Who’s the winner of computers vs humans in chess? At first, this seems like a question that can be answered in the blink of an eye. If we consider chess as a purely competitive endeavour, and our only goal to win the game, the obvious answer is: computers. Alternatively, this question could open up a discussion about the ultimate goal of chess. What if we believe, as Karpov, that chess is simultaneously art, science, and sport? Or that chess should simply exist to make you happy? From this perspective, you could reasonably argue that a significant goal of chess is, say, the search for beauty, in which case humans are perhaps better placed to triumph. Or, closer still, the best answer may be: a human with a computer.
So who wins in the match up of computers vs humans in chess? In this article, we’ll take a look at the history of computers in chess, how chess computers work, why chess is such an interesting challenge for those at the frontier of artificial intelligence and, most importantly, how the role of AI in chess has changed the way we play, learn, and enjoy the game (as well as how best to leverage this power to further and improve your own chess).
A Brief History of Computers vs Humans in Chess
Chess and computers have a long and increasingly overlapping history.
For those interested in discovering and furthering the boundaries of machine intelligence, chess has long been held as a prize target, both for its status as the signifier of human intelligence and, more practically, for the absolute nature of the game’s rules (which make it fertile ground for machine learning-based progress). For these reasons, building a computer that could outperform a human over a chess board has provided motivation for several mechanical offerings throughout history. These include such notable contenders as:
1. The Mechanical Turk (1769)
Also known as The Automaton Chess Player, this invention consisted of a lifesize model of a human head and body, sat behind a large table with a chess set on it. Created by a Hungarian engineer, Baron Wolfgang von Kempelen, in 1769, the Automaton had a relatively successful chess career. Whirring and clanking, the Mechanical Turk defeated several human opponents, including Napoleon Bonaparte and Benjamin Franklin among notable others, before eventually being exposed as an elaborate hoax: there was a (small but definitely human) master chess player concealed beneath the board, hiding behind the spinning gears and cogs and making masterful – but human – chess moves. The Mechanical Turk (the machine) was destroyed in a fire in 1854, thus closing the first chapter of machines vs humans in chess.
2. Deep Blue (1997)
Moving along somewhat in time and engineering sophistication, the pivotal moment in computers vs humans in chess came in 1997, when World Champion Garry Kasparov faced off against IBM’s supercomputer, Deep Blue, in the second of a pair of six-game matches under tournament conditions, and lost. Just the year before, Kasparov had tied the match against an earlier version of the same machine 3 ½ – 3 ½. The second match in particular was seen as a tipping point, and a victory for the power of brute-force computing in chess. Though far from the perfect chess computer, for openings alone, Deep Blue had 4,000 positions from 700,000 grandmaster games at its disposal for analysis, and the ability to calculate 200 million chess positions per second. This symbolic and controversial result marked the first time a computer program defeated a world champion in a match following classical tournament rules. If we were keeping score, this would be the moment that the machines edged ahead in the match of computers vs humans…
3. Stockfish, Leela, AlphaZero… (2021 and beyond)
These days, of course, the computers have us beat. A top player heading into a competitive game without computer preparation would be unthinkable. There are various engines available to the public, including the popular open-source Stockfish and Leela Chess Zero, that out-rate our greatest living player (hey Magnus!) by some 800 points. Special mention also goes to AlphaZero, Google’s neural network-driven program designed to master the games of go, shōgi, and chess, which changed the game for computers in chess when its then novel neural-network approach destroyed Stockfish’s best in a match in 2017, winning a closed-door 100-game match with 28 wins, 72 draws, and zero losses, all after studying chess for a total of nine hours. Due to these advances, now the more pertinent question is not how can humans beat chess computers, but how can we use them? Whichever engine (or Chessable course) you choose, in today’s chess world, you’ll be able to create a repertoire that has the backing of the world’s most intelligent software. The only limitation (for openings at least) is how far you can memorize the lines.
Next, let’s have a look at how chess engines work.
How Chess Computers Work vs How Humans Play Chess
Before turning on that engine, it is important to understand the difference between human play and moves made by a chess computer. If you have ever watched a chess game with expert commentary, you’ll often have heard a commentator refer to a move as a “computer move”. This is a good starting point to understand how computers calculate vs how humans calculate in chess.
