What Your Opening Book Doesn't Know About Your Opponents

Most opening preparation is written for a game that the majority of players will never play. A different approach — one grounded in what opponents at each rating level actually do — may produce better results for the vast majority of chess players.

There is a well-established procedure in chess improvement. A player decides to take openings seriously. They buy books, or subscribe to video courses, or spend several hours downloading database material and studying.
They learn fifteen or twenty moves of a well-recommended system, understand the ideas, recognise the structures. They feel prepared.
Then they play their next game. On move four, their opponent plays something the book never mentioned. Not a novelty — nothing deep or theoretical. Just an imprecise move that grandmasters would never make, and so the author never addressed.
The position on the board looks nothing like what was studied. The preparation was not wasted, exactly, but it was applied to a reality that does not quite exist for non-master levels.
This experience is so common as to be almost universal among club-level players. It is worth examining why it happens, and whether a different approach to opening preparation might produce better results for the players who experience it most often.

The Audience Problem in Opening Literature

Opening books are written by strong players, typically grandmasters or international masters. They are written, largely, about games between strong players. The moves analysed, the variations explored, the mistakes identified — all of it reflects the concerns and tendencies of players operating at a level that represents perhaps two or three percent of the active chess-playing population.
This is not a criticism. Deep, rigorous analysis of high-level games is exactly what strong players need. But it creates an audience mismatch that is rarely acknowledged: the books are consumed primarily by club players, online players, and improving amateurs who face opponents whose patterns of play look nothing like the games in the database.
A player rated 1300 does not face Carlsen's preparation. They face other 1300-rated players — with the specific tendencies, habitual errors, and characteristic move preferences of that particular level. These tendencies are not random. They are consistent, measurable, and quite different from what appears in theoretical literature.
Traditional opening preparation assumes a perfect opponent and asks: "What is the best move in this position?" The more useful question — the one that reflects who you will actually face — is: "Which positions will my opponents at this level actually create, and what is the strongest response to each of them?"

What the Data Shows

Large-scale databases of online games make it possible to examine, with some precision, how opening play actually differs across rating levels. The patterns are striking.
Consider the French Defense after 1.e4 e6 2.d4 d5 3.Nc3 — the most principled White response, played in over 40% of all games at every level. Theory identifies three main Black replies: 3...Nf6 (Classical), 3...Bb4 (Winawer), and 3...dxe4 (Rubinstein). These three moves generate the overwhelming majority of opening literature on the French. A well-prepared White player will have studied all three in considerable depth.
What the literature does not address is a fourth move: 3...c5. This move belongs to a different opening structure entirely — it is the standard response to the Advance Variation (3.e5), not to 3.Nc3. Against 3.Nc3, it is theoretically imprecise. And yet, drawn from a large sample of Lichess games, it is the single most common Black response at the 1400 and 1600 rating levels.
French Defense, 3.Nc3 — Black's third move by rating level (Lichess data)
3...Nf6 — Classical (theoretical main line)42%Lichess 2000+
3...Nf6 — Classical17–23%u1400 → u1800
3...c5 — Advance reflex (theoretically imprecise)7%Lichess 2000+
3...c5 — Advance reflex22–28%u1400 → u1600
3...dxe4 — Rubinstein20–22%stable across all levels
The pattern is instructive. At 1400 and 1600, the move that dominates — at 26–28% — is the one that does not belong to this opening at all. Players at these levels have encountered the Advance Variation more often than the Classical lines, and when they see a White pawn on e4 with d4 on the board, they reach for the response they know: ...c5. They are, in a sense, playing the wrong game.
By 1800, the four main moves are nearly equally distributed, each appearing roughly 22% of the time — the level at which opening knowledge begins to converge toward theory. By 2000+, the Classical and Winawer dominate and 3...c5 has fallen to 7%. The literature's focus is accurate here. But for the 1400 and 1600 player facing this position, preparing only the three theoretical main lines means being unprepared for the most common move on the board.
The Rubinstein (3...dxe4), notably, is the one move that barely changes across all five levels — appearing at 20–22% from 1400 to 2000+. It is the only reply whose frequency is not a function of rating.

A Different Question

Standard opening preparation asks: what is the objectively best move in this position?
A different approach asks: what do opponents at my rating level actually play here — and what is the best response to that?
These questions have different answers. In many positions, they overlap significantly. But in a meaningful number of positions — particularly after the first five or six moves — they diverge. The moves that appear most frequently in practical games at 1400 or 1600 are often not the moves that theoretical literature focuses on, because they are not the moves that strong players consider most interesting or challenging.
This creates an opportunity. If the moves that actually appear in your games can be identified with reasonable accuracy — drawn from a large sample of games at your rating level — then a targeted preparation focused on those moves will be more practically effective than a preparation focused on theoretical depth that your opponents will rarely reach.

