Computers Revolutionized Chess. Magnus Carlsen Wins by Being Human

The World Chess Championship was already a week old when something stunning happened in Game 6: after nearly eight hours of play last Friday, someone actually won.

It was the first time in five years that a championship classical game—the format played under long time controls—didn’t end in a draw. The chess world could hardly believe what it was seeing. In the age of supercomputer-trained super grandmasters, there were widespread fears that world championships were becoming dull and predictable. Preparation seemed to trump inventiveness.

Instead, this era brought out the brilliance of perhaps the most brilliant chess player ever.

Magnus Carlsen, of Norway, steamrolled Russia’s Ian Nepomniachtchi 7.5-3.5 in the best-of-14 series, capturing a decisive victory that solidified his legacy as the greatest in the history of the sport. He has been the world champion since 2013—this was his fifth win—and is the highest-rated player of all time.

What even his rivals marvel at is how Carlsen, 31, has weaponized the computer revolution against them. He does it not by overpowering opponents with calculation, but by harnessing that digital knowledge to turn games into more human battles.

“Magnus is proud of saying that he’s probably the top player who works the least with the computer and is the least influenced by the computer,” said Carlsen’s coach, Peter Heine Nielsen. “He wants to trust his own evaluation, his human touch and to keep that.”

For decades now, chess grandmasters have used the insight of computers to study lines of play and explore the game’s bottomless possibilities. That has become supercharged in recent years as powerful engines, capable of planning depth far beyond the human mind, can be loaded on anyone’s laptop. Not only do they speed up calculation—they are able to take game situations and effectively see into the future.

When players such as Carlsen and Nepomniachtchi have months to prepare for a world championship, they are armed with countless hours of study on this type of software. Based on the programs’ evaluations, they learn the optimal moves in the most probable situations.

Carlsen knows all of these as well as—or better than—anyone on the planet. But he also possesses a mind so inventive that his best move is often not playing the best move. He would rather lead the game down a more obscure path where the player across the board might get lost.

“They say chess is a deep dark forest full of snakes and thorns,” said Danny Rensch, the chief chess officer of Chess.com. “Magnus brings his opponents into that forest.”

Chess champions have been tinkering with the best way to deploy computers for half a century. Russian Anatoly Karpov, in 1978, reportedly trained with an experimental computer dubbed “Tolinka,” or Little Anatoly. But back then, man still held the machines in check.

The BC/AD moment for chess was the victory by an IBM-built supercomputer named Deep Blue over then-world champion Garry Kasparov in 1997. Though Kasparov argued that Deep Blue had cheated, there was no denying that non-human players were progressing at phenomenal speed. Less than a decade later, another computer named Deep Fritz defeated world champion Vladimir Kramnik by playing esoteric lines that threw Kramnik off-balance.

Since then, chess engines have become so widely available that even beginners have access to instant analysis tools that study positions 18 moves deep. At the cutting edge, they are so sophisticated that grandmasters have a range of state-of-the-art engines to choose from, each one with its own styles and quirks.

Before the 2018 championship—the first ever in which all the classical games ended in draws—a new neural-network-based engine called Leela Chess Zero ushered in an even more advanced era. By playing hundreds of millions of games against itself, the computers were growing more powerful all the time. Some worried they were turning human players into pawns.

So the job of being a chess grandmaster became as much about grasping the subtleties of the Petrov Defense as knowing how to wrangle the limitless amount of suddenly available data.

That’s why competitors at the world championship now pick their teams of cornermen—fellow grandmasters known as “seconds”—based on who might best manipulate the power of the engines to predict their opponents’ behavior and prepare an opening surprise.

But here’s the twist: the most lethal use of computer-based analysis isn’t to find something that only the machine can see. It’s figuring out what it sees and dismisses that might still be useful. The dream of any computer-savvy chess player is to discover a string of moves that an engine doesn’t necessarily favor, yet taps into a line that their opponent hasn’t prepared.

“That’s the Holy Grail,” said grandmaster Cristian Chirila, who assisted world No. 4 Fabiano Caruana when he faced Carlsen for the world championship in 2018. “If you can get there, that’s a huge advantage.”

In any given situation, the engines might recommend any number of moves and suggest that they are all relatively equal. Those are the obvious ones to study. But by playing a more obscure move—perhaps even one that the computers suggest is disadvantageous—Carlsen thrives by throwing his opponents into that unfamiliar territory.

And Carlsen’s instincts allow him to pick up the subtlest edges in these unscripted scenarios. That’s how he prevailed in the sixth game of this championship. After 136 moves, the longest game in championship history, he scored the first decisive result.

“I think Game 6 was one of the most exciting games in world championship history,” Caruana said.

One reason is that Nepomniachtchi came into the match with a similar reputation to Carlsen—aggressive, unpredictable, and human when it counts. Nepomniachtchi, a 31-year-old who sported a man bun until he curiously got a haircut midway through the competition, attributed that to laying his chess foundation with traditional study.

The man now backed by a supercomputer from the Skolkovo Institute of Science and Technology in Moscow, known as “Zhores,” didn’t have access to a decent chess engine until he was 12.

“I think the generation [born in] the early 90s is probably the last one, which was raised without some major, major computer influence,” said Nepomniachtchi. “Compared to some teenage players who are shining already today… I’d like to think that my take is a little bit more human.”

Unfortunately for Nepomniachtchi, he revealed his humanity a little too much in this championship with critical blunders. Carlsen, meanwhile, proved that few humans are  harder to read than he is.

“We can all probably replicate what the top computers are saying,” Nielsen said. “What’s going in Magnus’s mind, only Magnus knows.”

via Wall Street Journal