dtc nalimov 3 & 4 man databases

rats · Post by **rats** » Wed Nov 29, 2006 8:06 pm

Hi folks,

I'm a non too strong player currently trying to get a full grasp of QvR. I'm working from Averbakh's book and recently discovered Wilhelm & Nalimov tablebases. I was hoping to use these to check out various sequences I've been working on to fill in some of the gaps in the book and also just to practice, but all too often it just makes a couple of moves and then gives me the rook when it doesn't need to. I'm assuming this is because I downloaded the DTM 3 & 4 man databases when what I really need is the DTC versions.

Does anybody know if these are available (free) anywhere? Hope you can help.

Post by **Kirill Kryukov** » Thu Nov 30, 2006 12:48 pm

Hi rats!

DTM tablebases should do just fine. I guess it is giving away the rook when there is no difference either way. I mean there can be positions where giving away the rook does not lose faster than other moves. Can you post the example position where it happens?

guyhaw · Post by **guyhaw** » Thu Nov 30, 2006 3:55 pm

I have the DTC EGT for KQKR, but doubt that this is going to solve your problem. They are 1MB in total, so email me via http://www.tinyurl.com/law6k if you want still want them.
Whatever 'it' is that is giving away the bR 'unnecessarily', it sounds like 'it' is not playing correctly. WILHELM will give you DTM-minimaxing lines, or DTC-minimaxing lines if you have DTC EGTs.
Most engines, when given DTC EGTs, will walk up to a capture but then not effect it. This is because they think the DTC depth is a DTM depth, and since it gets greater after the capture, they don't opt for the capture.
g

rats · Post by **rats** » Thu Nov 30, 2006 9:23 pm

Can you post the example position where it happens?

For instance in position A attached, Wilhelm (black) with my DTM database plays Kf4 allowing an immediate fork, but I would rather he played to h6. (This is a continuation from position B which is one of the positions from Averbakh's book shifted two positions to the right, so the black move corresponding to Ra2 isn't considered in the book.) The reason, of course, is that he sees the first move as mate in 11 and the second as mate in 10, whereas from my point of view the ending I'm looking at is over in 2 moves in the former case but up to 5 in the latter.

Actually this happens quite a lot, that's just one example.

guyhaw · Post by **guyhaw** » Thu Nov 30, 2006 11:49 pm

Your position A is such that the DTC-optimal move is not the same as either DTM-optimal move. One might argue that, from a chessic point of view, a DTC-optimal move here would be better than a DTM-optimal move.
Remember that the metrics are not calculating what asks the opponent the biggest chessic question - they are just giving the metric-optimal moves.
The Tamplin/H study of DTZ and DTZ50 EGTs searched for positions with maximal DTZ50-DTZ difference, and thereby found positions where none of DTC-, DTM- or DTZ-optimal play find the move that DTZ50 finds.
g

rats · Post by **rats** » Fri Dec 01, 2006 8:44 am

Exactly.

Since my aim is to be able to play the ending from any position it's the chessic moves I'm interested in.

I'm assuming DTC50, DTC, DTZ or DTZ50 would all be the same for this ending.

rats

guyhaw · Post by **guyhaw** » Sat Dec 02, 2006 1:10 am

DTC == DTZ when there are no pawns on the board.
DTxk = DTx when no phase need be longer than 50 moves: certainly KQKR is one of these endgames.

There are plenty of chess studies where Black's supposedly 'best defence' move is not metric-optimal. Perhaps too often though, 'Black forces a unique-winning-move reply' is taken as evidence of 'best defence'.
g

CCRL Discussion Board