Perft(13), final fifth

Discussion of chess software programming and technical issues.

Moderator: Ras

User avatar
sje
Posts: 4675
Joined: Mon Mar 13, 2006 7:43 pm

After 52 weeks

Post by sje »

After 52 weeks:

The run has produced 750,598 draft 8 records, about 91% of the total needed. The average draft 8 record generation time is about 41.9 seconds.

There are now 294 draft 11 records, 73.5% of the 400 needed. The average draft 11 record generation time is about 29.7 hours.
http://dl.dropbox.com/u/31633927/Perft/Perft13/draft11

There are now 10 draft 12 records, 50% of the 20 needed. The average draft 12 record generation time is about 36.4 days.
http://dl.dropbox.com/u/31633927/Perft/Perft13/draft12

Recent output:

Code: Select all

rnb1kbnr/pppp1ppp/4p3/Q7/7q/4P3/PPPP1PPP/RNB1KBNR b KQkq - 3 3 8 4579936463741
r1bqkbnr/ppppp1pp/n6B/5p2/8/3P4/PPPKPPPP/RN1Q1BNR b kq - 3 3 8 173224918652
r1bqkbnr/ppppppp1/n6B/7p/2PP4/8/PP2PPPP/RN1QKBNR b KQkq - 2 3 8 401446513578
rnbqkbnr/pppppppp/8/8/2PP4/6P1/PP2PP1P/RNBQKBNR b KQkq - 0 3 8 317390444564
rnbqkb1r/ppppp1pp/5p1n/6B1/3P2P1/8/PPP1PP1P/RN1QKBNR b KQkq - 2 3 8 313803564178
rnbqkb1r/pppppppp/8/3n4/7P/2PP4/PP2PPP1/RNBQKBNR b KQkq - 0 3 8 369303233342
r1bqkbnr/ppppp1pp/n6B/5p2/8/3PP3/PPP2PPP/RN1QKBNR b KQkq - 0 3 8 252643235905
rnbqkbnr/pppppppp/8/8/2P5/4P1P1/PP1P1P1P/RNBQKBNR b KQkq - 0 3 8 258549140313
rnbqkb1r/ppppp1pp/5p1B/8/3P2P1/8/PPP1PP1P/RN1QKBNR b KQkq - 0 3 8 196969175448
rnbqkb1r/pppppppp/8/3n4/2P5/3P4/PP2PPPP/RNBQKBNR b KQkq - 0 3 8 368337756500
User avatar
sje
Posts: 4675
Joined: Mon Mar 13, 2006 7:43 pm

Perft(12) for 1 c3

Post by sje »

Perft(12) for the position after 1 c3 is 85,630,662,795,901,357 and this is the largest perft(12) found so far in the run. The 1 c3 worker thread has concluded and the 1 h3 worker thread has started.

The run is one calendar year old today and should finish in a month or two.
User avatar
sje
Posts: 4675
Joined: Mon Mar 13, 2006 7:43 pm

After 53 weeks

Post by sje »

After 53 weeks:

The run has produced 761,156 draft 8 records, about 93% of the total needed. The average draft 8 record generation time is about 42.1 seconds.

There are now 313 draft 11 records, 78.25% of the 400 needed. The average draft 11 record generation time is about 28.4 hours.
http://dl.dropbox.com/u/31633927/Perft/Perft13/draft11

There are now 11 draft 12 records, 55% of the 20 needed. The average draft 12 record generation time is about 33.7 days.
http://dl.dropbox.com/u/31633927/Perft/Perft13/draft12

Recent output:

