The sentence appears in a 2024 interview with a top-five grandmaster, asked what he had been working on between tournaments: “I am studying my openings less than I have in fifteen years.” The interview was not framed as controversial. The grandmaster was Wesley So. The interviewer did not push back. The remark went almost unnoticed in the chess press. Inside the small circle of players whose job is preparation, it was greeted as an obvious statement that should have been made earlier.

Opening theory — the systematic study of the first fifteen-to-twenty moves of a chess game — has been the central intellectual labour of elite chess for the better part of a century. The Encyclopedia of Chess Openings in five volumes; the New In Chess yearbooks; the ChessBase MegaBase with its eight million reference games; the personal databases of every world champion since Botvinnik. The activity has consumed more grandmaster-hours than any other single line of work in the game. As of the mid-2020s, it is being deprioritised at the top level for a reason that should have been obvious from 2008 onward: there is no theory left to discover.

The era of the novelty

The classical era of opening theory ran from roughly 1925 to 1995. The shape of the work was the novelty — a previously unseen move, ideally on move twelve or fifteen, ideally with a sharp follow-up, played in a serious game and analysed for years afterward. Players published opening monographs. Sidelines were discovered in correspondence chess and brought to the over-the-board world. A novelty in a Kasparov–Karpov championship match in 1985 could keep the strongest players in the world busy for a decade of analysis.

Most of those novelties were intuitive. Karpov’s 12…Be7 in the Caro-Kann was the result of two years of personal analysis. Kasparov’s 15.Nb1 in the King’s Indian Sämisch was conceived at home before being played in 1992. The novelty was an artefact of human judgement under uncertainty: nobody knew which move was best, the analyst guessed correctly, and the guess held under tournament-grade scrutiny.

The early engines — Fritz, Junior, the first Rybka — changed this slowly. They could check tactics quickly but could not evaluate the positional content of a novelty well enough to confirm it. The period from 1998 to 2010 was a transitional one in which engines accelerated analysis without replacing the analyst.

When engines outpaced theory

The Stockfish project’s rise from 2010 onward, the AlphaZero result in 2017, and the Leela Chess Zero project’s parallel emergence finished the transition. By 2018, every elite team had access to engines that evaluated positions stronger than any human. By 2020, the engines were running on hardware capable of analysing key positions to terminal evaluations within hours rather than weeks. The cycle of human discovery and engine verification reversed: engines proposed novelties; humans tried to remember them.

The structural consequence was that the frontier of opening theory — the set of positions still under analytical contention — collapsed. In most main lines of most main openings, the engines have run terminal analysis trees twenty moves deep. The result is invariably some shade of equal or near-equal. The Berlin Defence (which we covered in the Ruy Lopez Berlin Defense article) has been understood as a forced draw with best play since approximately 2015. The Marshall Attack has been similarly exhausted. The Najdorf English Attack has had its main tabiya analysed to terminal evaluations along all branches. Of the openings still considered theoretically alive in 2026, most are alive only in their fourth-or-fifth-most-popular sub-line.

Theory is finished. What remains is memory. — Vladimir Kramnik, ChessBase interview, 2023

The grandmaster who studies openings in 2026 is not discovering novelties. They are memorising engine-vetted move trees in positions where their opponent — if also well-prepared — will know the same trees. The work has become custodial rather than creative.

The Berlin Wall

The single most consequential opening result of the past quarter century was Vladimir Kramnik’s adoption of the Berlin Defence against Garry Kasparov in their 2000 world championship match in London. Kramnik played the Berlin in every classical game in which he had Black with one exception. He drew them all. Kasparov, who had built his entire career on attacking play against the Ruy Lopez, could not break through. He lost the match without losing a single Black game himself.

The Berlin in 2000 was a novelty in the older sense — Kramnik’s preparation had identified specific equality lines that Kasparov had not deeply analysed, and the system suited Kramnik’s positional preferences. In 2026, the Berlin is theory’s terminal position made into a tabiya: every elite player knows the main lines, the queens come off by move ten, the endgame structure is precisely understood, and the result with best play is a draw. Carlsen, Gukesh, and Ding have all played both sides of it in world championship contexts. None of them have introduced a meaningful theoretical novelty in the past three years.

The Petroff problem

Similar exhaustion has happened in the Petroff Defence (Petrov’s Defense in older usage), the Marshall Attack, the Sveshnikov Sicilian, the Catalan main lines, and the Slav Meran complex. The pattern is the same: a once-rich field of analysis has been resolved into a small set of equal endgames whose moves players memorise rather than discover.

The professional response to this has split. One school — represented by Caruana, Nepomniachtchi, and most of the eastern European top — continues to pursue exhaustive preparation, treating opening memory as a competitive baseline that simply cannot be conceded. The other school — represented by Carlsen, Gukesh in his post-championship play, and increasingly Pragg — has explicitly de-emphasised classical opening preparation, choosing systems for their middlegame richness rather than their theoretical edge.

The two schools meet in tournaments. In the past three years, both have won and lost roughly proportionally to their ratings. The data suggests that exhaustive opening preparation no longer produces a measurable Elo advantage at the top of the rating list — the strongest players draw most of their openings either way, and the games are decided in the middlegame.

Carlsen, who refused

Carlsen’s reaction to the death of theory has been the most visible and probably the most consequential. He plays opening systems he can sustain at the board with thirty minutes of preparation. He has stopped contesting the home-preparation race. His Berlin Defense games tend to leave theory by move twelve into positions engines call equal but he plays better than his opponents. His White-side openings cycle through the English, the Catalan, and the Italian without obvious theoretical motivation, by feel.

The pivot has been most explicit in his commitment to freestyle (Fischer Random) chess. Starting from one of 960 randomised positions with no openings theory possible, freestyle restores the practical-calculation primacy of the early game. Carlsen has won the Freestyle Grand Slam in each of its three editions and described the format as the most enjoyable chess he plays.

This is the limit case of the argument. If opening theory is exhausted, then a format that abolishes opening theory entirely is more honest to the underlying game. The argument is not yet mainstream but it is no longer marginal — half the world top ten now plays freestyle seriously.

What is left to study

For the player who studies chess for personal improvement rather than at world-championship level, the death of theory is mostly good news. The student of 2026 has access to engine-evaluated analysis of every meaningful opening position in human history, for free, on Lichess. The student does not need to memorise main-line theory because, almost certainly, they will face opponents who also do not. The work of becoming a strong amateur in 2026 looks more like the work of becoming a strong amateur in 1926 than it does like the work of becoming a strong grandmaster in 2010: study endgames, study tactics, understand middlegame strategy, play long games and review them carefully.

For elite chess, the situation is more interesting. Opening theory will not stay buried — it will be replaced by other forms of preparation. Position-specific preparation against a particular opponent is the obvious next layer. Endgame preparation, particularly the seven-piece tablebase regions, is becoming a serious area of study. Engine-vetted middlegame plans, in the form of “play this entire middlegame on autopilot” preparations, are the third growth area.

What is gone is the romantic image of the analyst at the board late at night discovering a move no one has seen. That image was real for sixty years. It has been replaced by a quieter labour: not the discovery of truth, but the memorisation of it. Chess is now a game whose objective frontier has been pushed to the edges of human computational capacity. Inside that frontier, the players play.

References

Cross-links inside Caissly: The Najdorf, examined and The Berlin Defense document two of the most heavily theory-exhausted openings of the modern game. The term novelty covers the historical concept that the engines have ended.

Issue Nº 003 · The Magazine · The Caissly Editorial