Chess2u
Would you like to react to this message? Create an account in a few clicks or log in to continue.

Chess2uLog in

descriptionweekly fun fact (periodical) Emptyweekly fun fact (periodical)

more_horiz



I will post various kinds of fun facts about chess here!

some will be useful
some others will be informative
and others will be just fun facts about chess

we can post all kind of chess related useful fun facts!

***

in the 1st match of world championship
that white's opening is the very opening I use a lot!
it is my favorite opening especially when the opponent is strong
it provides a soild defence
with easy start!

***

this patch is successful
and this update could have been only a cosmetic issue
but it induced actual increase of elo

LLR: 2.96 (-2.94,2.94) [0.00,6.00]
Total: 38968 W: 6113 L: 5837 D: 27018
sprt @ 60+0.05 th 1

update info:
Purely cosmetic issue or relevant to strength?

***

SF with tablebase proved to be detrimental in some ways

the official release of SF syzygy version has seen the dormant hibernation

many ppl say TB inducing poor elo is not true

some ppl argue that TB can bring about the slower searching speed of the engine

we are not sure which opinion is more accurate
but the following is the fact
that none TB version is doing a lot better at the tourey

*

none TB SF is this

N Engine Pts SB St Ko Ho Bo Na Gu

1 Stockfish 021113 6.0 21.00 = =1 == =1 1=
2 Komodo 1133 5.5 18.50 = == = 1= 11
3 Houdini 9601 4.0 16.75 =0 == 1 == =
4 Bouquet 1.8b 4.0 15.75 == = 0 =1 ==
5 Naum 4.6 3.0 13.00 =0 0= == =0 =
6 Gull R600 2.5 10.50 0= 00 = == =

*

with TB is like this

Nr Engine Pts SB Ko Bo Ho St Gu Na Hi Ry Cr Ju

1 Komodo 1121.05 11.5 100.50 =1 == 1= 1= == == 1= 01 1=
2 Bouquet 1.8b 10.0 87.25 =0 == 1= == == =1 == =1 ==
3 Houdini 9601 10.0 86.50 == == =0 == 1= == =1 == 1=
4 Stockfish 151013 10.0 84.00 0= 0= =1 0= =1 =1 == == 11
5 Gull 2.3 9.5 82.75 0= == == 1= =1 0= == == =1
6 Naum 4.6 8.5 76.00 == == 0= =0 =0 1= =1 =1 =0

***

I'm considering which opening book I'll use
PCC opening books are made from world's best matches
to know what is PCC
you can type "PCC chess" to find info in the search engines

I'm using PCC March
and they are releasing the opening book every month!
the newest one is PCC Sep
and PCC Oct will be released soon
PCC March is superb, I will decide if I'll move from March to Sep

descriptionweekly fun fact (periodical) EmptyRe: weekly fun fact (periodical)

more_horiz

a fun fact about SF!

SF must have a new official version number that is 5
actually SF should have gotten that before
there should have been stockfish 5 when the elo boost up was only 30 or so!!!
but now the elo boost up from SF4 is +60 elo
and with the newest version
the elo increase from SF4 is predicted to be about +70 or so
I also predict that the best timing to release SF 5 is
right before the grand final of nTCEC
in that way, SF 5 can gain the most amount of PR without any effort


***

a fun fact about SF and other engines!

who would thought SF can become that strong!

when houdini 3 was released
it was like 100~150 elo ahead of the 2nd best
that was critter or some other engines

and SF became that strong so quickly

most ppl couldn't predict SF could gain that much elo so quickly

and Komodo is famous for playing strong in longer TC matches
and Komodo and SF match is underway
and SF is leading slightly but
the match looks like a draw

N Engine Pts SB St Ko Bo Ho Na Gu

1 Stockfish 021113 6.0 24.75 = == =1 =1 1=
2 Komodo 1133 6.0 24.25 = == == 1= 11
3 Bouquet 1.8b 4.5 20.75 == == 0 =1 ==
4 Houdini 9601 4.5 20.50 =0 == 1 == ==
5 Naum 4.6 3.5 16.25 =0 0= =0 == ==
6 Gull R600 3.5 15.50 0= 00 == == ==

then

Stockfish 021113 6.5
Komodo 1133 6.5
Bouquet 1.8b 4.5

SF and Komodo seem to be in the same league

and who would think bouquet and houdini would be on the same level?