Code: Select all

rnbqkbnr/pppp2pp/4p3/5p1Q/8/4P2P/PPPP1PP1/RNB1KBNR b KQkq - 1 3 8 54543469330
r1bqkb1r/pppppppp/2n4n/6B1/2PP4/8/PP2PPPP/RN1QKBNR b KQkq - 2 3 8 472684965137
rnbqk1nr/pppppp1p/7b/6p1/3P2P1/8/PPP1PPBP/RNBQK1NR b KQkq - 0 3 8 237076911116
rnbqk1nr/pppppp1p/7b/6p1/6P1/4P3/PPPP1PBP/RNBQK1NR b KQkq - 0 3 8 147375719865
rnbqkb1r/pppppppp/8/5n2/4P3/3P2P1/PPP2P1P/RNBQKBNR b KQkq - 0 3 8 613273399703
rnbqkb1r/pppp1ppp/B3p2n/8/3P4/4P3/PPP2PPP/RNBQK1NR b KQkq - 2 3 8 1169610755323
rnbqk1nr/pppppp1p/7b/6p1/4P1P1/8/PPPP1PBP/RNBQK1NR b KQkq - 0 3 8 112358926982
rnbqk1nr/ppppppbp/6p1/8/8/3PPQ2/PPP2PPP/RNB1KBNR b KQkq - 2 3 8 821203146294
rnbqk1nr/pppppp1p/7b/6p1/6P1/5B2/PPPPPP1P/RNBQK1NR b KQkq - 3 3 8 65990004904
r1bqkb1r/pppppppp/2n4B/8/2PP4/8/PP2PPPP/RN1QKBNR b KQkq - 0 3 8 325721033816
User avatar
sje
Posts: 4675
Joined: Mon Mar 13, 2006 7:43 pm

Re: Perft(13), final fifth

Post by sje »

Perft(12) for the position after 1 g3 is 75,907,561,722,826,536 and this is the twelfth of of the 20 draft 12 results computed to date. The 1 g3 worker thread has concluded and the 1 h4 worker thread has started.

Each of the eight of the eight hardware supported hpyerthreads is working on a draft 12 calculation as has been the case since the start of the run. This will change as the current threads complete as there are no more draft 12 calculations to be started.
User avatar
sje
Posts: 4675
Joined: Mon Mar 13, 2006 7:43 pm

Re: Perft(13), final fifth

Post by sje »

Perft(12) for the position after 1 c4 is 97,580,419,694,352,415 and this is the largest of the thirteen draft 12 results computed to date. The 1 c4 worker thread has concluded and there are no more worker threads to be started. Seven worker hyperthreads are now executing on the four physical CPU cores.

I'll guess that the order of the remaining completions will be: 1 g4, 1 h3, 1 h4, 1 d3, 1 d4, 1 e3, 1 e4.
Daniel Shawul
Posts: 4185
Joined: Tue Mar 14, 2006 11:34 am
Location: Ethiopia

Re: Draft 12 result for 1 f4 completed

Post by Daniel Shawul »

While we are getting closer to the final values, I reviewed some of the techniques we came up with to approximate perft estimate. I now realize that the monte carlo methods we used have good theoretical background in statistical machine learning. Variance reduction, rejection/importance sampling, different markov chain monte carlo samplers and other methods are used for integration and making statistical inference in many fields. Our perft is also integration in a sense since we sum the results making it some what different from an alpha-beta type search where we take the max. Anyway this was sort of a reminiscent of the nice discussion we had about a year ago. Compare that to the useless garbage that takes up 30 pages nowadays..argh!
cheers
Daniel
User avatar
sje
Posts: 4675
Joined: Mon Mar 13, 2006 7:43 pm

Re: Draft 12 result for 1 f4 completed

Post by sje »

The run may be finished within a month; in 12 1/2 months it has computed about 93.5% of the draft 8 records.

The grand result perft(13) number from Paul Byrne to match is 1,981,066,775,000,396,239; soon we shall see what we shall see.

The largest draft 11 results so far:

