Playing with Databases

Mar 29, 2011, 10:30 AM |

In correspondence chess, players use books and databases to aid them in the opening, and sometimes in the ending as well. Tablebases, on the other hand, are generally forbidden when engines are not allowed. It's a rare game that reaches a position that can be entered successfully in the Shredder Endgame Database. Moreover, the consensus of most turn-based site arbiters appears to be that doing so is tantamount to engine use.

Computers have "solved chess" when six or fewer pieces remain, and they are hard at work on the seven piece, which might be completed in the next few years. Eight and nine piece solutions are years away, and solving the game from the opening move remains a theoretical pipe dream. Three and four piece tablebases have been included with Fritz software for quite some time, and I believe the five piece are part of the package now. My old notebook computer that I bought in 2001 lacks the five piece because its 20 GB hard drive cannot provide the slightly more than 7 GB of free space that is required. In contrast, 30 MB are sufficient storage space for all three and four piece endings.

Each piece dramatically increases the space needed. The six piece tablebases exceed the capacity of most home computers, as they require an estimated 1.2 terabytes of storage space (see David Kirkby's discussion at ChessDB). When computers finally manage to work out the seven piece endings, how much space will be needed to store the data?

Now, consider the beginning of the game when there are thirty-two pieces on the board. After one move--White and Black--there are four hundred possible positions that can be reached. White can lose by checkmate on the second move eight ways, and can deliver checkmate on the third via 347 unique sequences. By the end of the fourth move (eight plies), there are 84,998,978,956 possible move sequences. Let's round the number to eighty-five billion.

Read the rest at

The full post discusses my decision making process in this game: