Have there been changes in the performance of chess players since the onset of the pandemic?

The Corona And Chess Project

This project intends to analyse whether the performance and performance changes and variability of online chess players have changed since the Corona Pandemic started. To do so millions of matches of many thousand online Chess players will be evaluated. The first step of this project was creating a robust dataset of evaluated chess positions of the first 10 Moves (20 Plys) of each game, which will speed up later analysis of all matches as most of those positions arise multiple times in different matches and by storing the evaluation just once we can save computation time. This Dataset, which is not yet fully validated, is publically available via this link:

Opening Moves (20 Ply) Dataset

Matches of Playersample (1000 Players)

FEN with Evals - Chunk 1

As Second Step the Blitz Matches of 1000 Players who played almost every Quarter (at least 24 of 26) between Q1 2017 and Q2 2023 will be evaluated. This will be separated into Chunks of 100 Players. We are currently at Chunk 2 of 10.

Total Positions Evaluated so far: 2.268.000 of 43.715.046 of the current Dataset (updated 2024-03-11 11:30 UTC)

Currently validating the Results and working on some improvements, once that is finished the dataset will be published and the next dataset will be analysed.

Status: Idle

Finished Batches: 0

Time required for last Batch: -



Pressing the button runs a Chess Engine called Stockfish in your Browser using 1 CPU Core, about 300MB RAM and it downloads around 40Mb of Data which might be cached when revisiting the page depending on your browser and its settings.* Per Batch 1000 Chess Positions in Forsyth–Edwards Notation (FEN) are downloaded from a database, evaluated on your Computer and the results uploaded to the database again. Stockfish is configured for this project in a way which provides deterministic and reproducible evaluations of the positions. While there are existing evaluations of many starting positions they are not done with consistent node count which is very important to get consistent results of the performance of players and the engine configurations are also not also not consistent which results in non-reproducible results

*Which is probably less CPU, RAM and Data than a lot of websites use just for advertisments without doing any potentially useful computation ;)