Perft(7) challenge position #4

[sje]

Perft(7) challenge position #4

From line 24,145 of Perft(14) work unit #417:
[d]rnbqk1nr/p1pp1ppp/1p6/2b1p1B1/8/1QPP4/PP2PPPP/RN2KBNR b KQkq - 2 4[/d]
What is the value of perft(7) for the above?

One program says: 40,552,059,486
A second says: 40,552,058,742

[Ajedrecista]

Hello Steven:

Divided perft(7) of that position by JetChess 1.0.0.0:

Code: Select all

rnbqk1nr/p1pp1ppp/1p6/2b1p1B1/8/1QPP4/PP2PPPP/RN2KBNR b KQkq - 2 4

  1  qd8-e7  1492445476
  2  qd8-f6  1992746647
  3  qd8*g5  1421478416
  4  bc5-b4  1117769244
  5  bc5-a3  1150495631
  6  bc5-d4  1031537710
  7  bc5-e3   969283318
  8  bc5*f2    83974980
  9  bc5-d6  1037788240
 10  bc5-e7   957480333
 11  bc5-f8  1101421164
 12  bc8-b7  1996664787
 13  bc8-a6  1430782408
 14  nb8-a6  1138733001
 15  nb8-c6  1690767595
 16  ng8-e7  1319506568
 17  ng8-f6  1360087731
 18  ng8-h6  1442283113
 19   e5-e4  1412785552
 20   b6-b5  1349570785
 21   a7-a6  1126372794
 22   a7-a5  1450366251
 23   c7-c6  1205188937
 24   d7-d6  1641506423
 25   d7-d5  2113554228
 26   f7-f6   954778435
 27   f7-f5  1260913952
 28   g7-g6  1202196239
 29   h7-h6  1298880116
 30   h7-h5  1477191017
 31  ke8-f8  1323507651

Total:      40552058742

Result: 40,552,058,742 (move pathes after 7 half moves).

The second programme is right.

Regards from Spain.

Ajedrecista.

[xmas79]

Hello Steven:

Divided perft(7) of that position by JetChess 1.0.0.0:

Code: Select all

rnbqk1nr/p1pp1ppp/1p6/2b1p1B1/8/1QPP4/PP2PPPP/RN2KBNR b KQkq - 2 4

  1  qd8-e7  1492445476
  2  qd8-f6  1992746647
  3  qd8*g5  1421478416
  4  bc5-b4  1117769244
  5  bc5-a3  1150495631
  6  bc5-d4  1031537710
  7  bc5-e3   969283318
  8  bc5*f2    83974980
  9  bc5-d6  1037788240
 10  bc5-e7   957480333
 11  bc5-f8  1101421164
 12  bc8-b7  1996664787
 13  bc8-a6  1430782408
 14  nb8-a6  1138733001
 15  nb8-c6  1690767595
 16  ng8-e7  1319506568
 17  ng8-f6  1360087731
 18  ng8-h6  1442283113
 19   e5-e4  1412785552
 20   b6-b5  1349570785
 21   a7-a6  1126372794
 22   a7-a5  1450366251
 23   c7-c6  1205188937
 24   d7-d6  1641506423
 25   d7-d5  2113554228
 26   f7-f6   954778435
 27   f7-f5  1260913952
 28   g7-g6  1202196239
 29   h7-h6  1298880116
 30   h7-h5  1477191017
 31  ke8-f8  1323507651

Total:      40552058742

Result: 40,552,058,742 (move pathes after 7 half moves).

The second programme is right.

Regards from Spain.

Ajedrecista.

I wouldn't trust another program that uses hashed perft....

[uaf]

Chiron (w/ and w/o hashing):

Code: Select all

Chiron[B-4]> perft 7 1
clearing hash table...done
perft nodes: 40552058742
used time: 257.37s
nodes/sec: 157562657
Chiron[B-4]> perft 7 0
perft nodes: 40552058742
used time: 580.76s
nodes/sec: 69825726

[Ajedrecista]

Hello Natale:

I wouldn't trust another program that uses hashed perft....

Sure, but JetChess performed very well in a perft count of more than 1.1e+15 leaf nodes (first post and second post). I know that it is not a proof but anything is less.

That said, I have just tried qperft without hash:

Code: Select all

C:\[...]\qperft>qperft
Usage is&#58; perft <depth> &#91;H<hash size>&#93; &#91;-<split depth>&#93; &#91;<FEN string>&#93;
          <hash size> = 20 gives you 2^20 = 1M entries &#40;16MB&#41;

C&#58;\&#91;...&#93;\qperft>qperft 7 "rnbqk1nr/p1pp1ppp/1p6/2b1p1B1/8/1QPP4/PP2PPPP/RN2KBNR b KQkq - 2 4"
 - - - - - - - - - - - -
 - - - - - - - - - - - -
 - - r n b q k . n r - -
 - - p . p p . p p p - -
 - - . p . . . . . . - -
 - - . . b . p . B . - -
 - - . . . . . . . . - -
 - - . Q P P . . . . - -
 - - P P . . P P P P - -
 - - R N . . K B N R - -
 - - - - - - - - - - - -
 - - - - - - - - - - - -

Quick Perft by H.G. Muller
Perft mode&#58; No hashing, bulk counting in horizon nodes

perft&#40; 1&#41;=           31 ( 0.000 sec&#41;
perft&#40; 2&#41;=         1116 ( 0.000 sec&#41;
perft&#40; 3&#41;=        33828 ( 0.000 sec&#41;
perft&#40; 4&#41;=      1184142 ( 0.031 sec&#41;
perft&#40; 5&#41;=     36838554 ( 0.719 sec&#41;
perft&#40; 6&#41;=   1272676278 &#40;25.953 sec&#41;
perft&#40; 7&#41;=  40552058742 &#40;892.109 sec&#41;

C&#58;\&#91;...&#93;\qperft>

The result is confirmed without hash and this time JetChess was lucky of not mess the value with its hash, which can not be disabled by the way.

Regards from Spain.

Ajedrecista.

[sje]

64 bit signature false positive rate is now circa 1.7*10^-6 (1/575,000)

4 cases were located from 23 work units; each work unit has 100,000 unique positions.

Verified work units (19):
400-408 411-412 414 416 419 421-422 424 426 428

Work units with at least one false positive (4):
409 413 415 417

[pkumar]

I used 64-bit hashing. My engine agrees with other reported results. If two engines with different 64-bit hashing report identical perft results, can it be assumed that the error probability is as low as for 128-bit hashing?
(Deleted my last post as it was wrongly worded)

D5 36838554 (1.513e+000s)
D6 1272676278 (2.332e+001s)
D7 40552058742 (4.393e+002s)