Move-by-move dynamics of the advantage in chess matches reveals population-level learning of the game

The complexity of chess matches has attracted broad interest since its invention. This complexity and the availability of large number of recorded matches make chess an ideal model systems for the study of population-level learning of a complex system. We systematically investigate the move-by-move dynamics of the white player’s advantage from over seventy thousand high level chess matches spanning over 150 years. We find that the average advantage of the white player is positive and that it has been increasing over time. Currently, the average advantage of the white player is ~0.17 pawns but it is exponentially approaching a value of 0.23 pawns with a characteristic time scale of 67 years. We also study the diffusion of the move dependence of the white player’s advantage and find that it is non-Gaussian, has long-ranged anti-correlations and that after an initial period with no diffusion it becomes super-diffusive. We find that the duration of the non-diffusive period, corresponding to the opening stage of a match, is increasing in length and exponentially approaching a value of 15.6 moves with a characteristic time scale of 130 years. We interpret these two trends as a resulting from learning of the features of the game. Additionally, we find that the exponent {\alpha} characterizing the super-diffusive regime is increasing toward a value of 1.9, close to the ballistic regime. We suggest that this trend is due to the increased broadening of the range of abilities of chess players participating in major tournaments.

💡 Research Summary

**
The authors investigate how the white player’s advantage evolves throughout a chess game and how this evolution has changed over the past 150 years. Using a database of more than 70,000 high‑level games (mainly FIDE‑rated tournaments and World Championships), they extract the engine‑based evaluation of the position after each move, expressed in pawn units. A positive value indicates that white is ahead, a negative value that black is ahead.

First, they compute the average advantage ⟨A(m)⟩ as a function of move number m and then examine its long‑term trend by averaging over calendar years. The mean advantage was essentially zero in the early 20th century, but it has risen steadily to about 0.17 pawns today. Fitting an exponential approach to a plateau yields a limiting value A∞≈0.23 pawns with a characteristic time τ≈67 years. This suggests that the accumulated opening theory and the widespread use of databases have gradually increased the intrinsic benefit of playing white.

Second, the authors treat the temporal fluctuations of the advantage as a diffusion process. They calculate the mean‑square increment ⟨ΔA²(m)⟩, where ΔA(m)=A(m)−A(m−1). For the first few moves (≈0–10) the variance is practically zero, indicating a non‑diffusive regime that corresponds to the opening phase. Remarkably, the length of this non‑diffusive interval has been growing; an exponential fit predicts a saturation at about 15.6 moves with a time constant of roughly 130 years. After the opening, the variance follows a power law ⟨ΔA²⟩∝m^α. The exponent α started near 1.4 in older data and has risen to about 1.9 in recent games, approaching the ballistic limit (α = 2). This super‑diffusive behaviour reflects increasingly abrupt swings in advantage during the middle‑ and end‑game.

Statistical analysis shows that the distribution of advantage increments is heavy‑tailed rather than Gaussian, and the autocorrelation function exhibits long‑range anti‑correlations, meaning that a gain for white in one move tends to be partially offset in subsequent moves.

The authors interpret the two observed trends as signatures of population‑level learning and diversification. The elongation of the opening phase and the rise of the baseline advantage are attributed to the collective accumulation of opening knowledge—players now follow well‑studied lines that keep the position stable for longer. The increase of α toward the ballistic regime is linked to a broader spectrum of player strengths participating in elite tournaments; stronger players can create decisive imbalances more quickly, while weaker participants may be more vulnerable to large swings.

Limitations include reliance on a single engine’s evaluation (which can change with software updates and hardware), a dataset confined to top‑tier events (thus not representing amateur play), and the use of relatively simple exponential and power‑law models to describe complex strategic dynamics. Future work could incorporate multi‑engine consensus evaluations, extend the analysis to lower‑rated games, and explore agent‑based models that capture the interplay between opening preparation and tactical creativity.

Overall, the paper demonstrates how concepts from statistical physics—diffusion, scaling exponents, and non‑Gaussian statistics—can be applied to a classic intellectual sport to reveal measurable signatures of collective learning and evolving competitive dynamics.