CCRL Discussion Board

Mindful that a filename is not a guarantee of anything, and that 'datacomp' is a good check on the file before opening it ...
the Nalimov-published 3-3p EGTs KQNK(B/N/Q/R)P and KRBKNP appear to be available if not in Kirill's index.
This would leave us needing just the 16 3-3p EGTs that Eugene has not published at all.

On the 4-2p front, we appear to lack 10 QxxxKP EGTs to complete the set of 95. These are:
KQB(B/N)KP, KQNNKP, KQR(B/N/R)KP, KRB(B/N)KP, KRR(B/N)KP

g

Hi!

Well, the 33p's you mention are all shared by me, and possibly a few others.
As for the 42p's, I have found some of them, copy+check+hashing is in progress, I will update this message with the ones I found.

Stay tuned!

Well, here is some more stuff for you to download. The first 6 specified classes..........

More to come later................

Here are the last 4 classes mentioned above.

With this, I think I have nothing more to add to the project.

Now who was it that said that the last 15 classes were "one step away from EMule" ?

DW puts the final brick in the wall: all Nalimov-available EGTs sighted at least once. Congratulations.

Afaik, nothing heard from Eugene yet: two courier trips needed.

How shall we mine the data?! g

guyhaw wrote:Mindful that a filename is not a guarantee of anything, and that 'datacomp' is a good check on the file before opening it ...
the Nalimov-published 3-3p EGTs KQNK(B/N/Q/R)P and KRBKNP appear to be available if not in Kirill's index.
This would leave us needing just the 16 3-3p EGTs that Eugene has not published at all.

On the 4-2p front, we appear to lack 10 QxxxKP EGTs to complete the set of 95. These are:
KQB(B/N)KP, KQNNKP, KQR(B/N/R)KP, KRB(B/N)KP, KRR(B/N)KP

g

Any word if he is still working on them? I haven't seen or heard of any new 6men egtbs since Dr. Hyatts site went down a year or so ago. Though someone had mentioned .tbs exist. So not sure how that relates to the actual egtbs.

-Josh

John - I think this has been covered before.

The .tbs files are the stats files for the EGTs: one cannot have a genuine KPPKPP .tbs file unless all the EGTs have been created. I have what I take to be such a file from Eugene. Unfortunately, I do not have Marc B's KPPKPP .tbs file to x-check it with. In fact, I have .tbs files for all but 12 endgames.

I would bet that Eugene is not working on EGTs. Unless he's lost the files or is aware of some data/file-corruption problem, they are done and he's just too busy to deal with sending JT a disc.

g

guyhaw wrote: I would bet that Eugene is not working on EGTs. Unless he's lost the files or is aware of some data/file-corruption problem, they are done and he's just too busy to deal with sending JT a disc.

g

I just read your pdf on 6-men solved interesting read Have any thoughts or comments on mining the egtbs?

Also do you know where I can find a copy of "6-man data and strategy" that you, M Bourzutschky, J Tamplin wrote? btw what is mzugs?

-Josh

I'm constrained by Elsevier's copyright-protection rules not to put the pdf of the MB/JT/GH paper on '6-Man Data and Strategy' on my website, findable via search-query <"Guy Haworth">.

However, I have a pre-publication draft of the content, so if you email me, I can send you that.


'mzugs' are 'mutual zugzwangs' - or positions in which both sides would rather it was the other side's turn to move. Originally, this meant that having the move lost the side to move half a point or a whole point.

Now, with EGTs, it includes cases where DTx is less attractive if one has the move than if one does not have it. Of course, these sets of positions, which some call 'weak mzugs', are DTx-specific, so the DTM-set for an endgame may not be the same as the DTC-set or the DTZ-set.

If there are 'weak mzugs', I guess we call the original value-shedding positions 'strong mzugs'. The rarest, and therefore the most interesting such mzugs are full-point mzugs, fpzs.

