Investigating Experiential Effects in Online Chess using a Hierarchical Bayesian Analysis

Adam Gee, Sydney O. Seese, James P. Curley, Owen G. Ward · 2025

Read Investigating Experiential Effects in Online Chess using a Hierarchical Bayesian Analysis by Adam Gee, Sydney O. Seese, James P. Curley, Owen G. Ward (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-21713.txt.

Introduction

Gee and co-authors present a 2025 study on whether winning or losing streaks bend a chess player's next results. Sports lore claims such winner-loser effects are real, but data are usually thin. Online chess offers millions of games to test the claim. The team built a layered Bayesian model that lets each player carry a personal effect size.

Across the whole pool, they found little sign of a shared streak effect. A few single players did show signs in their own records. Streak data twist through time, so the authors checked their model with care. The paper is a cautious verdict: momentum in chess looks personal, not universal.

Catalog record

CREATORAdam Gee, Sydney O. Seese, James P. Curley, Owen G. Ward
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE44 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-21713

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Investigating Experiential Effects in Online Chess using a Hierarchical Bayesian Analysis Adam Gee, Sydney O. Seese, James P. Curley, Owen G. Ward Abstract The presence or absence of winner-loser effects is a widely discussed phenomenon across both sports and psychology research. Investigation of such effects is often hampered by the limited availability of data. Online chess has exploded in popularity in recent years and provides vast amounts of data which can be used to explore this question. With a hierarchical Bayesian regression model, we carefully investigate the presence of such experiential effects in online chess. Using a large quantity of online chess data, we see little evidence for experiential effects that are consistent across all players, with some individual players showing some evidence for such effects. Given the challenging temporal nature of this data, we discuss several methods for assessing the suitability of our model and carefully check its validity. 1 Introduction It is rare that the success or failure of an individual competitor or sports team is not discussed in the context of their recent performance. The idea of momentum in competition, broadly defined, has been examined repeatedly in the scientific literature. In sports analytics this is often investigated in the context of the “hot hand” phenomenon, where players experience streaks of success that appear unlikely to occur due to chance [ 34 , 38 , 40 ] . Similarly, other studies have directly considered the question of momentum in sport, both within and between games [ 20 , 46 ] . A large body of research in the psychology and animal behavior literature explores a specific formulation of this concept of momentum. It is commonly referred to as winner-loser or experiential effects 1 1 1 We will use these two names interchangeably throughout this work. , defined as success (or failure) in recent competition leading to an increased probability of success (or failure) in future competition [ 12 ] . The presence of these winner-loser effects have been hypothesized to facilitate the formation of stable dominance hierarchies [ 12 , 27 ] . In animals, loser effects appear to consistently be stronger than winner effects (e.g. sticklebacks [ 3 ] ; crayfish [ 6 ] ; copperhead snakes [ 42 ] ). In humans, several studies have revealed some evidence for winner and loser effects in competitive sports including tennis, football and judo [ 8 , 20 , 21 , 39 ] . For example, winners of tennis tie-breaks are more likely to win subsequent sets when players are evenly matched [ 39 ] . Winner and loser effects have also been experimentally demonstrated in humans playing competitive video games [ 44 ] . However, there is still considerable discussion about the existence and consistency of these effects across individuals, contexts and species. Indeed, there are still several outstanding key questions regarding the presence of such experiential effects, including – Are winner/loser effects strongest when considering the immediate previous outcome or several previous outcomes? – Do the size of winner/loser effects vary across rating levels such as between players/teams of stronger versus lesser ability? – Are winner/loser effects stable across time? – Do winner/loser effects vary within individuals across contexts such as across different formats of competitions? One of the main challenges posed by these questions is the availability of appropriate data to investigate these phenomena. Online chess has enjoyed widespread popularity in recent years, with millions of regularly active users on several major sites. The most common format of chess played on these sites is “Standard” rated games. These games are often short “bullet” games, lasting less than three minutes. It is common for players to play many such games in succession. In these games, players are randomly paired with another user currently wishing to play at the same time and of similar estimated ability. The color that each player is assigned (White or Black) is randomly chosen. Games of this form provide an excellent setting to investigate the presence of psychological momentum and experiential effects in such sports and can be used to consider the key questions posed above. Any individual plays a wide range of players, who may utilise different styles and strategies of chess. The random assignment of opponents and individual game settings, coupled with the vast amount of games many users play, makes online chess an excellent setting to investigate the presence of winner-loser effects in a controlled environment. In this paper we use a Bayesian hierarchical logistic regression model to investigate the presence of winner-loser effects in online chess, using publicly available data. We carefully motivate and describe the model chosen. Then, we use this model to examine if and when these effects occur, and how they vary across player ability and game format. While we do not believe this model is a complete representation of the process underlying these games, we use multiple model checking procedures to ensure our model is reasonable. We find little evidence for strong persistent winner-loser effects, with this conclusion consistent across the range of players and abilities analysed. Individuals appear to show some variability, and there is some evidence that some players demonstrate such effects. However, these effects do not appear to be systematic. This article is organized as follows. We first provide a brief history of the study of these effects across disciplines in Section 1.1 . In Section 2 we describe in detail the data we use here, highlighting key properties which make it well suited for the study of winner-loser effects. Section 3 gives the proposed hierarchical Bayesian regression model, detailing the model and prior specification. In Section 4 we fit the proposed model, highlighting the results for two cohorts of chess players of differing abilities. Section 5 provides a detailed discussion of several procedures used to ensure our model is able to correctly capture the temporal structure present in such data. Finally, in Section 6 we summarise the results of this model and highlight important considerations for future work. 1.1 A history of the study of experiential effects in human and animal competition Winner-loser effects have consistently been demonstrated across species engaged in various competitive interactions [ 48 , 43 , 28 , 35 , 25 , 31 , 36 , 1 , 26 , 37 , 4 , 13 , 47 , 30 ] . The most common experimental design utilizes a series of dyadic matchups whereby individuals are induced to win or lose against weaker or stronger opponents and are then evaluated in subsequent competitions. These studies have demonstrated that loser effects are typically more long lasting than winner effects [ 48 , 28 , 1 , 47 , 30 ] . The likelihood of winning subsequent encounters has also been shown to increase following unresolved aggressive encounters (e.g. East African cichlid fish [ 10 ] ) or even following recent losses (olive fruit fly [ 5 ] ). Further, variation in competitive contexts or individual variability in skill level of fighting [ 7 ] , growth rate [ 33 ] , age [ 16 ] , and life experience with fighting [ 29 ] can all influence the strength of winner-loser effects. Functionally, loser effects may serve to reduce

Read the complete text