descriptionweekly fun fact (periodical) EmptyRe: weekly fun fact (periodical)

more_horiz




I ran test runs and benchmarking tests

H3 versus SF Nov 7

when no opening book is used
from the basic default starting position
among the very first 20 moves

TC: 1 sec per move, 1 core tourney
14, 16 (depths = the deepest depths searched for each engine)
number of matches produced for one day: 400

TC: 1 sec per move, 2 cores
16, 18 (depths)
number of matches produced for one day: 200

TC: 2 seconds per move, 1 core
17, 18 (depths)
number of matches produced for one day: 200

***
***

just a fun fact
I think match means
a bunch or rounds

so one match means many rounds
1 round means 1 single round

***
***

SF Nov 7 benchmarking

from the default starting position
rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1

TC: 1 sec, 2 cores
depth: 13.5

2 s
14.5

5 s
16.5

TC: 10 sec, 2 cores
18

TC: 20 sec, 2 cores
19.5

TC: 30 sec, 2 cores
20

***

1 core benchmarking
SF Nov 7

from the default starting position
rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1

1s: 14
(400 rounds per day)

2s: 14.5 or 15
(200 rounds per day)

5s: 15
(80 rounds per day)

10s: 16.5
(40 rounds per day)

20s: 18
(20 rounds per day)

30s: 19

***

as a result of benchmarking

1 core = 2 core (almost same depths)

use of 1 core is chosen

***

I have to choose the seconds that will be used

the best option is 1 sec per move
= rapid TC

***
***

there should be the manual for SF
probably in the zip file for the official releases
you know I think there is a big zip file
which contains the official version of SF
along with source files and other things
perhaps in that zip file
there should be the manual
but Aggressiveness, Cowardice should mean
Aggressiveness is being aggressive when SF is winnning
Cowardice is being defensive when SF is losing
that's my guess

***
***

from watching tens of thousands of rounds among engines
for years
I came to the conclusion that
60 moves per 1 min is more flexible
than 1 sec per move
even when the two TCs are same at the first glance
by using 60 moves per 1 min
when the board is crowded with chessmen
engines can think for 2 sec or so
and when the board is not crowded like in the ending
the engines can think faster

descriptionweekly fun fact (periodical) EmptyRe: weekly fun fact (periodical)

more_horiz

these are the rating lists gathered
for an easy viewing for the ppl




weekly fun fact (periodical) 1qrvqw10


***


weekly fun fact (periodical) 2qwbtq10


***


weekly fun fact (periodical) 3qetbq10


***


weekly fun fact (periodical) 4qbtqq10


***


weekly fun fact (periodical) 5qbwtq10


***


weekly fun fact (periodical) 6qbtqf10

descriptionweekly fun fact (periodical) EmptyRe: weekly fun fact (periodical)

more_horiz



this thread is all about and dedicated to fun facts of chess!

***

ppl usually think SF's eval is overblown and exaggerated a lot
but after the recent patch
SF's eval seems to be close to what is considered to be normal

***

SF can sometimes search extremely deep sporadically
when in certain situations

for instance,
if the average searching depth in the middle phase is 20 plies
sometimes SF can search for 50 plies
when certain board situations kick in

***

after watching strelka 5.1's matches
strelka's eval is super exaggerated

in one board position
Houdini 4 says
H4 is losing by 0.7 eval
when Strelka's eval shows that strelka is winning with 2.6 eval

in another board position
SF's eval shows
SF is winning by +4.0 eval
and strelka 5.1's eval says
it is losing with -6.0 eval


descriptionweekly fun fact (periodical) EmptyRe: weekly fun fact (periodical)

more_horiz
Permissions in this forum:
You cannot reply to topics in this forum