The scaling with time of opening books

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

User avatar
Laskos
Posts: 10948
Joined: Wed Jul 26, 2006 10:21 pm
Full name: Kai Laskos

Re: The scaling with time of opening books

Post by Laskos »

Guenther wrote:

Code: Select all

StockfishASM_150916NB-64 => TestNB.bin
StockfishASM_150916-64 => GS_medium.bin
StockfishASM_150916AB-64 => Noomen2016.ctg via Aquarium bookadapter

Book details:
----------------
GS_medium.bin = general book, a bit handtuned, max depth 30 plies, 3.8MB

TestNB.bin = just a general 4 plies book for variation/randomization for the 'NB' engine (avoid dupes and only few lines), 9KB

Noomen2016.ctg = set to max 80 plies, tuned ex-commercial book, 1.05GB! (one caveat is that some settings cannot be set in the book adapter,
e.g. min games, but at least it seems to play only marked tournament moves)

Time control 20s+0.25s, 1000 games, 2 cores, Ponder Off, 256MB (per core it seems? Taskmanager says so), Syzygy 5 men, WB 4.80b, adjudication at move 120, manual result correction (BTW no time losses)

Code: Select all

   # PLAYER                      :   RATING  ERROR  POINTS  PLAYED   (%)
   1 StockfishASM_150916-64      :  3208.53   4.97   524.0    1000  52.4
   2 StockfishASM_150916NB-64    :  3191.47   4.97   476.0    1000  47.6

White advantage = 40.94 +/- 5.23
Draw rate (equal opponents) = 79.67 % +/- 1.26

Head to head statistics:

1) StockfishASM_150916-64   3208.53 :   1000 (+140,=768,-92),  52.4 %

   vs.                              :  games (   +,   =,  -),   (%) :     Diff,     SD, CFS (%)
   StockfishASM_150916NB-64         :   1000 ( 140, 768, 92),  52.4 :   +17.07,   5.07,  100.0
Same conditions as above, but in progress, 411 games played, no manual correction yet

Code: Select all

   # PLAYER                      :   RATING  ERROR  POINTS  PLAYED   (%)
   1 StockfishASM_150916AB-64    :  3247.81   9.02   259.0     411  63.0
   2 StockfishASM_150916NB-64    :  3152.19   9.02   152.0     411  37.0

White advantage = 52.87 +/- 9.37
Draw rate (equal opponents) = 76.32 % +/- 2.37

Head to head statistics:

1) StockfishASM_150916AB-64 3247.81 :    411 (+125,=268,-18),  63.0 %

   vs.                              :  games (   +,   =,  -),   (%) :     Diff,     SD, CFS (%)
   StockfishASM_150916NB-64         :    411 ( 125, 268, 18),  63.0 :   +95.61,   9.21,  100.0
-17 for NB SFAsm against GS_medium
-100(ca.) against Noomen2016
More detailed data in the next days.
I finished my testing of Stockfish with BookX against Stockfish with a generic 2-move book for randomization of the openings. My conclusion would be: the importance of the opening book diminishes with longer time control, but it's not clear whether it's due to generally higher draw rate, Wins/Losses ratio remains pretty constant.

Book advantage:

20s + 0.2s --> 111 ELO points 400 games
360s + 3.6s --> 59 ELO points 200 games
3600s + 36s --> 35 ELO points 60 games

The last 3600s + 36s match finished +6 -0 =54, no-book Stockfish didn't manage a single win in 60 games.

For curiosity, I had a book competition, Brainfish book blows away the strongest of Polyglot books:

Games in 10s+0.1s:

Code: Select all

   # PLAYER             : RATING  ERROR    POINTS  PLAYED     (%)
   1 Brainfish          : 3143.7   15.4     307.0     500    61.4
   2 SF My Friends W    : 3113.7   13.8     282.5     500    56.5
   3 SF Book_X          : 3072.2   14.1     248.0     500    49.6
   4 SF Performance     : 3059.7   13.9     237.5     500    47.5
   5 SF Generic 12-m    : 3058.5   14.5     236.5     500    47.3
   6 SF Generic  2-m    : 3000.0   15.1     188.5     500    37.7

White advantage = 31.41 +/- 5.06
Draw rate (equal opponents) = 71.37 % +/- 1.33
shrapnel
Posts: 1339
Joined: Fri Nov 02, 2012 9:43 am
Location: New Delhi, India

Re: The scaling with time of opening books

Post by shrapnel »

I have a question.
Up to what Move Number is it safe to use even a strong Opening Book in a LTC game ?
I heard most of the Books around have been tuned for STC games.
I'd like to avoid the unpleasant surprise of coming out of a supposedly strong Book and finding myself -1 or -2 down in a LTC game (while my opponent had come out of Book many moves back ).
i7 5960X @ 4.1 Ghz, 64 GB G.Skill RipJaws RAM, Twin Asus ROG Strix OC 11 GB Geforce 2080 Tis
shrapnel
Posts: 1339
Joined: Fri Nov 02, 2012 9:43 am
Location: New Delhi, India

