Vishy Anand on Staying in the Game

Vishy Anand talks with Tyler Cowen about what chess engines did to the sport’s store of general knowledge, what elite intuition and calculation actually feel like from the inside, and how he rebuilt his practice to stay in the world top ten past fifty.

Key ideas

1. Computers erased the value of accumulated experience in opening preparation. A player who has thoroughly checked one line with an engine can beat a specialist who has spent years on the same opening, because ‘the computer evaluation gap is just too strong’ — a verified numerical verdict overrides understanding built over a career. The practical countermeasure top players use is avoidance: gauge an opponent’s likely preparation and steer around it rather than try to out-understand it.
2. Chess has become irreducibly concrete since AlphaZero. Anand agrees with a diagnosis he attributes to Maxime Vachier-Lagrave: engines have dissolved general principles, so moves once dismissed as ‘a complete lack in chess education’ now have to be evaluated case by case rather than ruled out on dogmatic grounds. Nothing is safely generalisable any more; everything is a specific claim about a specific position.
3. Preparation is a certainty problem, not a depth problem. An engine-checked line is worthless unless verified completely — ‘a 99 percent guarantee… is simply not good enough’ before relying on it in a match — and that certainty decays on its own: new hardware and new engine versions force players to re-verify old preparation every six months to a year.
4. Escaping computers by changing the game doesn’t work — engines dominate the variants too. Fischer Random chess, which randomises the starting position, was built to defeat opening theory and, later, hoped to defeat computers; Anand reports engines are ‘just as good’ at it as at ordinary chess. Format variety survives anyway, not as an anti-computer strategy but as a way of keeping human-versus-human competition fresh.
5. His own expertise is pattern extraction and a felt sense that something is wrong, not raw calculation. Asked to name the cognitive ability that makes him distinctive, Anand points to distilling the right idea out of a mass of remembered games and noticing, from very few visual cues, that a position has gone wrong — closer to how he describes Magnus Carlsen’s play than to Fabiano Caruana’s more exhaustive calculation.
6. Competitive decline is invisible while it is happening and resists any deliberate fix. Anand’s own 2011–2013 slump illustrates a pattern he thinks afflicts most elite players: sustained success invites sustained study by rivals, and ‘a staleness gets into your game’ well before its cause is visible to the player living through it. The only remedy he has found is to stop diagnosing the slump and go back to playing without an agenda.
7. Staying competitive past fifty meant redesigning practice, not defending old habits. Anand semi-retired after the pandemic, plays a small number of chosen events a year, and rebuilds match fitness in the four or five months before each one — deliberately narrowing an ‘infinite variety of choices’ back down to the single move in front of him. Losing, he says, has only become more painful with age, not less.

Content

How computers levelled preparation

Anand dates the shift to the years around his world championship matches: once engines got strong enough to hand any diligent player a verified, superior move, the value of a grandmaster’s accumulated feel for an opening collapsed relative to raw verification. ‘If you have good computer moves — you understand them well, and you play them where they’re supposed to be played — then understanding cannot make up for it,’ he says. ‘The computer evaluation gap is just too strong. Imagine that the computer says, “This move and you are plus one.” I’ll take that against anyone. I’ll take that even against a specialist, because those are pretty good odds.’ The consequence, he notes, is that a club player who has thoroughly checked one line can now hold their own against a top grandmaster inside that specific opening — so the grandmaster’s real skill has shifted from out-preparing opponents to reading who has likely done the work and steering away from it.

That verification has to be total, not merely good. ‘The problem in preparation is not whether it’s detailed. The problem is whether you can believe in it. If you have not checked it 100 percent, it’s worthless. It’s like a 99 percent guarantee before you go into the operation. You want to have a 100 percent guarantee, right? 99 percent is simply not good enough.’ And the guarantee is temporary: every new engine version and every hardware upgrade reopens old analysis, so lines neglected for even six months have to be rebuilt almost from scratch before they can be trusted again.

Chess after AlphaZero: the end of general principles

Cowen puts to Anand a claim he attributes to the French grandmaster Maxime Vachier-Lagrave: that the post-AlphaZero lesson is how concrete chess really is, and how few durable generalisations survive. Anand agrees, and dates the trend broader than AlphaZero alone. ‘There are no general principles anymore. In fact, new general principles emerge. Nowadays, we push the h and a pawns whenever we want, and it’s not a bad move. Twenty years ago, these moves would have been considered betraying a complete lack in chess education. Now, it’s not there.’ Where a player once relied on inherited rules of thumb, everything now gets evaluated move by move, and roughly half of what used to be dogmatically ‘bad’ turns out to be fine in the right position.

He extends this to the practice of daring, unconventional-looking moves: chess used to require a human ‘stamp of approval’ — a top player playing an odd move first, so others felt safe copying it — and ‘the computer is now the new stamp of approval.’ The King’s Indian Defence, an opening some strong players privately regarded as unsound, illustrates how unstable even engine verdicts are: it was judged lost by the computer several years ago, later reconfirmed as lost, and has still found new life through move-order tricks that keep it ‘not quite refutable.’ Computers, Anand says, contradict themselves every couple of years, so even the pendulum of engine opinion swings.

Escaping the computer: format variety over rule changes

Fischer Random chess — which shuffles the pieces’ starting squares to deny memorised opening theory — was originally conceived to defeat prepared analysis, then hoped by some to defeat computers as well. Anand is direct about that hope failing: ‘Now, we realize we are not avoiding computers at all. They’re just as good at Fischer Random or No Castling as they are at chess.’ He describes experimenting himself with a ‘No Castling’ variant, sitting with an engine and discovering that almost every known opening has to be re-evaluated once a king cannot reach safety in the usual way — evidence, to him, that engines adapt instantly to any rule change rather than being tied to conventional chess’s accumulated theory.

