7 Man Syzygy and SSD

[MikeB]

MikeB your example position is not good. ALL engines I try play -Qxh7 immediately with only 5 piece syzygy and Fritz even with 4 piece Nalimov! BTW I have found, that testing syzygy benefit (self test) needs 5000 games minimum.

The poster was looking for a possible example and although my engine also found t quickly, it was not on ply 1 like it was with 7 man and my runoff 400 was "just for kicks" as noted - I often why people here are so quick here to say "your example is no good". Was just trying to answer someone's question and you get attacked for it. geeez .... next time you can answer it...

[syzygy]

Would you know the elo gain over if you had those 7 man over 6 man on your set up?

I think the gain even from 6-man is quite small, because it is seldom you go into an ending that can't be solved by search or handled by special case evals. Those endings certainly exist, the question is how often do they occur in real games? I think not often.

But perfect play once a 6-men ending has been reached on the board is only a small part of the story. The phase before it is much more important (especially at longer time controls where it will indeed be rare that the engine can't figure out a 6-men ending on its own).

[MikeB]

I don't seem to be able to find a list of MD5 hashes for 7 piece tables. I'd like to verify my downloads are not corrupt. Could someone point me to a list of hashes?

Thanks
Mike

SHA256

Apparently this file, which was distributed, did not have hash updates to some 7 man egtb re-runs:

Code: Select all

ec4785cf89824483c3eed88de4b8d87983f0520510c65b52ff6144ac55b744dc  KRNPvKQN.rtbz
c3bc270aa634b3e4398001e4b9e98347cfa660798fc22783sha032s1867afb47  KRRNvKRR.rtbz
6bc3fe64f334601e1c28c4770a35c065e0a9f1b736532fb49edd65da417d76ac  KRBBNvKQ.rtbz
275c616256ed173177ff8f68e9119335d88b6a8ef8e47ac8a8ca0b949aaf44fe  KQRPvKRB.rtbw
275d5c38fc52d25c0cd70a4ba095b13311d7bf878f2c3db41c9341d8475d8426  KNNPPvKR.rtbw
5dfd7562750f0b1873d3b91b14aa82c8690504d324a36ff76f3c2f2199dcdee7  KQBNvKQN.rtbw

If you start having engine "crashing" issues with 7 man syzygy, but not when you use 6 man syzygy, it is likely you have a 7 man corrupted file, Running sha256 is a little bit time consuming, but you will find the suspect file when you do the hash compare - even when both files are the same exact size.

[MikeB]

http://talkchess.com/forum3/viewtopic.p ... 17#p769925

Thanks for posting - re-did the 800 MB list based on the Human Occurrence Ranking (HOR) :

Code: Select all

TB		HOR	Size
----------------------------------------
KRPPvKRP	1	23,261,437,520
KBPPvKBP	2	30,746,474,768
KPPPvKPP	3	2,107,379,792
KNPPvKNP	4	29,087,503,696
KNPPvKBP	5	30,955,844,688
KBPPvKNP	6	28,727,898,832
KQPPvKQP	7	22,632,171,664
KBPPvKRP	8	31,606,770,192
KRPPPvKR	9	1,075,778,832
KRBPvKRP	10	48,381,043,216
KRNPvKRP	11	58,025,199,248
KNPPvKRP	12	29,911,219,856
KPPPvKBP	13	7,584,064,976
KRBPvKRB	14	80,280,358,288
KRPPvKRB	15	28,687,360,336
KPPPvKNP	16	6,680,648,720
KBPPvKPP	17	7,973,639,248
KRPPvKRN	18	27,834,688,208
KRPPvKPP	19	5,054,627,664
KNPPvKPP	20	9,689,699,280
KRNPvKRN	21	74,082,453,136
KRNPvKRB	22	72,853,318,096
KRPPvKNP	23	9,632,432,848
KRPPvKBP	24	11,435,368,336
KPPPvKRP	25	4,645,229,072
KBNPvKRP	26	82,917,193,040
KRPPvKQP	27	16,689,932,048
KNPPPvKN	28	959,916,944
KQPPPvKQ	31	3,745,607,056
KNPPPvKR	32	7,312,293,712
----------------------------------------		
		Total	794,577,553,312

[Ozymandias]

re-did the 800 MB list

You mean 740 GB?

[MikeB]

re-did the 800 MB list

You mean 740 GB?

I do - thank you, good catch.

[Nordlandia]

So is this list considered the most important files to keep on SSD as of now when volume is limited.

Is is possible to rearrange the list based on occurance and probability, for Samsung T5 500 GB (465 GB actually). Christmas gift from my parents. So current list has to to be reduced considerable.

Read performance is 512 MB/s through USB 3.1 Gen 2, despite USB is bad solution for tablebase probing because of the delay. Nevertheless i can use it now until i get something better. I keep 5 and 6-men on Samsung 960 EVO and want to utilize Samsung T5 for 7-piece tables.

