Greetings,
I asked this ages ago on ccc but can't find the responce.
Anyone know what are the various metrics for egtb and a description on each?
-Josh
egtb metrics
-
guyhaw
- Posts: 489
- Joined: Sat Jan 21, 2006 10:43 am
- Sign-up code: 10159
- Location: Reading, UK
- Contact:
Endgame metrics
The metrics can all be given acronyms DTx where 'DT' means 'depth to (some event)' and x symbolises that event.
It is assumed that the side with the win is minimising depth, i.e. the number of moves, to this event and the side facing the loss is maximising depth to that event.
Thus:
DTM == DT Mate
DTC == DT C(onversion), i.e. to mate and/or change of force
DTZ == DT (move-count) Z(eroing move),
i.e. to P-push, conversion and/or mate
DTZk == Depth to (move-count) Z(eroing move) given that a k-move draw-claim rule is in place. Thus, today, DTZ50 would be a useful metric to use in an EGT.
One could also conceive of metrics like DTCk and DTMk but maybe one would be happy to go with DTZk.
DTZ leads to the smallest EGTs as the data, ranging over a smaller range of integers, compresses better. DTM is the metric used in the currently prevailing Nalimov EGTs.
g
It is assumed that the side with the win is minimising depth, i.e. the number of moves, to this event and the side facing the loss is maximising depth to that event.
Thus:
DTM == DT Mate
DTC == DT C(onversion), i.e. to mate and/or change of force
DTZ == DT (move-count) Z(eroing move),
i.e. to P-push, conversion and/or mate
DTZk == Depth to (move-count) Z(eroing move) given that a k-move draw-claim rule is in place. Thus, today, DTZ50 would be a useful metric to use in an EGT.
One could also conceive of metrics like DTCk and DTMk but maybe one would be happy to go with DTZk.
DTZ leads to the smallest EGTs as the data, ranging over a smaller range of integers, compresses better. DTM is the metric used in the currently prevailing Nalimov EGTs.
g
-
guyhaw
- Posts: 489
- Joined: Sat Jan 21, 2006 10:43 am
- Sign-up code: 10159
- Location: Reading, UK
- Contact:
Edwards' Databases
I think Edwards' databases were DTM.
KBBKN and KBNKN were published in the ICCA Journals: I didn't see evidence of others.
Ken Thompson used DTZ for KRPKR and KQPKQ when he was pushed for RAM and did the endgames in slices, each slice corresponding to a subset of positions.
http://www.is.reading.ac.uk/people/G.Ha ... blications has some papers introducing and working with metrics other than DTM. The work was largely enabled by non-DTM programs, first by John Tamplin and then by Marc Bourzutschky.
g
KBBKN and KBNKN were published in the ICCA Journals: I didn't see evidence of others.
Ken Thompson used DTZ for KRPKR and KQPKQ when he was pushed for RAM and did the endgames in slices, each slice corresponding to a subset of positions.
http://www.is.reading.ac.uk/people/G.Ha ... blications has some papers introducing and working with metrics other than DTM. The work was largely enabled by non-DTM programs, first by John Tamplin and then by Marc Bourzutschky.
g