-

Empirical Advantages in Chess Openings

This was a personal project developed as a capstone for the Google Professional Certificate in Data Analytics.

The scenario for the project was as follows: a fictional company, Chess-360, has an mobile App for learning chess. They are searching for new ways to approach the teaching chess openings.
In this project we explore chess databases looking for empirical data where certain oppenings appears to offer sensible advantages.

A total of 6 millions of games from 3 different datasets were included. Filtering criteria were applied to minimize odds that the advantage was based on skill or spurious , e.g. only include skilled players (ELO>2000) with small differences in skill (ELO difference <50). The data was also categorized for its game formats and winning conditions.

Considering the best winning rates for the white pieces, black pieces and termination type, a set of 24 openings were selected. Among these 24 openings only one is included in the top 10 most common chess openings.

It was found that certain openings can be exploited or avoided to increase the winning chances by up to 25% against an even skilled player.
Surprisingly, the highest advantages were found in time-forfeit termination games, which points to a new strategy to be implemented.
It was suggested that these openings be considered for inclusion in the company App.