Beyond the Swiss System — A Manifesto for the Scientific Reform of Chess Tournaments

Rethinking Chess Tournaments

Let us imagine for a moment that the game of chess had been invented only last year. Suppose that its popularity had suddenly exploded, producing large tournaments with hundreds of players that must be completed within a few days. The natural question would arise immediately:

How should such tournaments be organized, and how should a reliable ranking of the players be determined?

A full round-robin tournament would clearly be impossible in large fields. Some form of incomplete competition would therefore be necessary. But if the problem were posed today — in 2026 — would the solution resemble the system currently used in chess?

A problem already studied elsewhere

The organization of competitions under incomplete information is not a new problem. Over the past century, a vast body of scientific knowledge has developed around closely related questions. Relevant fields include:

• statistical ranking models (Zermelo, Bradley–Terry, Davidson models);
• graph theory and ranking on networks;
• experimental design;
• social choice and voting theory;
• robust statistics;
• sports analytics and tournament design.

In all these domains, researchers have developed sophisticated mathematical tools to estimate relative strengths from incomplete comparisons and to design fair and informative competitions.

If a group of modern statisticians and mathematicians were asked today to design a tournament system for chess under these constraints, they would approach the problem using this extensive theoretical toolkit.

Would they end up reinventing the current FIDE Swiss system, together with its well-known tie-break rules?

Almost certainly not.

Structural limitations of the current system

The Swiss system used in chess attempts to solve several problems simultaneously:

• scheduling pairings,
• balancing colours,
• producing a final ranking.

These objectives are combined into a single algorithm based on the accumulation of points (1, ½, 0). When ties occur, a collection of tie-break systems — Buchholz, Sonneborn–Berger, Median Buchholz, Cut-variants, and others — are applied.

These tie-breaks are described in considerable detail in the regulations. Yet, remarkably, their theoretical justification is often weak or entirely absent.

Over time, practical experience has revealed a number of structural issues, including:

• strong dependence on the pairing path;
• sensitivity to results of unrelated games;
• statistical noise in opponent-score tie-breaks;
• colour imbalance affecting outcomes;
• incentives for non-competitive behaviour in the final rounds — and sometimes even in the first round, as players may try to avoid being pushed into a stronger field of opponents.

These phenomena are widely observed in practice, even if they are rarely analyzed in a systematic way.

A modern perspective

A contemporary approach to tournament design would likely begin with a conceptual separation between two distinct problems:

pairing and ranking.

• The pairing algorithm determines which games are played.
• The ranking algorithm estimates the relative strength of the players from the observed results.

In the current chess tradition these two roles are entangled. But there is no mathematical necessity for this.

Instead of embedding all objectives into a single mechanism, one could design pairings that explicitly maximize fairness — for example by balancing colours or avoiding structural biases.

This raises a simple question that surprisingly is almost never asked in the chess world:

Why should players in a tournament receive a different number of games with White and Black?

In modern databases of high-level games, White typically scores several percentage points more than Black.

Yet in most Swiss tournaments with an odd number of rounds — for instance the widely used nine-round format — some players inevitably finish with:

5 games with White 4 games with Black

while others receive:

4 games with White 5 games with Black

This asymmetry is generally accepted as unavoidable. But it is worth reflecting on what this really means. If White has a statistical advantage, then a player receiving five Whites has enjoyed an advantage that another player did not. Conversely, someone receiving five Blacks has been placed at a systematic disadvantage. In other words, the tournament itself introduces an unequal condition between competitors.

Traditionally, this is justified by arguing that such imbalances “average out” over many tournaments. Yet this argument implicitly admits that each individual event contains a structural bias.

One may therefore ask a very natural question:

Why not design tournaments where every player receives exactly the same number of Whites and Blacks?

This would require an even number of rounds and a pairing system explicitly designed to enforce colour symmetry.

Such a condition is difficult to guarantee within the traditional Swiss framework, where pairing, colour allocation, and ranking are all intertwined.

But once pairing and ranking are treated as separate problems, achieving strict colour balance becomes considerably easier.

And with it, tournaments can move closer to a fundamental principle of competitive fairness: all players should compete under the same conditions.

The ranking algorithm should compute rankings using statistically coherent models that take the entire network of games into account. In such a framework, the need for tie-break rules is greatly reduced. Moreover, since every result influences the global estimation of strengths, players cannot rely on a single final-round draw to secure their position in the standings.

Traditional chess tournaments produce many ties because the ranking is based on a coarse scoring system (1, ½, 0). With only a small number of rounds, several players inevitably finish with identical scores. The elaborate hierarchy of tie-break rules used in modern regulations is essentially an attempt to break these artificial ties.

Statistical ranking models operate differently. They estimate player strengths using all available information in the network of games, producing continuous ratings rather than discrete point totals. As a result, exact ties become extremely rare, and the complex machinery of tie-break rules largely disappears.

The proliferation of tie-break rules in chess tournaments — some of them rather exotic — is therefore not a sign of sophistication, but rather a symptom of the limitations of the underlying scoring system.

Instead of questioning the foundations of the system, the typical response has been to introduce yet another tie-break rule — expanding the regulations without addressing the underlying problem.

Moreover, modern Swiss pairing procedures have become so complex that they are effectively executed only by specialized software. Very few arbiters are able to reproduce the full procedure manually without assistance. This illustrates an interesting paradox: a system originally designed for practical simplicity has gradually evolved into a highly technical algorithm whose behaviour is understood mainly through its software implementations.

Tradition versus scientific progress

The persistence of the current system is largely historical. The Swiss system was introduced more than a century ago as a practical method for organizing large tournaments.

At that time, the mathematical and computational tools available today simply did not exist.

Since then, however, the scientific understanding of ranking systems and incomplete competitions has progressed enormously. Yet the chess world has remained remarkably conservative, continuing to rely on tournament mechanisms conceived in a very different scientific era.

Instead of rethinking the underlying structure, the response has often been to introduce incremental modifications: new tie-break variants, minor adjustments, or cosmetic changes.

These adjustments address symptoms rather than causes.

The purpose of this site

This site aims to examine these issues in a systematic way. Over time we will:

• analyze the limitations of traditional tie-break systems;
• show how they can be interpreted as crude approximations of more principled statistical methods;
• explore alternative approaches grounded in modern mathematical theory.

Our goal is not to dismiss the historical achievements of tournament organizers or arbiters. The Swiss system was a brilliant practical invention in its time. But chess deserves tournament formats that reflect the knowledge available today.

A call for dialogue

We hope that players, arbiters, organizers, and researchers will engage with these questions.

Many participants in the chess world are simply unaware that entire areas of modern mathematics address problems very similar to those encountered in tournament ranking.

This site therefore seeks to build a bridge between the academic community and the chess community.

Ultimately, one may hope that the international chess federation will one day assemble an interdisciplinary task force — bringing together mathematicians, statisticians, and computer scientists — to address the structural challenges of tournament design. Not merely cosmetic adjustments, but a serious reconsideration of the foundations.

Chess players deserve tournament systems that reflect the knowledge of the twenty-first century, not the limitations of the nineteenth.

The Swiss system was introduced more than a century ago as a practical method for organizing large tournaments. It was a brilliant solution for its time. Today we have the scientific tools to do better.