A “computer move” is a move that, to the human eye, looks unnatural, awkward, or just plain confusing. Often this is because the point or tactical justification of the move will not reveal itself until six or seven moves later (or more), when the point of, say, retreating your bishop back to c8 on move 14 becomes clear. It’s also worth mentioning that a computer move could involve neglecting an obvious (to humans) path to an advantage for a more obscure (but objectively superior) variation, which a human would almost never think to play. Today’s chess engines have access to a mind-boggling amount of positions, historical games, endgames, tablebase draws, you name it. In short, they are ruthlessly efficient, they don’t get tired, and they basically never blunder!
The way chess computers work is complex and varied, ranging from brute-force calculation and position evaluation (coldly calculating their next move by evaluating its position using different parameters ranging from piece value, king safety, piece mobility, pawn structure, and so on) to using neural networks that more closely resemble the way the human mind works (honing in on one or two specific plans and investigating). The top chess engines of today combine these strategies. The practical result of this is that chess computers with sufficient processing strength will always be able to beat any human, calculating the most efficient move that will lead them to the best possible result from any position. Although chess computers today have not yet reached the status of “perfect” play, they should never lose to a human.
Another interesting result of this progress is that, as a human, having access to the internet, a laptop, or even a phone is now equivalent to having a super-super GM at your side, available at all times to show you how to play faultless chess.
We are all Patzers
As GM Alex Colovic put it in his instructional blog post, How to Analyse your Own Games, compared to the engine: “We are all patzers.” And, he’s right! Chess engines have powers of calculation that are simply not achievable by the human brain. For this same reason, often enough a computer’s evaluation or recommended move will present nothing but confusion to a human mind.
In contrast to a “computer move”, a “human move” could be described as one that is in fact based on the limitations of the human brain’s ability to calculate. This could present itself as a mistake, a move that has an obvious refutation, or one that is natural-looking but which overlooks an objectively better move that a computer would have made. A human move will likely follow the rules of essential chess principles, and its logic will (usually) be understandable to a fellow human chess player. Interestingly, a human player will understand nuances beyond the capacities of the machines, including the ability to gauge how far their opponent is able to calculate, what their plans are, and — of course — to consider the psychological pressures that also affect humans in chess.
Even though a chess engine can win at chess regardless of these factors, it is due to blind spots like this that AI remains far away from the ability to master intelligence outside of the defined rules of a game like chess. The ability to read and fully comprehend humans, with their infinite range of expressions, body language, cultures, emotions, and so on, remains out of machine reach (for now). Returning, then, to the discussion of the ultimate purpose of chess, if the ability to feel a move, to pull off a tactic based on an opponent’s perceived blind spot, or to experience the joy of coming out of an uncertain line unscathed is untouchable by chess computers, can they really claim to have mastered chess? We may be patzers, but we have access to the simple pleasures of learning and improving.
How Chess Computers Changed Chess
After all this, we can (curiously) conclude that the focus is no longer on computers vs humans in chess. Even Magnus Carlsen, generally acknowledged as the greatest chess player of our time, has been clear that he has no interest in competing with a computer. There would simply be no point to it.
The reason we can objectively say that we are currently seeing the best chess of all time is largely thanks to the role of computer analysis in chess. The theory, preparation, and investigation tools at the disposal of today’s grandmasters vastly outpower what was available even 30 years ago. The interesting question, now, is: What can Magnus (and by extension, humans in general) achieve with a computer? Aside from the great democratizing power that access to chess technology has had on chess (any child in, say, Norway can pick up a course on Chessable and learn how to play a flawless Sicilian Dragon, for example), the possibilities for human play at the top levels have greatly extended.
In the media furore following Kasparov’s loss to Deep Blue, one headline referred to the bout as “The Brain’s Last Stand”. But, as we now know, the chess world did not crumble, it only became stronger. While the development of technology and the bid to ‘solve’ chess will continue, the joy of chess, and chess education, remains firmly in human hands. So, while this blog article will sadly not provide a definitive answer to the computers vs humans chess (who is better) debate, the more we consider the intertwined history of human chess and machine chess, the more it becomes evident that it would be an error to view it as a simple competition with a single winner and loser. Instead, we should consider how chess computers have helped us improve chess as a whole, and how lessons learned from the training ground of the chess board in turn helps AI to develop in fields like medicine, economics, the study of the universe…
If you still want to calculate like Stockfish and plan like AlphaZero, there’s no course for that. But you can learn a practical alternative, with the course from GM Daniel Gormally: Sharp Middle Games: A Practical Guide to Chess Calculation. Or you can learn to think like former World Champion, Vladimir Kramnik, or how to convert endgames like Magnus Carlsen.