Since Ns and Ps tend to feature in fpzs, the cognoscenti like to see fpzs with as few of these as possible. I'd say that the sole KBBNNKQ fpz found by Marc Bourzutschky recently is 'star' at the moment:
8/8/8/8/4q3/2k4N/5B2/N1K2B2: wtm, 0-1 in DTZ/M = 12; btm, 1-0 in DTZ/M = 42.
See http://www.volny.cz/evcomp/tablebase.htm, 12.03.06 for the file.
There are no 3-6-man P-less fpzs.

Usual targets for data-mining are:
- DTx-maximal positions, wtm/btm, 1-0 & 0-1
- subsets of these positions which are both DTx- and DTy-maximal
- mzugs in the various types: =/1-0, 0-1/=, 0-1/1-0
- DTx-maximal mzugs (for 0-1/1-0, add the two depths together)
- rare events such as under-promotion; other examples in the MB/JT/GH paper

g

Thanks Dan! I added your new links to the index page. So, with this addition the 42p collection seems completed? Great work everyone! Let's keep them up now, until they all have 10 sources.

BTW, I checked the availability for the pawnless sets, and found many more sets with 10 sources, and updated the smileys. Now all 33 are super-shared, and many of 42 sets.

guyhaw wrote:John - I think this has been covered before.

The .tbs files are the stats files for the EGTs: one cannot have a genuine KPPKPP .tbs file unless all the EGTs have been created. I have what I take to be such a file from Eugene. Unfortunately, I do not have Marc B's KPPKPP .tbs file to x-check it with. In fact, I have .tbs files for all but 12 endgames.

I would bet that Eugene is not working on EGTs. Unless he's lost the files or is aware of some data/file-corruption problem, they are done and he's just too busy to deal with sending JT a disc.

g

Hi, Guy. If that is the reason, I would be willing to drive up to Seattle, and bring an external USB disk, and do all the work of copying and verifying the data, if someone could contact him and ask if he would be willing. Its a 3-1/2 hour drive for me to Redmond, but I wouldn't mind at all.

kp

Well, it's a fair offer - but someone has to hear from Eugene first.
No-one has heard from Eugene since January, so it's his offline status that is the blocker at the moment.
g

guyhaw wrote:Well, it's a fair offer - but someone has to hear from Eugene first.
No-one has heard from Eugene since January, so it's his offline status that is the blocker at the moment.
g

Are you sure he is okay? Everyone seems to say the same thing, he does not answer e-mails. Is this normal for him, or could it be something more serious? If anyone knows him well enough to call, it might be a good idea, unless you know why he is not responding.

kp

I have asked people, who have a longer association with him than I do, to help allay our concerns that something might be amiss.
If it is, then EGTs are not his priority. I can only suggest patience.
g

Now KK has declared on the last few 'Nalimov published' EGTs in his index, I have sizes:

|{3m to 5m DTM EGTs}| = 007.4 GB
|{3-3 & 4-2 DTM EGTs}| = 220.6 GB
|{3-3p so far DTM EGTs}| = 311.2 GB, maybe going to 408.8 GB
|{4-2p DTM EGTs}| = 512.3 GB

... and not forgetting the 5-1(p) EGTs, that's over a TB.

g

guyhaw wrote:Now KK has declared on the last few 'Nalimov published' EGTs in his index, I have sizes:

|{3m to 5m DTM EGTs}| = 007.4 GB
|{3-3 & 4-2 DTM EGTs}| = 220.6 GB
|{3-3p so far DTM EGTs}| = 311.2 GB, maybe going to 408.8 GB
|{4-2p DTM EGTs}| = 512.3 GB

... and not forgetting the 5-1(p) EGTs, that's over a TB.

g

You're right, I forgot to update the total size. 311 GB.

BTW, thanks to everyone who shares files, and to Dan and Renze for the last 3 sets links.

Please note that I edited the FEN for the sole full-point (Type 3) zug in KBBNNKQ:

It should have been ... 8/8/8/8/4q3/2k4N/5B2/N1K2B2: wtm, 0-1 in DTZ/M = 12; btm, 1-0 in DTZ/M = 42.

Marc Bourzutschky remarks that this is notable for being pawnless AND only having two Ns: see EG Vol 11-162 (2006) position '1220' on p499.

g