20 Years. 450,000+ Games Analyzed. The definitive research report on the evolution of correspondence and daily chess is now live.
One month ago we wrote a blog post dealing with Chess960. And already now we feel that it would be good to write a more elaborate update. Reason: the results of our ’20 years LSS Research Report’.
When Bobby Fischer unveiled his shuffled-back-rank variant in Buenos Aires in 1996, he was not trying to break chess. He was trying to rescue it — from itself. Nearly thirty years later, after our own twenty-year study of 452,542 LSS games confirmed that Master-level standard correspondence chess now draws a remarkable 97.3% of the time, while Chess960 at the same level draws only 54.2%, I have become convinced of something I want to argue carefully and honestly in this piece:
Chess960 is not a gimmick, not a novelty, and not a dilution of “real” chess. It is the purest available test of chess understanding — the ability to navigate a position you have never seen, with knowledge you cannot have memorized.
Let me explain what I mean by understanding, why engines themselves struggle to tame it, and why — for the next twenty years at LSS — we are making it a central pillar of what we do.
1. Memory Ends Where Understanding Begins
In regular chess, even a reasonably strong club player — not just a super-grandmaster — can travel the first 10, 15, sometimes 20–25 moves of a well-trodden line on something close to autopilot. The opening is recognized before it is analyzed. Then, move by move, the position grows a little less familiar, until you are finally on your own. The trouble is that, thanks to centuries of theoreticians and decades of engines, by the time you arrive in genuinely unknown territory, you are very often already standing in a drawish position. The fight has been pre-decided in a library.
Chess960 dissolves that comfort almost immediately. The unknown begins not after move twenty, but after move two or three. If you open the Lichess Chess960 database, or examine the games between the strongest engines, the “theory” is astonishingly thin — not the hundreds or thousands of games stacked behind a single ECO code (and remember, the Encyclopaedia of Chess Openings runs to five volumes, each with a hundred coded chapters, each code further branching by two digits), but often a mere handful of games for a given starting position.
And here is the part people underestimate: you cannot even learn reliably from your own opponents. In standard chess, beat a rival in a line once and you carry that experience into the rematch. In Chess960, the next time you meet that same opponent, it is overwhelmingly more likely you will face one of the other 959 positions than the one you have already studied together. Recognition gives you almost nothing. Only understanding travels.
That, to me, is the definition worth defending. Memorization is the storage of answers. Understanding is the capacity to generate good answers when no stored one exists. Chess960 forces the second and starves the first.
2. The Centaur Paradox: Why Even the Engine Can’t Rescue the Draw
Here is the empirical jewel of our twenty-year study, and the fact that most surprises people: in Chess960, Masters draw only 54.2% even with engines allowed. The machine — supposedly the great equalizer, the great draw-maker — does not flatten the game into a handshake the way it does in standard play, where it pushes Masters to 97.3%.
Why not?
My own view is that engines are, in a meaningful sense, like us. They were built by strong chess players who were either programmers themselves or who could communicate well with programmers. The modern AI layer adds enormous power, but the underlying logic is unchanged: an engine learns from prior mistakes and prior experience — repeating and refining what works, discarding what does not. That is precisely a human method, executed at superhuman scale.
But scale needs material to feed on. With 960 starting positions instead of one, the branching tree of plausible early play is vastly wider and the accumulated “experience” vastly thinner. After move twenty or thirty, a Chess960 game normalizes — it comes to look like any game lifted from a classical tournament. The hard part is everything before you arrive there, where you must solve real problems with a fraction of the background material.
I have felt this directly in my own games. The single biggest eye-opener was this: in Daily and correspondence chess I weigh my own judgement — including gut feeling and my preference for certain structures — against the engine’s recommendation, knowing there is rarely one best move but often a cluster of reasonable, roughly equal ones. In Chess960, I have repeatedly watched engine evaluations swing far more violently from one move to the next. The machine is less certain. And that uncertainty is a signal: the player should give the engine’s verdict a lower weight, and his own understanding a higher one.
