How to improve at chess: Uncovering insights using regression analysis

Chess is an intellectually challenging game which has grown in popularity recently. The game benefits young players by improving cognitive and problem-solving skills, memory, concentration, and math/reading skills. It also benefits older players by slowing down mental decline. Since chess is centuries old, there exists a lot of advice, rooted in conventional wisdom, for improving performance. For advanced beginners (players with International Chess Federation ratings between 1,500 to 1,700), such advice advocates for aggressive practice techniques either in breadth or depth of play. Specifically, five factors characterize these techniques: 1) number of daily games played, 2) the proportion of higher-rated opponents played, 3) the number of unique openings played, 4) the number of unique opponents played and 5) the extent to which players resign and lose games. We hypothesized that advanced beginners would improve faster at chess when they follow these aggressive practice techniques. We tested these factors using a data set comprised of 162,000+ games played by 1,250+ players on the Lichess platform over a six-month period. We used the difference in rating points as a proxy for improvement and used simple and multiple linear regression analyses. Our analysis showed that such techniques had a mixed impact on individuals’ ratings. Specifically, playing more daily games, playing more higher-rated opponents, and losing through fewer resignations were inversely related to improvement. Conversely, having greater depth through fewer unique openings seemed to help, while the number of unique opponents was not a significant factor. Coaches and players can use these findings to tailor their practice programs with a more balanced approach.