Pre-evaluated chess position and its features

One of the hypothesis underlying this project is that the evaluation of chess positions can be improved by systematically analyzing various board features, such as material balance, mobility, central control, king safety, and piece connectivity. These features are extracted from 200,000 chess positions represented in the FEN (Forsyth–Edwards Notation) format, with each position evaluated using predefined heuristics.

The dataset consists of 200,000 chess positions, each represented by a FEN string along with a pre-calculated evaluation. The FEN strings describe the positions of all pieces on the board, while the evaluations are numerical values indicating the relative advantage of either side. The data was gathered from real games from https://github.com/r2dev2/ChessDataContributor

To interpret this data, several metrics are computed for each position:

Material count: This measures the material difference between the two sides by assigning weights to different pieces (e.g., pawns, knights, bishops, rooks, and queens). Positive values indicate a material advantage for white, while negative values favor black.
Total material: This metric sums the total material on the board, regardless of color.
Mobility: This reflects the number of legal moves available to each side.
Central control: Central squares attacks (e4, d4, e5, d5) are critical in chess.
King safety: This examines the vulnerability of each king by evaluating whether the squares surrounding the king are under attack.
Connectivity: This assesses how well the pieces support each other.
Position of the pieces: the squares where they are and the average evaluation when they are on these squares, creating a sort of heat map.
There is also a graph representation of the raw board, which indicates the list of squares attacked by another square. Depending on the piece on the square and the overall board configuration, a square can "point" to many others, which allow us to calculate the above features.

These features are then stored in a relational database for future analysis, enabling more advanced techniques such as an evaluation function to predict how good a position is.

The notable findings from this dataset suggest that a holistic approach to chess evaluation that considers both static (material) and dynamic (mobility, connectivity) factors provides a more comprehensive picture of a position.

The data can be interpreted by analyzing the relationships between these metrics and the final evaluations. For example, a high material count often correlates with a favorable evaluation, but in some cases, superior king safety or central control can compensate for a material deficit.

In conclusion, this study provides valuable insights into the evaluation of chess positions by analyzing key positional features.