Site views: 119811



Stockfish Polyglot: Updated November, 15 2018


Intro: Stockfish Polyglot is keeping to make great results in the latest Playchess machines tournaments, I didn't expect such a success. Anyway I have seen better performance in computers with alot of cores, such as mainframes, not in normal computers. I still have to make the What's the strongest tournament of this month, being busy for my book, but I bet the results can be different from computer to computer, especially comparing a normal quad-core like mine to a 16 cores powerful computer, or Xeon servers. My theory is that the original Stockfish works better on these latest ones, while asmFish, BrainFish and Raubfisch does a better work on normal computers.
This new release has alot of updates and elo raising, you will find the complete changelog below the 'Download' button.

Stockfish can reach up to 512 cores. From >8 cores it automatically enables the NUMA optimization, as far I have understood by reading the latest official Stockfish GitHub update.


Stockfish Polyglot is none but the very own Stockfish with the only addition of the possibility to use Polyglot (.bin) books. In order to do that I used the library polybook.h from Brainfish and merged with the Stockfish code. This new update features the 'OwnBook' checkbox as requested, so that you can easily switch from the Polyglot book to the local CTG book (Chessbase softwares) or ABK book (Arena GUI). I've wandered through GitHub to recover the 'OwnBook' checkbox code (it needed just two mere rows..), luckily I've managed to find it and extract the part I needed to use.

This distribution is enhanced by the pgo-build option.

Stockfish Polyglot comes with almost any architecture available, hence also for old 32bit operating systems like Windows XP, old machines, 64bit systems, standard machines, modern machines and power computers. It runs on both Windows and Linux. NOT FOR MACOS OR ANDROID, IF SOMEONE CAN COMPILE THIS SOFTWARE FOR THESE TWO SYSTEMS PLEASE WRITE ME CLICKING 'CONTACTS' ABOVE.


Stockfish Polyglot


Maximize its strength by using my opening book Goi.


Stockfish GitHub page.


Brainfish source code.




- Stockfish_Polyglot_2018-11-13_32bit.exe
- Stockfish_Polyglot_2018-11-13_32bit_general.exe
- Stockfish_Polyglot_2018-11-13_32bit_old.exe
- Stockfish_Polyglot_2018-11-13_32bit_popc.exe
- Stockfish_Polyglot_2018-11-13_x64.exe
- Stockfish_Polyglot_2018-11-13_x64_bmi2.exe
- Stockfish_Polyglot_2018-11-13_x64_general.exe
- Stockfish_Polyglot_2018-11-13_x64_modern.exe
- Stockfish_Polyglot_2018-11-13_x64_popc.exe




Complete changelog:

- Changelog: This is the expansion of the Stockfish version having timestamp "1542051187".

From the previous version of Stockfish Polyglot the following changes/benchmarks were made by the official Stockfish team:

"Replace the PassedDanger array by an equation

This equation seems to do as well as the current PassedDanger array.

Master values were: 3, 7, 11, 20
The new values given by the equation are: 3, 6, 11, 18

LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 84301 W: 18155 L: 18156 D: 47990 Elo -0.00

LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 7940 W: 1358 L: 1217 D: 5365 Elo +6.17

We stopped a LTC run after 70000 games:
LLR: 0.74 (-2.94,2.94) [0.00,4.00]
Total: 70257 W: 11319 L: 11064 D: 47874 Elo +1.26

Bench: 3913185


Remove redundant king square parameter

We don't need to pass the king square as an explicit parameter to the functions
king_safety() and do_king_safety() since we already pass in the position.

LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 69686 W: 14894 L: 14866 D: 39926 Elo +0.14

No functional change.


Simplify tropism. (#1807)

We calculate tropism as a sum of two factors. The first is the number of squares in our kingFlank and Camp that are attacked by the enemy; the second is number of these squares that are attacked twice. Prior to this commit, we excluded squares we defended with pawns from this second value, but this appears unnecessary. (Doubly-attacked squares near our king are still dangerous.) The removal of this exclusion is a possible small Elo gain at STC (estimated +1.59) and almost exactly neutral at LTC (estimated +0.04).

LLR: 2.96 (-2.94,2.94) [-3.00,1.00]
Total: 20942 W: 4550 L: 4427 D: 11965 Elo +2.04

LLR: 2.94 (-2.94,2.94) [-3.00,1.00]
Total: 56941 W: 9172 L: 9108 D: 38661 Elo +0.39

Bench: 3813986


Pawn and Piece Values Tuned at LTC

Failed STC
LLR: -2.96 (-2.94,2.94) [0.00,4.00]
Total: 27487 W: 5846 L: 5903 D: 15738 Elo -0.72

Passed 1st LTC
LLR: 2.95 (-2.94,2.94) [0.00,4.00]
Total: 38503 W: 6270 L: 5999 D: 26234 Elo +2.45

Passed 2nd LTC
LLR: 2.95 (-2.94,2.94) [0.00,4.00]
Total: 34016 W: 5584 L: 5326 D: 23106 Elo +2.64

This pull request lead to an interesting discussion about testing
methodology for Stockfish:

Bench: 3647775


Extension for king moves changing castling rights

passed STC:
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 8463 W: 1919 L: 1747 D: 4797 Elo +7.06

passed LTC:
LLR: 2.95 (-2.94,2.94) [0.00,5.00]
Total: 142590 W: 23263 L: 22587 D: 96740 Elo +1.65

Bench: 3607243


Simplify mobility danger

Check sign only after adding mobility danger term.

LLR: 2.95 (-2.94,2.94) [-3.00,1.00]
Total: 9090 W: 2001 L: 1856 D: 5233 Elo +5.54

LLR: 2.94 (-2.94,2.94) [-3.00,1.00]
Total: 123466 W: 19766 L: 19805 D: 83895 Elo -0.11

bench: 3630207


Stockfish_Polyglot_2018-11-13_x64.exe single processor benchmark:



Stockfish_Polyglot_2018-11-13_32bit.exe single processor benchmark: