Perft(13) [3.4 GHz Core i7-2600, 16 GB RAM]

[Ajedrecista]

Hi Steven:

In the thread of Perft(13), BF's were discussed a while. On the date August, 6th I uploaded some Excel files and little more stuff regarding relative percentages of each move. If you missed it, here is the link.

According with some calculations by me (with adjusting polynomials, so not very serious), the greater BF should be around 34 (for e4) while the less BF should be near 28 (for f3). I think I posted it elsewhere. A month ago (more less), Daniel Shawul, running UCT methods (maybe I am wrong) obtained preliminary results that more less fit with my own estimates, so maybe I am not totally wrong. Since draft 11 records up to now are not of d3, d4, e3 and e4 initial moves, this is why the average draft 11 record is greater than each one of the 32 drafts 11 you got (but I suppose that you already know it...). Drafts 11 records starting with e3 and e4 should be the greater ones; OTOH: I guess that f3, f6 should be the smallest draft 11 record for Perft(13).

Thank you very much for keeping us well informed about this huge perft count.

Regards from Spain.

Ajedrecista.

[sje]

Because Apple is dropping support for its MobileMe data hosting, I have pulled all my files from the service and this includes the perft data. It may be the case that the upcoming iCloud service can offer reliable hosting, but Apple's track record here is not so good.

In the interim, I'm using DropBox and can share the appropriate data files from there soon.

And the perft(13) calculation continues. After 41 days, Symbolic has produced about 160,000 draft 8 records At completion, there will be about 822,000 such records.

And a new draft 11 result has appeared, the 33rd so far:
[d]rnbqkbnr/pp1ppppp/2p5/8/8/N7/PPPPPPPP/R1BQKBNR w KQkq - 0 2[/d]
The perft(11) for the above is 2,768,271,331,293,873.

33 (8.25%) down, 367 (91.75%) to go.

[sje]

Two more draft 11 records (#34 and #35):

Code: Select all

rnbqkbnr/pp1ppppp/2p5/8/8/2N5/PPPPPPPP/R1BQKBNR w KQkq - 0 2 11 3782980467479785
rnbqkbnr/pp1ppppp/2p5/8/8/P7/1PPPPPPP/RNBQKBNR w KQkq - 0 2 11 2342166292286351

35 (8.75%) down, 365 (91.25%) to go.

[sje]

Three more draft 11 records; all posted to encourage independent verification:

Code: Select all

rnbqkbnr/pp1ppppp/2p5/8/8/1P6/P1PPPPPP/RNBQKBNR w KQkq - 0 2 11 3152460705609456
rnbqkbnr/pp1ppppp/2p5/8/8/7N/PPPPPPPP/RNBQKB1R w KQkq - 0 2 11 2844253997560744
rnbqkbnr/pp1ppppp/2p5/8/8/5N2/PPPPPPPP/RNBQKB1R w KQkq - 0 2 11 3602834392535550

38 (9.5%) down, 362 (90.5%) to go.

The first draft 12 records will appear when the draft 11 count is up to 160 or so, perhaps in late October.

[sje]

After 43 days of wall time, draft 11 record number 39:
[d]rnbqkbnr/pp1ppppp/2p5/8/P7/8/1PPPPPPP/RNBQKBNR w KQkq - 0 2[/d]
The perft(11) for the above is 3,377,958,687,547,657.

39 (9.75%) down, 361 (90.25%) to go.

Assuming a 200 day run time, the average production rate of draft records will be about two per day. The rate so far has been just under one per day; this will slowly speed up over time. On the last day of the calculation, the last ten or so draft records should appear.

[sje]

After 44 days of wall time, draft 11 record number 40:
[d]rnbqkbnr/pp1ppppp/2p5/8/1P6/8/P1PPPPPP/RNBQKBNR w KQkq - 0 2[/d]
The perft(11) for the above: 3,342,480,370,972,141.

40 (10%) down, 360 (90%) to go.

There are 168,000 draft 8 records output so far; that's about 20% of the needed count. The mean calculation time per draft 8 record is about 22.6 seconds.

[sje]

After 47 days of calculation (and a few power outages), the current estimated completion date is March 26, 2012. The increase in the estimated run time is based on the increase in the frequency of "hard" positions. By "hard", I mean positions that have multiple center pawns advanced in the early ply.

Here's an example of a moderately hard draft 8 position:
[d]rnbqkbnr/pp1p1ppp/8/2p1p3/1P6/4P2N/P1PP1PPP/RNBQKB1R b KQkq - 0 3[/d]
The perft(8) for the above is 949,549,001,945. In comparison, perft(8) for the initial array is rather less at only 84,998,978,956.

[sje]

Here is some output from the top status utility on the machine running the calculation. Note the 247,945 minutes (ca. 142 days) run time for the process. Linux bases the CPU time on hyper threads, not core threads and there are eight threads in play.

Code: Select all

Tasks: 170 total,   2 running, 167 sleeping,   0 stopped,   1 zombie
Cpu(s): 99.9%us,  0.1%sy,  0.0%ni,  0.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Mem:  16444684k total, 16346192k used,    98492k free,   208668k buffers
Swap: 16752636k total,   347964k used, 16404672k free,  2714520k cached

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND                                                           
 1756 sje       20   0 12.4g  12g 1336 S  797 76.5 247945:37 Symbolic                                                           

[sje]

After 51 days, the calculation has produced about 186,000 draft 8 records. The new estimated completion date is April 2nd, 2012.

The most recent results:

Code: Select all

rnbqkbnr/pp1p1ppp/4p3/2p5/N3P3/8/PPPP1PPP/R1BQKBNR b KQkq - 1 3 8 806672063359
rnb1kbnr/pp1ppppp/1q6/1Bp5/8/P3P3/1PPP1PPP/RNBQK1NR b KQkq - 2 3 8 668595979700
rnbqkbnr/pp1ppppp/8/8/1P1p4/3Q4/P1P1PPPP/RNB1KBNR b KQkq - 1 3 8 990832906438
rnb1kbnr/pp1ppppp/Bq6/2p5/4P3/N7/PPPP1PPP/R1BQK1NR b KQkq - 2 3 8 853727646979
rnbqkbnr/pp1pp1pp/5p2/2p5/3P4/5N2/PPPKPPPP/RNBQ1B1R b kq - 1 3 8 215280367515
rnbqkbnr/pp2pppp/3p4/2p5/8/4P2N/PPPPKPPP/RNBQ1B1R b kq - 1 3 8 279126515546
rnbqkbnr/pp2pppp/8/2pp4/6Q1/1P2P3/P1PP1PPP/RNB1KBNR b KQkq - 1 3 8 2067234988127
rnb1kbnr/pp1ppppp/1q6/2p5/4P3/N7/PPPPNPPP/R1BQKB1R b KQkq - 2 3 8 474724753102
rnbqkbnr/pp1p1ppp/4p3/2pN4/4P3/8/PPPP1PPP/R1BQKBNR b KQkq - 1 3 8 1039880555492
rnb1kbnr/pp1ppppp/Bq6/2p5/8/P3P3/1PPP1PPP/RNBQK1NR b KQkq - 2 3 8 807514925906

[sje]

As mentioned earlier, some of the draft 8 record counts can be fairly large. Here's a verification run of one of the recent results done all by its lonesome self on a separate box:
[d]rnbqkbnr/pp2pppp/8/2pp4/6Q1/1P2P3/P1PP1PPP/RNB1KBNR b KQkq - 1 3[/d]

Code: Select all

[] emptran 8
Bxg4 18,078,680,415
Bd7 74,010,050,049
Be6 62,470,891,901
Bf5 86,483,789,607
Na6 63,991,602,356
Nd7 47,745,655,289
Nh6 66,057,131,001
Nf6 77,039,439,053
Nc6 93,982,773,746
Qa5 85,744,207,928
Qc7 105,522,085,151
Qb6 106,367,651,283
Qd6 126,258,823,266
Qd7 72,422,310,827
a6 64,127,351,261
b5 68,807,933,023
a5 72,619,084,231
b6 59,822,349,656
c4 66,664,415,654
d4 73,968,079,707
f5 51,278,529,873
f6 54,475,865,708
e6 86,806,620,450
e5 104,787,610,258
g5 64,876,649,246
h6 62,529,998,177
g6 72,370,147,718
h5 77,925,261,293
Depth: 8   Count: 2,067,234,988,127   Elapsed: 490.107  (4.21793e+09 Hz / 2.37083e-10 s)