Code: Select all

rnbqkbnr/ppp1pppp/8/3p4/4P3/8/PPPP1PPP/RNBQKBNR w KQkq - 0 2 11 28808031383517106
rnbqkbnr/ppp1pppp/8/3p4/8/4P3/PPPP1PPP/RNBQKBNR w KQkq - 0 2 11 22334345588975690
rnbqkbnr/ppp1pppp/3p4/8/4P3/8/PPPP1PPP/RNBQKBNR w KQkq - 0 2 11 17344682849928475
rnbqkbnr/ppp1pppp/8/3p4/3P4/8/PPP1PPPP/RNBQKBNR w KQkq - 0 2 11 17022349302490272
rnbqkbnr/ppp1pppp/3p4/8/8/4P3/PPPP1PPP/RNBQKBNR w KQkq - 0 2 11 16408433376099590
rnbqkbnr/ppp1pppp/3p4/8/3P4/8/PPP1PPPP/RNBQKBNR w KQkq - 0 2 11 15948394992914745
rnbqkbnr/pppp1ppp/8/4p3/8/3P4/PPP1PPPP/RNBQKBNR w KQkq - 0 2 11 15651641608357144
rnbqkbnr/pppp1ppp/4p3/8/8/3P4/PPP1PPPP/RNBQKBNR w KQkq - 0 2 11 15175911933142687
rnbqkbnr/ppp1pppp/8/3p4/8/3P4/PPP1PPPP/RNBQKBNR w KQkq - 0 2 11 14915052041878031
rnbqkbnr/pp1ppppp/8/2p5/3P4/8/PPP1PPPP/RNBQKBNR w KQkq - 0 2 11 12598425892815128
User avatar
sje
Posts: 4675
Joined: Mon Mar 13, 2006 7:43 pm

After 54 weeks

Post by sje »

After 54 weeks:

The run has produced 770,624 draft 8 records, about 94% of the total needed. The average draft 8 record generation time is about 42.4 seconds.

There are now 330 draft 11 records, 82.5% of the 400 needed. The average draft 11 record generation time is about 27.5 hours.
http://dl.dropbox.com/u/31633927/Perft/Perft13/draft11

There are now 13 draft 12 records, 65% of the 20 needed. The average draft 12 record generation time is about 29.1 days.
http://dl.dropbox.com/u/31633927/Perft/Perft13/draft12

Recent output:

Code: Select all

r1bqkbnr/pppppppp/7B/8/1n1P2P1/8/PPP1PP1P/RN1QKBNR b KQkq - 2 3 8 327424752321
r1bqkbnr/ppp1pppp/n7/3p3P/8/8/PPPPPPPR/RNBQKBN1 b Qkq - 0 3 8 273797688254
r1bqkbnr/ppp1pppp/n7/3p4/7P/8/PPPPPPPR/RNBQKBN1 w Qkq - 2 3 9 7755936033636
rnb1kbnr/pppp1ppp/5q2/4P3/8/7P/PPP1PPP1/RNBQKBNR b KQkq - 0 3 8 1869180399528
rnb1kbnr/pppp1ppp/5q2/4p3/3P4/7P/PPP1PPP1/RNBQKBNR w KQkq - 1 3 9 45479519342339
r1bqkbnr/ppp1pppp/2n5/3p4/7P/8/PPPPPPP1/RNBQKBNR b Qkq - 3 3 8 319474068292
rnbqkbnr/ppppppp1/8/7p/5B2/3P3P/PPP1PPP1/RN1QKBNR b KQkq - 0 3 8 305383092119
rnbqkbnr/pppp2pp/4p3/5p2/4P3/3B4/PPPPKPPP/RNBQ2NR b kq - 1 3 8 609954622531
r1bqkbnr/pppppppp/8/8/1n1P2P1/8/PPPQPP1P/RNB1KBNR b KQkq - 2 3 8 399688128688
rnb1kbnr/ppppqppp/8/4p3/3P4/7P/PPPNPPP1/R1BQKBNR b KQkq - 2 3 8 485078915810
User avatar
sje
Posts: 4675
Joined: Mon Mar 13, 2006 7:43 pm

Re: About the largest draft 11 record of Perft(13).

Post by sje »

The latest and greatest draft 11 result (1 e3 e6): 29,547,429,295,793,347.
User avatar
sje
Posts: 4675
Joined: Mon Mar 13, 2006 7:43 pm

Re: Perft(13), final fifth

Post by sje »

Perft(12) for the position after 1 g4 is 65,566,769,916,823,116 and this is the fourteenth draft 12 result computed to date. The 1 g4 worker thread has concluded and there are no more worker threads to be started. Six worker hyperthreads are now executing on the four physical CPU cores: 1 d3, 1 d4, 1 e3, 1 e4, 1 h3, 1 h4.