How you practise chess matters far more than how much. In a six-month study of 44,213 online players, time spent on structured training raised ratings about three and a half times faster than time spent playing games, even though players devoted more than 90% of their hours to games.
The study was led by Daniel A. Southwick of the University of Pennsylvania and the University of Utah, with Kyle W. Harwell of the U.S. Army Research Institute for the Behavioral and Social Sciences and Garrett Wright, Joseph A. Olsen and Benjamin M. Ogles of Brigham Young University. It was published in Psychological Science on June 23, 2026.
Settling an old argument with logged data
For decades, psychologists have argued over “deliberate practice,” the idea, set out by Ericsson and colleagues in 1993, that skill grows fastest through focused training with clear goals, immediate feedback and repetition. Critics say its influence has been overstated, partly because many studies relied on people recalling how they practised.
This study used no recall at all. Chess.com shared time-stamped records for users who registered between March 1 and March 6, 2024, established a stable rating and stayed active for six months. Starting ratings ranged from 100 to 2,505.
The researchers tracked four activities: playing rapid games, reviewing past games with the site’s engine-assisted tool, watching video lessons, and solving tactics puzzles. Before analysing the data, they preregistered reviews, lessons and puzzles as practice that can meet deliberate practice principles, and games as practice that cannot.
Reviews and lessons pay off, puzzles do not
An hour of deliberate practice activities was linked to a 3.11-point rise in rapid rating, compared with 0.86 points for an hour of play. That is the 3.61-fold advantage in learning efficiency reported in the paper.
Players nonetheless spent an average of 90.39% of their time playing. Reviews took 4.61% of their time, puzzles 4.01% and lessons just 0.99%.
Broken down by activity, the gaps were large. Each hour of game review was associated with 4.41 rating points and each hour of video lessons with 5.21 points. Puzzles, at 0.73 points per hour, did no better than simply playing.
The authors suggest one possible reason. Game reviews target a player’s own mistakes and let them retry the position, while some speed-based puzzle modes may reward quickly solving puzzles a player already understands. They stress that this explanation is speculative and untested in their data.
Strong players train differently
The benefits were biggest for beginners. Among players rated below 400, lessons were linked to 17.81 points per hour and reviews to 6.46, against 1.40 for games. Among advanced players rated 1,000 and above, only game reviews remained significantly more efficient than play.
The best players also spent their time differently. Expert-level players rated 2,000 or more spent 14.31% of their time reviewing games, compared with 3.77% for the lowest-rated group.
The authors argue this could help explain why players who start ahead tend to pull further ahead, a pattern often credited to talent. In their words, the gap can widen “not because of talent, but because of training choices.”
Caveats from the authors
The data could not capture off-platform training such as club games or private coaching, and contained no measures of ability or motivation. The design is observational, the sample is mostly recreational players, and anyone who quit Chess.com during the six months was not included.
One practical note stood out to the authors: the improvement they measured required no human coach. The fastest-learning players were using the site’s automated review tool on their own games.
Study Details:
- Title: Not All Practice Is Created Equal: Longitudinal Evidence From Over 40,000 Chess Players
- Authors: Daniel A. Southwick, Kyle W. Harwell, Garrett Wright, Joseph A. Olsen, Benjamin M. Ogles
- Journal: Psychological Science
- Publication Date: June 23, 2026
- DOI: 10.1177/09567976261452568