Perft(13), second half

[sje]

The first Perft(13) thread has gotten a bit large, so I've started this new thread to report results for the second half of the calculation.

After 25 weeks and 3 days:

The run has produced 445,182 draft 8 records, about 54% of the total needed.
Mean time per draft 8 record: ca. 35 seconds

The run has produced 133 (33.25%) of the 400 draft 11 records.
Mean time per draft 11 record: ca. 32 hours

http://dl.dropbox.com/u/31633927/Perft/Perft13/draft11

160 draft 11 results are needed for the first eight draft 12 results. The first of these long-awaited draft 12 records should show up in the next five or six days.

[sje]

After 26 weeks:

The run has produced 457,568 draft 8 records, about 56% of the total needed.
Mean time per draft 8 record: ca. 34 seconds

The run has produced 144 (36%) of the 400 draft 11 records.
Mean time per draft 11 record: ca. 30 hours

http://dl.dropbox.com/u/31633927/Perft/Perft13/draft11

[sje]

After almost exactly six months of calculation, the run has produced its first draft 12 record. There will be 19 more of these with seven more showing up over the next few days.
[d]rnbqkbnr/pppppppp/8/8/8/N7/PPPPPPPP/R1BQKBNR b KQkq - 1 1[/d]
The perft(12) for the above is 63,558,937,554,457,795

http://dl.dropbox.com/u/31633927/Perft/Perft13/draft11

http://dl.dropbox.com/u/31633927/Perft/Perft13/draft12

[sje]

Draft 12 record number 2 of 20:
[d]rnbqkbnr/pppppppp/8/8/8/P7/1PPPPPPP/RNBQKBNR b KQkq - 0 1[/d]
The perft(12) for the above is 54,239,338,583,061,004

This matches the result shown in an earlier post (by Byrne, I believe), although I can't find it at the moment.

[sje]

Draft 12 record number 3 of 20:
[d]rnbqkbnr/pppppppp/8/8/8/2N5/PPPPPPPP/R1BQKBNR b KQkq - 1 1[/d]
The perft(12) for the above is 84,663,535,534,493,517

There will be five more draft 12 records appearing shortly and then a couple of months before the rest begin to show.

http://dl.dropbox.com/u/31633927/Perft/Perft13/draft11

http://dl.dropbox.com/u/31633927/Perft/Perft13/draft12

[sje]

Draft 12 record number 4 of 20:
[d]rnbqkbnr/pppppppp/8/8/8/7N/PPPPPPPP/RNBQKB1R b KQkq - 1 1[/d]
The perft(12) for the above is 64,658,238,653,217,709

For more draft 12 records to go in this batch; threads for 1 c3/c4/d3/d4 have been started.

[sje]

Draft 12 record number 5 of 20:
[d]rnbqkbnr/pppppppp/8/8/8/1P6/P1PPPPPP/RNBQKBNR b KQkq - 0 1[/d]
The perft(12) for the above is 72,470,867,205,869,240

The thread for 1 e3 has started, the thread for 1 e4 is the next to be started.

[Ajedrecista]

Hello Steven:

The perft(12) for the above is 54,239,338,583,061,004

This matches the result shown in an earlier post (by Byrne, I believe), although I can't find it at the moment.

Great work! I copy here the link of the result given by Paul Byrne (the results match exactly):

http://talkchess.com/forum/viewtopic.ph ... 31&t=40392

Perft 12: 54,239,338,583,061,004

-paul

It was easy to find: just check in Paul's profile. I recommend to the readers take a look not only at this post but to the whole thread.

I remember that I did two clumsy estimates for each of the twenty draft 12 records of Perft(13), applying the idea of estimating the branching factors (idea taken from Adam Hair). These two estimates were almost the same (only changing a little the adjusting polynomials). Here are my posts:

http://talkchess.com/forum/viewtopic.ph ... ht=#419839

http://talkchess.com/forum/viewtopic.ph ... 82&t=39678

First estimate:

Code: Select all

Move            Estimate 

a3       54,531,136,234,162,156 
a4       79,063,508,502,158,903 
b3       72,639,688,816,621,455 
b4       73,709,067,003,701,511 
c3       86,056,827,153,858,571 
c4       97,970,589,000,615,670 
d3      141,383,458,087,446,826 
d4      213,018,757,033,778,459 
e3      240,725,195,611,700,379 
e4      246,589,794,037,145,382 
f3       43,493,804,033,203,347 
f4       61,162,666,611,330,980 
g3       76,232,972,134,254,158 
g4       65,814,044,230,043,052 
h3       54,049,205,191,901,014 
h4       80,829,651,636,234,054 
Na3      63,795,543,395,998,863 
Nc3      84,531,400,071,549,975 
Nf3      82,344,973,694,923,533 
Nh3      64,898,200,828,736,569 
------------------------------- 
SUM = 1,982,840,483,309,364,857

-------------------------------

Second estimate:

Code: Select all

Move            Estimate 

a3       54,531,136,245,433,857 
a4       79,063,508,508,444,871 
b3       72,639,688,830,427,074 
b4       73,709,067,010,541,664 
c3       86,056,827,162,078,542 
c4       97,970,589,005,764,498 
d3      141,383,458,092,703,225 
d4      213,018,757,047,938,236 
e3      240,725,195,616,730,295 
e4      246,589,794,029,993,095 
f3       43,493,804,037,847,797 
f4       61,162,666,623,173,938 
g3       76,232,972,139,647,889 
g4       65,814,044,240,929,760 
h3       54,049,205,206,297,538 
h4       80,829,651,642,814,500 
Na3      63,795,543,402,742,775 
Nc3      84,531,400,076,854,934 
Nf3      82,344,973,704,170,665 
Nh3      64,898,200,838,916,087 
------------------------------- 
SUM = 1,982,840,483,463,451,240

The two adjusting polynomials were quartic functions, and I think the coefficients were very similar between them, this is why these estimates are almost clones.

I can calculate the relative error in percentage [(estimate) = (1 + z)·(real)], where the relative error (in percentage) is 100·z; more less I get (roundings included):

Code: Select all

Move:      Error (%):
 Na3       +0.372262%
  a3       +0.537982%
 Nc3       -0.156071%
 Nh3       +0.371124%
  b3       +0.232951%

Hoping no typos; I did an overall estimate for Perft(13) using my own (and clumsy) method last summer, and my error was ~ -0.03022% (better than estimating the branching factor of each initial move). Here is a post by mine (giving my estimates), the day before Paul posted Perft(13) value:

http://talkchess.com/forum/viewtopic.ph ... 22&t=41032

Regards from Spain.

Ajedrecista.

[sje]

Draft 12 record number 6 of 20:
[d]rnbqkbnr/pppppppp/8/8/8/5N2/PPPPPPPP/RNBQKB1R b KQkq - 1 1[/d]
The perft(12) for the above is 82,432,607,453,980,338

[sje]

Draft 12 record number 7 of 20:
[d]rnbqkbnr/pppppppp/8/8/1P6/8/P1PPPPPP/RNBQKBNR b KQkq - 0 1[/d]
The perft(12) for the above is 73,447,518,515,970,566

Threads for 1 e4 and 1 f3 have started; the thread for 1 f4 will be started next.

The perft(12) result for 1 a4 should appear shortly.