What keeps the sport interesting, in his account, is not rule engineering against computers but deliberate variety for human competitors: alternating classical, rapid, and blitz formats, and mixed fields — young, unprepared juniors playing alongside heavily prepared veterans, as at the Tata Steel tournament — to reintroduce unpredictability. He also points to a genuine shift in the sport’s audience, which he attributes to the ‘revolution’ following the Netflix series The Queen’s Gambit: casual fans now outnumber serious ones, and classical chess’s share of attention has fallen, in his estimate, from effectively 100 percent a few decades ago to roughly 30–40 percent today, with faster formats absorbing the rest.

Intuition, calculation, and the shape of a winning idea

Cowen raises a distinction Magnus Carlsen drew on another interview: some players see only short lines but evaluate positions superbly (Carlsen’s self-description), while others, like Caruana, calculate much further ahead but trust evaluation less. Anand endorses the distinction as a real difference in how brains process the game — ‘some people fill in the gaps intuitively… you’re guided more by this sense of what is good and not’ — while noting that Caruana’s more exhaustive, almost computer-like approach is less prone to missing exceptions, at some cost in speed.

Anand’s own description of his defining ability supports this self-placement on the intuitive side: ‘the ability to pull out details from a mass of information that I’ve seen… being able to extract useful ideas from a lot of information. Also, good visual intelligence’ — plus, he adds, the more basic ability to sense from very few details that ‘something is wrong’ in a position. He illustrates the mechanism with a celebrated 2013 game against Levon Aronian, where he found a winning attacking idea not by exhaustive calculation but by half-remembering a piece pattern from a classic 1907 game (Rotlewi versus Rubinstein) that strong players carry in memory. He spent twenty-five minutes searching by elimination, ruling out lines that plainly failed, until an image of one particular square ‘flashed’ in his head; from there, he says, ‘the rest fills in very fast,’ the way ‘80 percent of the map’ resolves once the missing piece is found. A separate memory — that Vladimir Kramnik had once allowed a losing queen manoeuvre against the computer program Fritz in a near-identical structure — was enough on its own to warn him off a move that ‘would spoil a very nice position.‘

The mechanics of a slump

Anand’s own form dipped from roughly 2011 to 2013, a pattern he thinks recurs across elite players — Caruana went through something similar more recently — and one he finds genuinely hard to explain. ‘It’s a staleness that gets into your game. It’s probably an accumulation. If you’re doing too well for too long, others have been working nonstop trying to understand you. At some point, without you or them realizing it, the gap is closed. Then you’re encountering more resistance. You’re doing what worked perfectly before, but you’re encountering more resistance, and you can’t see why.’ The player experiencing it cannot see it coming and, once inside it, cannot diagnose a way out through more analysis. The fix he has found by experience is closer to a mental reset than a technical one: stop trying to solve the slump and go back to playing ordinary games for ordinary reasons, lightened of the pressure to fix anything. He is candid that this cannot be systematised — if it could, he says, top players would not have slumps at all.

Redesigning practice to stay in the game past fifty

At 53, and still ranked in the world’s top ten with no close rival near his age, Anand attributes his longevity to a deliberate change in how he trains rather than to unusual natural durability. He semi-retired from full-time competition after the pandemic — a decision the pandemic itself made easier, since everyone was effectively semi-retired at the time — and now plays a small number of chosen events a year, treating tournaments partly as social occasions with old colleagues. He typically trains hard for four or five months before an event he cares about, deliberately rebuilding the mental frame that day-to-day chess philosophy cannot substitute for: ‘The most important skill in chess is, if I was sitting at a board now with the clock ticking, what would I make in this position? Then suddenly, the infinite variety of choices you have becomes irrelevant. You have to make only one move.’ He also credits having something outside chess to return from — citing Hikaru Nakamura’s streaming career, which he says has taken pressure off Nakamura by giving him a second source of validation beyond results at the board, even though Anand doubts streaming improves anyone’s actual play.

Losing, and what he looks for in the next generation

Anand rejects the idea that he has made peace with losing. ‘I’m not a good loser. I’m a good actor. I know how to stay composed in public. I can even pretend for five minutes, but I can only do it for five minutes because I know that once the press conference is over, once I can finish talking to you, I can go back to my room and hit my head against the wall because that’s what I’m longing to do now.’ If anything, he says, it has worsened with age, as each loss now comes with self-directed anger at having repeated a mistake he feels he should have learned from long ago.

Asked what he looks for in the wave of young Indian talent from his home state, Tamil Nadu — players like Praggnanandhaa and Gukesh — Anand separates two qualities by age. In the very young, he wants to see fanaticism: total absorption in the game. As players mature, he looks instead for resilience — the capacity to absorb a bad defeat and keep competing rather than let one result compound into a spiral, a quality he says both Praggnanandhaa and Gukesh have already shown under real pressure. Where the chess understanding is not yet there, he looks for the same ‘sporting qualities’ that have kept him in the game himself: tenacity, fitness, and the refusal to let a losing position collapse into resignation before it has to.

See also

• Vishy Anand — guest; five-time World Chess Champion
• Tyler Cowen — host
• Kenneth Rogoff — Harvard economist and lifelong grandmaster-level chess player Anand cites directly in this conversation
• Kenneth Rogoff on Monetary Moves, Fiscal Gambits, and Classical Chess — companion Conversations with Tyler episode covering the same terrain of computer-era preparation, Fischer Random, and the classical-versus-faster-formats question
• Adam Neely on AI Music, Musical Deskilling, and the Ethics of AI Training Data — Neely rests his case on chess having ‘survived AI because the game itself is unaffected’; Anand, who actually played through it, reports the opposite — the engines rewrote what counts as a good move