That will not help a weaker player — he will still, on balance, blunder more than the engine. But at the very top, the best Chess960 players probably do hold defensible preferences against the machine, at least for now. Remember: most engines were born as regular chess engines, later adapted with a few modifications to handle 960. That is not the same as an engine built from the ground up for the variant. The Centaur, in Fischer Random, is a more equal partnership than it has become in standard play — and that is exactly why the games stay alive.
A note on what a draw really is. I am not, in any way, against a beautiful draw. That chess, when properly defended, tends toward the draw is not a flaw — it is a sign that the players are good. Checkers and draughts were solved into draws long ago, and they still crown champions. A game in which Black must fight for the half-point is not a lesser game. The problem Chess960 addresses is not the existence of draws. It is the frustration of a sport in which too many tournaments produce no clear winner at all.
3. Chess as a Science Improving — and a Sport That Still Needs a Winner
We should be honest that the rising draw rate is mostly just chess getting better. Engines improve step by step; so do the human super-grandmasters; so do the theoreticians and scholars of the game. Treat chess as a science and you should expect the standard of play to rise, and — chess being a finite, closed game — you should expect that some openings will eventually be classified as simply unplayable for Black, and others as dead drawn. White’s small first-move edge is real. None of this makes the game less interesting.
I think the right analogy is financial analysis. Today’s analyst has bigger databases, more theory, and a wider investable universe than his predecessor of thirty or fifty years ago. Is he therefore a lesser analyst? Of course not. And an analyst restricted to a single country or sector is genuinely disadvantaged against one with the whole universe — even if, in a lucky year, the narrow specialist may outperform by coincidence. Chess as a science behaves the same way.
But chess is also a sport, and here the three-result structure — 0, ½, 1 — becomes a problem. A tournament can end with every game drawn except one, decided by circumstance rather than brilliance. I have heard of (and you can imagine) the grim version: a field of draws, save a single agonizingly slow game against an unsporting procrastinator, won only because the honorable older opponent — already unwell at the start — passed away mid-game. Dramatic, yes; absurd, certainly; but it captures why a sport that too often produces no real victor is a sport in trouble.
There are remedies, and we take them seriously:
- Asymmetric scoring — rewarding a win with, say, 1.5 points, so that winning one and losing one (1.5) beats drawing two (1.0), is a big incentive for fighting chess.
- The engine as jury — asking a very strong engine to judge accuracy, averaging each player’s centipawn precision across the tournament, and breaking ties in favor of the more accurate player. Chess becomes, in part, a judged sport, like gymnastics or figure skating.
- And, above all, format — and this is where Chess960 earns its place.
Crucially, our own data shows the draw problem is overwhelmingly a top-level phenomenon. At lower categories, draws have risen over twenty years too, but far more slowly. It is the Formula 1 effect: among the very best drivers in the very best cars, the gaps compress and the field bunches; lower down, with lesser machinery, the spread is wide and the results decisive. Chess960 widens the field back out — precisely where standard chess has narrowed it to a sliver.
4. Not Destruction — Resurrection
There is a tempting but wrong way to describe what Chess960 does to opening theory. It is tempting to say it destroys theory. It does not.
The opening theory of regular chess has been built over centuries, like a science, and on average it has improved — in every variation and every opening. What Chess960 changes is the scope of any one body of knowledge: you can dig deeper and deeper into one of the 960 positions, but what you learn there does not automatically transfer to the other 959. A strong chess player’s training probably makes him a marginally better Go, draughts, or even sudoku player — but those are genuinely different fields, with their own rules and demands. The brain is shared; the specifics are not.
So Chess960 does not destroy theory. It resurrects something theory had quietly buried: the need for general principles on which to build, whatever direction you choose to explore. A recent Freestyle Chess commentary captures this beautifully — Vincent Keymer and Peter Leko’s “laws” for 960 openings are not memorized lines at all, but rediscovered first principles: watch for early tricks; exploit the one-move fianchetto when a bishop starts in the corner; never lock your queen in behind a fixed pawn phalanx; play for harmony so every piece works together. These are the very principles classical opening theory was originally derived from, before memorization made it unnecessary to think about them at the board. As GM Levon Aronian put it: “Chess960 is healthy and good for your chess… it really improves your chess vision.”