Any recommendations / advice?

Code: Select all

TB		HOR	Size
----------------------------------------
KRPPvKRP	1	23,261,437,520
KBPPvKBP	2	30,746,474,768
KPPPvKPP	3	2,107,379,792
KNPPvKNP	4	29,087,503,696
KNPPvKBP	5	30,955,844,688
KBPPvKNP	6	28,727,898,832
KQPPvKQP	7	22,632,171,664
KBPPvKRP	8	31,606,770,192
KRPPPvKR	9	1,075,778,832
KRBPvKRP	10	48,381,043,216
KRNPvKRP	11	58,025,199,248
KNPPvKRP	12	29,911,219,856
KPPPvKBP	13	7,584,064,976
KRBPvKRB	14	80,280,358,288
KRPPvKRB	15	28,687,360,336
KPPPvKNP	16	6,680,648,720
KBPPvKPP	17	7,973,639,248
KRPPvKRN	18	27,834,688,208
KRPPvKPP	19	5,054,627,664
KNPPvKPP	20	9,689,699,280
KRNPvKRN	21	74,082,453,136
KRNPvKRB	22	72,853,318,096
KRPPvKNP	23	9,632,432,848
KRPPvKBP	24	11,435,368,336
KPPPvKRP	25	4,645,229,072
KBNPvKRP	26	82,917,193,040
KRPPvKQP	27	16,689,932,048
KNPPPvKN	28	959,916,944
KQPPPvKQ	31	3,745,607,056
KNPPPvKR	32	7,312,293,712
----------------------------------------		
		Total	794,577,553,312

[MikeB]

So is this list considered the most important files to keep on SSD as of now when volume is limited.

Is is possible to rearrange the list based on occurance and probability, for Samsung T5 500 GB (465 GB actually). Christmas gift from my parents. So current list has to to be reduced considerable.

Read performance is 512 MB/s through USB 3.1 Gen 2, despite USB is bad solution for tablebase probing because of the delay. Nevertheless i can use it now until i get something better. I keep 5 and 6-men on Samsung 960 EVO and want to utilize Samsung T5 for 7-piece tables.

Any recommendations / advice?

Code: Select all

TB		HOR	Size
----------------------------------------
KRPPvKRP	1	23,261,437,520
KBPPvKBP	2	30,746,474,768
KPPPvKPP	3	2,107,379,792
KNPPvKNP	4	29,087,503,696
KNPPvKBP	5	30,955,844,688
KBPPvKNP	6	28,727,898,832
KQPPvKQP	7	22,632,171,664
KBPPvKRP	8	31,606,770,192
KRPPPvKR	9	1,075,778,832
KRBPvKRP	10	48,381,043,216
KRNPvKRP	11	58,025,199,248
KNPPvKRP	12	29,911,219,856
KPPPvKBP	13	7,584,064,976
KRBPvKRB	14	80,280,358,288
KRPPvKRB	15	28,687,360,336
KPPPvKNP	16	6,680,648,720
KBPPvKPP	17	7,973,639,248
KRPPvKRN	18	27,834,688,208
KRPPvKPP	19	5,054,627,664
KNPPvKPP	20	9,689,699,280
KRNPvKRN	21	74,082,453,136
KRNPvKRB	22	72,853,318,096
KRPPvKNP	23	9,632,432,848
KRPPvKBP	24	11,435,368,336
KPPPvKRP	25	4,645,229,072
KBNPvKRP	26	82,917,193,040
KRPPvKQP	27	16,689,932,048
KNPPPvKN	28	959,916,944
KQPPPvKQ	31	3,745,607,056
KNPPPvKR	32	7,312,293,712
----------------------------------------		
		Total	794,577,553,312

The order is in occurrence for human play games , so just go down the list starting with number 1 until you run out of space. Without all the files (including rtbz) , the engine may not play out the ending reliably, however the .rtbw files will let you know if it's an EGTB mate. You can always use the link below to play out a position, just keep clicking the top move on the upper right.
https://syzygy-tables.info/?fen=3N4/4b3 ... _w_-_-_0_1

[Nordlandia]

The lists provided are based on human occurrences. Is it likely that the list will be much different if engine games are used as reference ?

[MikeB]

The lists provided are based on human occurrences. Is it likely that the list will be much different if engine games are used as reference ?

It depends on your settings , if you use game result adjudication, it will more closely resemble human games. If you play out games to the bitter end, you will have endgames that human players would never play out since it would be clear if the position is won or lost to the humans way before you get down to 7 pieces. I would consider using the human games as the "common sense" approach since you really do not need endgames when the position is ones-sided. Humans tend not to play out one-sided games UNLESS they are playing the computer and they are WINNING lol ;>)

Another notch in my belt regards,
MikeB