Chess home page  10x8 Chess home page  Live 10x8 Games  Gothic Chess is licenced under U.S. Patent #6,481,716 by Ed Trice

The Carreblanca rating list

The aim of this list is to rate as many 10x8 engines as possible, according to their general capability to play 10x8 Chess. The testing conditions will thus not focus on one particular variant, such as Gothic of Capablanca, but test the engines on a mix of opening positions. This composite nature of the list is expressed in its name: "Carreblanca" is a composite of the surnames of Pietro Carrera (who invented 10x8 Chess wit Archbishop and Chancellor) and Jose Raul Capablanca (who made it popular).

The possibility to start from many different opening positions has the practical advantage that games between the same engines will automatically be different, even if the engines play deterministic. In normal Chess this problem is circumvented by using (often external) opening books, randomly picking the opening lines. In 10x8 Chess this procedure would be less satisfactory, as opening theory is not really well developed, and most engines do not have opening books at all, or extremely limited ones.

The initial positions

The opening positions used in making the Carreblanca list are all positions from Capablanca Random Chess. This means they start with a closed line of pawns on 2nd and 7th rank, and have all other pieces on 1st and 8th rank, for white and black, respectively. And the array of pieces on the 8th rank reflects that on the 1st rank, to make sure the positions are not biased, except for the unavoidable first-move advantage. The latter is eliminated by always playing the games in pairs, with reversed colors.

Not any position from CRC is suitable, though. We limit ourselves to positions that have the Kings in the f-file, and the Rooks in the corners, to make sure that normal castling rules apply. In addition, to balance the piece power over the board, we require there be at least one Bishop and one Knight on each side of the King. Like in CRC, the Bishops will always be on oppositely colored squares. This leaves 216 different opening positions, which can be played twice.

The choice of opening positions does not need to be entirely random; we assume that none of the positions is so wildly unbalanced that the procedure of playing the games in reversed-color pairs will not be able to cope with the unbalance due to the leading move. If statistics would suggest that such unbalanced positions exist, we will eliminate those as possible initial positions. Preferred positions for testing will be the opening arrays of Capablanca Chess, Embassy Chess (mirrored to have normal castling), Bird Chess, Carrera Chess and Gothic Chess. It will not be held against an engine if it is not capable of playing all positions; engines somehow limited in their ability to load positions (for technical or legal reasons) will simply be tested with positions that they can handle.

Testing conditions

The conditions will be corrected for hardware as much as possible, according to the bench-marking procedure developed by CCRL. We will maintain separate lists for blitz and normal play.

Time-odds games

Initially, the number of engines will be so small, and the engines that exist will be spread out over such a large rating range, that it will be difficult to find enough different opponents against which testing will be meaningful. As a temprary solution to this problem, we will include a number of time-handicapped versions of the stronger engines in the list. By giving these 'pseudo-engines' time odds of 3, 10, 30 or 100, it should be possible to reduce their strength by upto about 500 Elo points. (Typically a factor 2 time odds reduces the rating by 50-70 Elo points.) This will provide a larger variety of possible opponents in any region of the rating scale. In addition it is interesting in itself to determine how much exactly the playing strength varies with search time. As the time-odds games will take much less time, this is not very costly.