The Role of Annotation Level

A related issue in opening preparation is the level at which annotations are formulated. A grandmaster explaining the English Attack in the Sicilian Najdorf will, naturally, use concepts like prophylaxis, long-term pawn weaknesses, and minor piece coordination.
For a player at 1400, these concepts may be genuinely unfamiliar — they can read the words without being able to use them in their own games. The annotation that would actually help them is different in depth, vocabulary, and emphasis.
Rating-calibrated preparation suggests that annotations should be written for the level they serve — more detailed and foundational at lower levels, more concise and assumption-laden at higher levels. This is a simple idea but it requires that preparation material be built around specific rating levels rather than a general audience.

Practical Implications

None of this suggests that traditional opening theory is without value. For players operating at higher levels, where opponents do play the theoretical main lines, deep preparation is essential. The divergence between theoretical play and practical play narrows substantially above 1800 or 1900, and at the top levels, virtually every game follows well-charted theoretical paths for many moves.
But for the large majority of active chess players — those in the 800 to 1800 range — the practical value of deep theoretical preparation may be lower than commonly assumed, and the value of preparation specifically calibrated to what opponents at their level actually play may be higher.
The data to support this kind of calibrated preparation exists and has existed for some time. Online chess platforms generate millions of games each month, and the statistical patterns of opening play at each rating level are identifiable with considerable accuracy. What has been less common is building opening preparation explicitly around this data rather than around the theoretical canon.

One Implementation of This Approach

RBOTChess — the platform the author has built around this methodology — works as follows. For a given opening and rating level, the platform identifies the moves opponents statistically play most frequently at that level, drawn from Lichess data. For each of those moves, the platform presents the strongest available response, verified at full analytical depth and annotated for the specific rating level being studied.
The practical workflow for getting into a new opening has three stages. The first is building the conceptual foundation: a combined strategy and gameplan document that covers the opening's character, the typical structures that arise, the plans available to both sides, and a concrete variation-by-variation account of what to do against each common opponent response. Reading this before touching the tree gives the framework that makes individual moves meaningful rather than arbitrary. When the game goes outside the trained lines — as it inevitably will — this foundation is what allows the player to reason forward rather than guess.
The second stage is the tree study itself. In the tree, the mechanism is specific: an opponent's move — statistically the move a player at your rating level will actually play — is answered by the RBOT move. The RBOT move is not a statistical average or a human guess. It is the move a full-strength engine identifies as the strongest response to what your opponent just played. Realistic on one side, optimal on the other.
This is worth being precise about, because it distinguishes RBOT from analysis tools and opening explorers. RBOT does not present options for you to evaluate. For every move your opponent plays, there is one recommended response — the RBOT move. The decision has already been made. Your job is to learn it, understand why it is correct, and recognise it when it appears on the board. The cognitive work of choosing between alternatives — the work that makes opening preparation overwhelming for most players — has been done in advance. What remains is understanding and repetition.
Other tools in this space train your ability to react to rating-specific moves. RBOT gives you a complete system built to win at your level — the difference between practising reactions and having a prepared answer.
The third element is the piece maps. For each opening and rating level, RBOT has played 200 consecutive winning games against a Maia-rated engine calibrated to the relevant level. The results of those games are analysed to show where each piece type most commonly stood on the board in winning positions, as well as the most frequently occurring pawn structure at that level. Three views are available: a heatmap for individual piece types, a combined view showing peak positions for all pieces simultaneously, and a pawn skeleton showing the typical structural patterns.
The piece maps serve a specific purpose: when the game has gone beyond the trained tree and neither the strategy document nor the gameplan covers the exact position, the piece maps give the player a visual reference for where their pieces should be heading. Good piece placement in the Scandinavian at 1400 looks different from good piece placement at 1800 — the maps make that difference concrete and immediately comprehensible.
This combination — conceptual foundation, statistical realism, engine precision, and visual reference — is what distinguishes the approach from both traditional opening books (which treat both sides as playing optimally) and from tools that simply show you what players at your level play (which tells you about tendencies without telling you what to do about them).

The argument here is not that club players should ignore theory. It is the more modest claim that they might benefit from preparation that takes seriously the question of who they will actually face — and builds from there, rather than from the top of the theoretical pyramid downward.
Whether one calls this approach rating-based preparation, practical opening study, or something else, the underlying logic is the same: preparation that reflects the reality of your games is more useful than preparation that reflects an idealised version of chess you rarely play.

The complete London System — opening tree, strategy and gameplan documents, piece maps, 200 downloadable winning games, and Guess the Move training — is free with registration at rbotchess.com.
Premium subscribers get access to 14 additional openings, selected by statistical popularity across Lichess rating levels.

Stefan Rebner is the developer of RBOTChess (Rating Based Opening Training), a platform for opening preparation built around Lichess game data at specific rating levels.