5. A Candid Word on the Freestyle Boom
It would be dishonest not to address the obvious. Chess960 has enjoyed a real surge lately, driven by Magnus Carlsen’s open embrace of the format and Jan Henrik Büttner’s substantial sponsorship of the Freestyle Chess organization, which has drawn most of the world’s top players into its orbit. Carlsen became the FIDE World Fischer Random champion in 2026, following Wesley So (2019) and Hikaru Nakamura (2022).
Let me be both diplomatic and candid. There is nothing remotely wrong with professionals doing something for money — in Chess960, in coaching, in any event. They are professionals, not amateurs. But I would gently push back on a lazy inference: the fact that nearly every elite player joined the Freestyle World Cup is not proof that they all feel about the variant the way Fischer did, or the way Carlsen seems to. Fischer and Carlsen may well be genuine exceptions — true believers. What I would find genuinely interesting is to know, player by player, who among the world’s best loves Chess960 as much as classical chess, who loves it more, who loves it less — and, in each case, why. That is a far more revealing question than a sponsor’s prize fund can answer.
6. Why LSS Is Betting on the 960 Mind
So where does this leave LSS Daily? Firmly committed. We are not a research institute; we are a sports platform where opponents fight to win. The fact that Chess960 produces fewer dead draws matters to us precisely because of being a sports platform in the first place. But it is not the only reason.
We believe there are tremendous cognitive lessons in a game that compels players to think from first principles rather than recall — the kind of mental flexibility that has value far beyond the sixty-four squares. We will not forsake regular chess; it remains the heart of LSS, and Daily Chess is our home format for serious players who want depth without a six-hour clock. But over the coming years we will deliberately increase the number of Chess960 competitions, and the prizes within them. We will reward the genuine multi-talent through mixed rankings that combine standard and 960 performance. And, because the existing 960 databases are thin and poorly organized, we intend to build a proper one — a resource worth having.
Two practical notes for the purists. First, we run most of our Chess960 events as paired color-reversed games — opponents contest the same starting position once with White and once with Black — so that any inherent imbalance in a given position is shared equally. Which brings me to the second point: the recent observation that not all 960 positions are equally balanced or complex is, frankly, not an objection at all. Of course they aren’t — and we know that, which is exactly why we pair them. Standard chess is no different. Ask any specialist about the gulf between a quiet Berlin Defense in the Ruy Lopez and the fireworks of a Sämisch King’s Indian or a King’s Gambit. Some positions are simply more exciting than others. That has never been a reason to abandon a format. It is a reason to design tournaments intelligently.
To every player who loves Chess960 and is not yet with us: come and play. Many of our members are already active across several platforms — FIDE for over-the-board, ICCF for slow correspondence — and choose LSS as their home for the faster, sharper world of Daily Chess. There is room at our board for the 960 mind, however you spend the rest of your chess life.
The Test That Cannot Be Memorized
Twenty years and 452,542 games taught us that engines did not kill chess — they raised its floor. But they also taught us that the standard game, at the very top, is approaching a solved, drawn calm. Chess960 is the most robust answer we have found, not because it is easier — it is unquestionably harder — but because it relocates the contest from the library back to the mind.
It is harder for everyone, and it humbles the strong in a particular way: a top player’s gut-feel and recognition are usually his greatest assets, and 960 quietly switches them off. Weaker players are caught by surprise more often; stronger players feel the shock of a blunder more deeply — though the very best, as in any sport, are also the best at recovering from it.
Fischer wanted to give the game back its “live creative spark.” On the evidence of two decades, he succeeded. Chess960 is not the death of theory; it is the resurrection of understanding. And for the next twenty years at LSS, that is exactly the kind of mind we intend to celebrate, reward, and grow.
The position is one you have never seen. The clock is yours. Now — what do you understand?