacrifice.

The pawn cannot be taken for it would lead to a complete collapse of the black position. Instead the game continues with 32.Rg8 33.Qh6 Qf7 34.f6 Kd8 35.Kd2 Kd7 36.Rc1 Kd8. Here white switches targets and after attacking black on the kingside for the whole game it switches to the queenside in just a couple of moves and the infiltration of black’s position is inevitable. On move 52 stockfish resigned.

In 2018 Google Deepmind and the Stockfish team agreed to a rematch on equal terms to eliminate any doubts that Alphazero won due to better equipment. The second match had been played on equal terms. They played 1000 games against each other and this time Alphazero won 155 – 6 with 839 draws. This is a small example of how alpha zero completely changed chess computers forever. 210 games from the second match have been published on chessgames.com. There is also a very interesting book by grandmaster Matthew Sadler “Game Changer”, in which he looks at 20 of them which he found most interesting. If this topic is interesting to you I really recommend reading. Let me know in the comments which other Alphazero games do you like the most and if it still would have been the strongest chess engine on earth if Google Deepmind kept developing the project.
The game notation:
1.d4 e6 2.e4 d5 3.Nc3 Nf6 4.e5 Nfd7 5.f4 c5 6.Nf3 cxd4 7.Nb5 Bb4+ 8.Bd2 Bc5 9.b4 Be7 10.Nbxd4 Nc6 11.c3 a5 12.b5 Nxd4 13.cxd4 Nb6 14.a4 Nc4 15.Bd3 Nxd2 16.Kxd2 Bd7 17.Ke3 b6 18.g4 h5 19.Qg1 hxg4 20.Qxg4 Bf8 21.h4 Qe7 22.Rhc1 g6 23.Rc2 Kd8 24.Rac1 Qe8 25.Rc7 Rc8 26.Rxc8+ Bxc8 27.Rc6 Bb7 28.Rc2 Kd7 29.Ng5 Be7 30.Bxg6 Bxg5 31.Qxg5 fxg6 32.f5 Rg8 33.Qh6 Qf7 34.f6 Kd8 35.Kd2 Kd7 36.Rc1 Kd8 37.Qe3 Qf8 38.Qc3 Qb4 39.Qxb4 axb4 40.Rg1 b3 41.Kc3 Bc8 42.Kxb3 Bd7 43.Kb4 Be8 44.Ra1 Kc7 45.a5 Bd7 46.axb6+ Kxb6 47.Ra6+ Kb7 48.Kc5 Rd8 49.Ra2 Rc8+ 50.Kd6 Be8 51.Ke7 g5 52.hxg5 1-0
The game: https://www.chessgames.com/perl/chessgame?gid=1899422
The article does a great job showcasing how AlphaZero transformed chess AI, but it could dig deeper into how this innovation impacts human learning or gameplay strategies. Also, exploring how AlphaZero’s approach compares to traditional engines in real-world applications might add more depth. It’s a fascinating read, but a bit more context on its influence beyond chess could make it even stronger.
Very interesting! Are there fundamental principles of chess that humans have overlooked, which AlphaZero has now uncovered?
This article brilliantly captures the groundbreaking impact of AlphaZero on chess and its revolutionary approach to strategy and creativity. The detailed analysis of AlphaZero’s unique gameplay, including its unconventional moves like 17.Ke3 and the stunning 30.Bxg6 sacrifice, highlights how it redefined chess engines’ capabilities. The comparison with Stockfish provides valuable context, and the mention of the rematch under equal terms underscores AlphaZero’s dominance. The piece not only showcases a pivotal moment in chess history but also serves as a testament to the potential of AI to reshape traditional domains.
Great breakdown of AlphaZero’s impact! The fearless king walk and 30.Bxg6 sacrifice were stunning. The 2018 rematch proved its dominance—imagine if DeepMind kept developing it! ♟️????
AlphaZero revolutionized chess engines by demonstrating unmatched creativity and strategic depth. In its historic match against Stockfish, AlphaZero’s unconventional moves, like sacrificing pieces to control the board and dismantle the opponent’s position, stunned the chess world. Its ability to consistently outplay traditional engines proved that deep learning could surpass even the strongest rule-based systems. AlphaZero didn’t just beat Stockfish; it redefined the approach to chess computation, pushing the boundaries of what computers can achieve in strategy games.