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

Chess2uLog in

descriptionhow to run a super-amazingly customized tourney  Emptyhow to run a super-amazingly customized tourney

more_horiz

this is about book depth, max number of moves (move limit), 50 move rule
this has very deep analyses and in-depth info
about the tourney settings
and engine parameters
and the mechnism behind engine matches

***

book depth: 16

the book depth I use is 16
ppl usually use like 12
and 16 is like 4 more than 12
but I found out 16 is okay
depth of 16 proved to be very good
I have watched hundreds of matches and perhaps 1000 matches
with this setting of 16

the result should be somewhat different from book to book
if we run 100 matches with book depth of 16
from my observation
85 matches on average are perfectly started
that is approx. 85 matches will have eval of 0.3 or less
as soon as the book is no longer used
(= from the move number 17)
so that score of 0.3 or less is natural
and no one side is penalized heavily from the get-go
by having evals like -0.8 or something like that from the move number 17
for the rest 15 matches
the eval diff is something like a little more than 0.3
mostly slightly more than 0.3
so that's quite reasonably okay

***

50 move rule: 16 move rule

when using 50 move rule
the matches usually have a long winded movements back and forth
without certain purpose
both sides usually just move around pieces back and forth
hither and thither without certain goal to earn more time
and the length of the matches will be very long
I have observed and watched carefully for more than hundreds of matches (perhaps 1000)
with this 16 move rule setting

16 move rule is working like a magic
if I may use the figurative speech
or to simply put
16 move rule is working well almost without salient flaws or errors
and it is working very well from my observation

*

as I mentioned above
the book depth of 16 is working 100% perfect for 85 out of 100
and the depth of 16 is working reasonably acceptably well for 15 out of 100

so let's extrapolate the score of the setting
85 * 1.0 + 15 * 0.5 = 85 + 7.5 = 92.5

the usage of book depth of 16 has the score of 92.5%
which is like A and just below A+ score

*

and as for this 16 move rule
this 16 move rule setting is also working well
from my observation

for 100 matches I have carefully obversed
about 94 matches were intact and perfect
with using 16 move rule

but for other 6 matches
the result was interfered
that is the match result could have been a win
but it results in a draw cuz of this 16 move rule

for example,
if white was leading with +3.0 score
at the ending phase
but the board position then is a zugzwang or a bottleneck
so even if white was leading with 3 points
after 16 moves
the match became a draw from the automatic GUI adjudication

if white had 34 more moves (hence total of 50 moves to go)
white could have solved the bottleneck or zugzwang situtaion
there is a chance that white can't solve the bottleneck or zugzwang
even after 50 moves
but, there also is a chance that white can solve the bottleneck or zugzwang
before the 50 moves
it is half and half

so for 6 matches of 100
about half of the match result became draw
when the actual result should have been a win
so 3 matches were interfered
and other 3 matches were intact

and as for the remaining 94 matches
the match result was perfect and intact
from my observation
so no problem there

so let's postulate the score of this 16 move rule setting
94 matches = 100% = 94 * 1.0 = 94%
6 matches = 3 is 100%, 3 is 0% = total is 3%

so the holistic overall score is 94% + 3% = 97%
and that is a sure A+ score

***

maximum number of moves: 120

the maximum number of moves is 120
many ppl will think 120 is way short
there will be so many matches which go over 120 moves!

at the first glance
it seems like that
but keep in mind that we have the 16 move rule in place
16 move rule and 120 rule work in tendum and cooperation with each other
perfectly well in harmony

engines have contempt and contemptscore parameters
and this is the most important aspect when it comes to this topic

*

let's say if we have the standard 50 move rule
and max of 200 rule
which are generally used

we need 200 rule cuz we have 50 move rule
to elaborate more

cuz we are usually using 50 move rule
the engines will make back and forth,
hither and thither moves without certant and overt purposes
a lot for a long period

that's why when we use 50 move rule
engines make a lot of those
so-called "back and forth hither and thither moves without any certain overta purpose"
(by the way, does any one know the proper name or terminology for this?)

let's just shorten that name until we come up with other better names
let's just use "back and forth moves"

when we use 50 move rule
engines make so many back and forth moves
that's why we need max of 200 rule

but if we use 16 move rule
then engines don't make such back and forth moves a lot
since the engines can calculate usually over 16 moves
if they make such back and forth moves
the engines know that they will have a draw
and this is something the engines don't like
having a draw is something the engines don't like
so the engines will not try to do the back and forth moves

*

having a draw is something the engines don't like

and this is a very important statement
and there is more to than meets the eye
if black is losing and white is winning
black wants a draw
so black will make lots of back and forth moves
but white will not let it happen
white will not do back and forth moves
white will be the one which doesn't do back and forth moves

cuz this statement
having a draw is something the engines don't like
holds true 100% all the time

when the match is even and balanced out
(= when the eval is close to 0.0
and not in favor of any one particular side)
both engines don't want to draw)
there are 2 engines which don't do back and forth moves

when the match is not even
that is one side is winning
and another side is losing
the losing side is playing for a draw (it wants a draw)
so the losing side will do back and forth moves a lot
but the winning side will never do the back and forth moves
since the winning side doesn't want a draw

so when the match is not even
there is at least one side
that doesn't want a draw

so in any case (among the two possible senarios)
the statement
having a draw is something the engines don't like
holds true 100% all the time
and works perfectly well without any error or flaw

*

in conclusion,

cuz this statement
having a draw is something the engines don't like
holds true all the time

16 move rule and 120 rule work in tendum and cooperation with each other
perfectly well in harmony

so cuz we have 16 move rule
the engines rarely make the so-called
back and forth moves

so we don't need max of 200 rule
max of 120 rule is sufficient most of the time

max of 120 rule is more than enough most of the time

from my observation of watching hundreds or perhaps 1000 matches
with these settings

if we run 100 matches with max of 120 rule (along with 16 move rule)
almost all 100 matches don't go over 120 move limit
there are only a few matches that go over that 120 move restriction
but those are only a few
and when we run like 300 matches for one engine
that (a few matches exceeding 120) doesn't make any visible difference in elo ratings

but here is a catch
this is another point
a few matches among 100 matches will go over 120 move limit
but those a few matches are almost always drawish anyway
so even if there is no max move limit at all
those long and prolonged matches will be mostly draws anyway

usually only those matches which are quite even and in a bottleneck
will reach 120 move limit

*

so when we surmise the score of this maximum number of moves being 120

almost all the time, the matches end before 120 moves
= 100% score

only for some rare exceptions (a few matches per 100 matches)
the matches go over 120 limit
but most of these are drawish
and draws anyway
= almost 100% score

we merge the two aforementioned factors
and we get 99% or so

the overall score is 99%

***

book depth of 16 has the score of 92.5%

16 move rule is used
instead of 50 move rule
and the 16 move rule's score is 97%

the score for
maximum number of moves: 120 is 99%

***

as a result of the combination of all these settings
we can have a match faster
the total length of match on average is reduced by about 30%
and the total time needed to run a tourney with hundreds of matches is decreased by also 30%
while the accuracy of elo ratings is almost intact and perfect
so if we needed 10 hours to finish a tourney
now we only need 7 hours
and that's something important

descriptionhow to run a super-amazingly customized tourney  EmptyRe: how to run a super-amazingly customized tourney

more_horiz
Nice post thorhammer. thumb up

I love reading your posts. It's great to be able to interact with someone who knows what he's talking about.
I am always checking back to see your posts and articles.
Permissions in this forum:
You cannot reply to topics in this forum