Re: The scaling with time of opening books

Post by shrapnel »

Laskos wrote: My conclusion would be: the importance of the opening book diminishes with longer time control, but it's not clear whether it's due to
I should have thought that the reason was obvious.
In LTC game, the Engine has more time to figure out the best move, which may be as good if not better than the Book move.
While in an STC or Blitz game, the Engine has barely a few seconds at the most to find the best move, so a good Opening Book is of more value.
Only logical I should think.
Hence my question above.
i7 5960X @ 4.1 Ghz, 64 GB G.Skill RipJaws RAM, Twin Asus ROG Strix OC 11 GB Geforce 2080 Tis
Jouni
Posts: 3293
Joined: Wed Mar 08, 2006 8:15 pm

Re: The scaling with time of opening books

Post by Jouni »

FYI Thomas Zipproth wrote:

"Now a month after the release and much feedback, I have a better understanding of it.

With a time of 1 minute for the whole game, Cerebellum is of little value, because Stockfish often is not able to continue correctly after the last move out of the book, which is in many cases calculated with a time of about 1 minute on a fast quadcore. And with that time, the book can easily be outplayed by a statistical book simply by playing many 1 minute games, and collecting statistical data about the errors Stockfish is making st that time.

Around 5 minutes is the break even point with statistical books, and above that, for example at 12+1, is where Cerebellum really begins to shine, that's what the results showed so far."
Jouni
shrapnel
Posts: 1339
Joined: Fri Nov 02, 2012 9:43 am
Location: New Delhi, India

Re: The scaling with time of opening books

Post by shrapnel »

Doesn't really answer my question, but thanks anyway for your reply.
i7 5960X @ 4.1 Ghz, 64 GB G.Skill RipJaws RAM, Twin Asus ROG Strix OC 11 GB Geforce 2080 Tis
User avatar
Ozymandias
Posts: 1535
Joined: Sun Oct 25, 2009 2:30 am

Re: The scaling with time of opening books

Post by Ozymandias »

Jouni wrote:FYI Thomas Zipproth wrote:

"at 12+1, is where Cerebellum really begins to shine, that's what the results showed so far."
The Engine Masters say otherwise.
User avatar
Laskos
Posts: 10948
Joined: Wed Jul 26, 2006 10:21 pm
Full name: Kai Laskos

Re: The scaling with time of opening books

Post by Laskos »

Ozymandias wrote:
Jouni wrote:FYI Thomas Zipproth wrote:

"at 12+1, is where Cerebellum really begins to shine, that's what the results showed so far."
The Engine Masters say otherwise.
What do they say? I had a mediocre result too at 1 hour + increment time control with Cerebellum book against a generic 2-mover, +3 -0 =37, no better than MyFrineds book, but it might have been a statistical fluke.
Thomas Zipproth
Posts: 59
Joined: Sat Jan 01, 2011 5:33 pm

Re: The scaling with time of opening books

Post by Thomas Zipproth »

Ozymandias wrote:
Jouni wrote:FYI Thomas Zipproth wrote:

"at 12+1, is where Cerebellum really begins to shine, that's what the results showed so far."
The Engine Masters say otherwise.
That's not so obvious as it my seem. For example I myself (Gilgamesch) playing with Brainfish and Cerebellum ended 4th place on a quadcore, that should be ok.
The only other person which I know for sure that he uses Brainfish + Cerebellum was Longin on 15th place, with a 4 or 6 core I think.

Some people are using only Brainfish but not Cerebellum, because it is a bit faster than most Stockfish compiles or because it has Numa support. Or they are using Cerebellum with an additional Gui book.
The players on the last two places used Brainfish, but as far as I can see from checking their games, with an additional Gui book, so that is something different.

Thomas
shrapnel
Posts: 1339
Joined: Fri Nov 02, 2012 9:43 am
Location: New Delhi, India

Re: The scaling with time of opening books

Post by shrapnel »

Hi Thomas
When are you going to release the Commercial version ?
Right now every Tom, Dick and Harry is using your excellent Cerebellum book, which you make with so much effort, for free....gives me no advantage.
People who are prepared to pay money, should get an advantage over those who want only free gifts !
i7 5960X @ 4.1 Ghz, 64 GB G.Skill RipJaws RAM, Twin Asus ROG Strix OC 11 GB Geforce 2080 Tis
User avatar
Ozymandias
Posts: 1535
Joined: Sun Oct 25, 2009 2:30 am

Re: The scaling with time of opening books

Post by Ozymandias »

Thomas Zipproth wrote:Some people are using only Brainfish but not Cerebellum, because it is a bit faster than most Stockfish compiles or because it has Numa support. Or they are using Cerebellum with an additional Gui book.
The players on the last two places used Brainfish, but as far as I can see from checking their games, with an additional Gui book, so that is something different.
I assumed they were using the whole package, using a subpar book instead of Cerebellum never crossed my mind. More data is needed if actually just two players were using it correctly. For the moment, i retract my statement.