It is possible we might have engine based instead of tablebase based 8 and 9 piece in the future, but it would need to be a different training system that was more tolerant to not having 100% accurate results. ![]() The total number of positions in the tables more than 500 trillions (500 000 billions). The tablebases contain the exact evaluations (draw or moves to mate) for all positions with no more than 7 pieces on the board. 8 piece tablebases are a long way off unless there is a big breakthrough in drive capacity in the next few years, as 7 piece tablebases already take over 18 TB of data (and that is without depth to mate information), and I've seen estimates that 8 pieces would be around 100 times that, and currently nobody has got anywhere near producing a set (some may have produced individual single endgame tablebases for a specific set of pieces with 8 pieces, but there is definitely no full set out there). Generation of tablebases is an important step as in chess theory and practice so in improving of retrograde analysis algorithm. ![]() But for the OTB player, who wants to improve I believe they are next to useless. ![]() We rely on tablebases to determine which moves are winning/drawing/losing, and while engines are ok at this, it can take a long time to get an accurate result, and even with a long think time, they can still get it wrong (and unlike tactics we can't pre-calculate the engine analysis because there are a massive number of possible paths through a problem, especially in practice and benchmark modes). The 7 piece tablebases included in Rybka Aquarium are also useful (necessary) for endgame study composers who have to make sure that their compositions are correct and don't have extra solutions.
0 Comments
Leave a Reply. |