Various statistics

• Pairings
• Draw rate
• ECO statistics
• Opening statistics
• Opening positions (sorted by move sequence)
• Opening positions (sorted by white advantage)
• Game length
• Releases per year

Ponder hit

Ponder hit stats show the proportion of correctly predicted moves in a match of two engines. Book moves and forced moves are ignored (or in fact any moves made in 0 seconds), as well as all moves with abs(evaluation) of 9 pawns or more (this takes care of tablebase and mating lines). Also drawn games are ignored entirely.

Most similar and most different pairs

• List of most similar pairs
• List of most similar pairs (different families only)
• List of most different pairs
• List of most different pairs (same families only)

Ponder hit cross-tables

• Matrix for best versions, sorted by ELO
• Matrix for best versions, sorted by name
  
  

Evaluation difference

Average difference in evaluation on consecutive moves is compared for each pair of engines that played a match. Only expected moves are used for this.

Example: For this sequence of moves: "1.e4 {+0.15} c5 {+0.08} 2.Nf3 {+0.25}" the evaluation difference will be computed as (abs(0.08-0.15) + abs(0.25-0.08))/2 = 0.12 (in pawns).

Most close and most different pairs

• List of most similar pairs
• List of most similar pairs (different families only)
• List of most different pairs
• List of most different pairs (same families only)

Evaluation difference cross-tables

• Matrix for best versions, sorted by ELO
• Matrix for best versions, sorted by name