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

Chess2uLog in

descriptionstockfish forecast Emptystockfish forecast

more_horiz

http://tests.stockfishchess.org/tests

I think it would be a good idea for somebody to do the forecasting
of the new releases of the new engines
like the weather forecast

it's gonna be like the weather forecast
it cannot be 100% right
but it'll be reasonably accurate

I can do the forecasting of the new releases of SF
based on the data and info I can see
from the testing framework site

I can analyze the data and info
along with the numeric things
to forecast when a new version will be released
(how many days we should wait for the new version to be released)
how much elo gain will be there
what kind of improvements there will be in the new releases
and so forth

it is like the weather forecasting ppl analyzing data
to make a prediction on what weather they will have later

this SF forecasting will be like that!


***

perhaps many of us just waited indefinately for a new SF to be released

we just waited without knowing when a new version will be available

so that was uncertain and precarious
but, now we don't have to just wait for it uncertainly
we will know in advance when a new version will come out!

***


the following is my speculation based on the data and info
from the site

(this is just my first prediction
so the accuracy of the prediction will go up gradually
as I accumulate more know-hows on how to do the prediction)


***


the following is the speculation

there will a new version that will contain these functions

LTC: more bonus for safe checks

Adapt SE margin to move gain
(almost included, but cancelled)

Bonus for more advanced chained pawns. I use 1/3 value of CandidatePassed[] array as a starting point.
(almost included, but cancelled)


this new version (1st version) will be released within one day
from the time I'm writing this

and there will be some elo gain (a few points)

but after this new release
there will not be a new release for at least 2 days
since many of the newer tests show
negative LLR results

as more tests are done
and as more LLR results are known
we will know how many days we have to wait
until the new version (2nd version) will be released

legend:
1st version means the new version after Oct 15th
2nd version means another newer version after the 1st version

descriptionstockfish forecast EmptyRe: stockfish forecast

more_horiz

new forecast
this is the up to dated new speculation and prediction on
when SF will be released!

***

I thougth the new version will be released shortly
but the new version's release have to wait for at least 1 more day

due to the following reason

previously
one of the following functions (you can see
which function I refer to if you compare my previous forecast
with this new forecast
or you can directly compare the engine testing process from the site)
is thought to be not good
but after more tests on that function
it seems that it turned out to be good
(the new dev versions are under test
and whether the new function is good or bad
we don't know cuz the function is not fully tested yet
and the function's authenticity is not finalized and confirmed by
all layers of tests)


***

new functions:

more bonus for safe checks take 2

Chained pawn, Take 2. Reduce basis value for a chained pawn to compensate for the rank bonus.

Chained pawn, Take 3. Increase rank bonus.

***

the basic algorithm of testing looks like

take 1 -> take 2 -> take 3

if a new function is made then
that function is tested under short TC

if the result is positive

that function must go thru another long TC test

only when the function can pass even the long TC

that function is embedded into the code

moreover
only functional changes (elo gain accrued)
are the ones we can predict I guess
I think non-functional changes (no elo gain)
is not mentioned on the site

that's the guess
that is what it seems

***

so one of the fuctions above
was thought to be not good
but after more tests
that said function seems good
so more testing will be done

and after more testing
that said function and some other functions
will be embedded to the final code
so the new release of engine will take like 1 more additional day
and possibly up to a few days
from the previous forecast (1 post above)

descriptionstockfish forecast EmptyRe: stockfish forecast

more_horiz

***

update:

(cont'd from the post above)


I thought
"so the new release of engine will take like 1 more additional day
and possibly up to a few days"

but from the latest data and info from the site
it looks like the new release is imminent
so I updated that quoted statement (prediction)

from "so the new release of engine will take like 1 more additional day
and possibly up to a few days"

to "so the new release of engine will take like 1 more additional day"


***


I thought the testing framework is normal cloud network
but I think it is not the same kind of cloud network
I think the framework is making use of traditional cloud technique

I think in juxtaposition of many sort of clouds
we can dichotomize cloud into traditional cloud and next-gen cloud
- traditional cloud
- next-gen cloud

***

there will be next-gen consoles
and the next-gen video consoles I think can utilize next-gen cloud technique
by that
I guess it kind of means that
if there are 100 next-gen consoles (1 console has 8 cores)
all the 100 consoles can be interconnected so
100 consoles act like one giant supercomputer
so that one giant supercom (a conglomeration of 100 consoles) is like
one com with 100 x 8 = 800 cores

so that one giant com with 800 cores can run one chess engine

and I guess this is my guess on the next-gen things


also new SF engines gained more elo
but some can call it next-gen software-wise
and some don't call it next-gen

new things on the next-gen video consoles have more polygons
so they look sharper and better graphically
but ppl don't just say
the new things are next-gen just cuz the new things
look better in terms of graphics

ppl say the new things on the new next-gen consoles are next-gen
cuz the new things have more than just better graphics, like
- better character animation
- better particle effects
- better AI of char
- better story
- better physics
- new things and new concepts which were impossible with old-gen hardware

likewise, on the same notion with above
I guess SF will be called next-gen stockfish by most ppl
when SF can do something really next-gen

gaining just more elo can't satiate the requirement to be next-gen

there should be more, like
the new engine should
- do next-gen positional play
- have human like thinking
- do next-gen tactics and strategies

it is my opinion that I think winboard 4.7 is the next-gen GUI
WB 4.7 has a totally different graphics that is amazing and good
WB 4.7 (or other recent versions) has a powerful tourney managing tool
cuz of these 2 prominent and overt functions
WB 4.7 is the next-gen WB GUI

if SF can have something or some function that stands out of the crowd
SF is next-gen!


I think the stockfish framework doesn't use this next-gen cloud
I guess stockfish framework utilizes the traditional cloud



stockfish forecast Vvasrq10



take a look at that diagram
there are many computers (with many cores) which are active and connected
but they don't act like one giant supercomputer

each computer there is working on each engine separately
- one com is running SE
- another com is running safe check
- another com is running chained
- and so on

so I think it looks like it is traditional cloud technique

descriptionstockfish forecast EmptyRe: stockfish forecast

more_horiz



http://tests.stockfishchess.org/tests

if we read and analyze the content on this website for a few days
we will know every tidbits and detailed info on how it works
and we will know when a new update will be released!

so we can predict when and what kind of update will be made available
if we read this homepage for some time


***


here is a tip

a new change will be tested for 15 sec per match STC
then if it passes
it will go thru the next stage that is 60 sec per match LTC
if it passes this LTc
it usually becomes the new update
but sometimes there are exceptions
even if a new change passes all the stages
it can go thru extra testing if needed
and even if it passes all the tests
it doesn't become the new update
when it is considered and deemed not good in the long run for SF's growth

like 10 days ago
there was a new patch
on variable contempt
raise contempt if SF has more time and something like that
that new patch has big win/loss ratio
so that one patch alone implies 10 elo boost up
but it is not applied
I don't know why but probably cuz of the patch being not suitable

***

so if we just read the websites for a few days
then we can become the master in stockfish forecasting!


descriptionstockfish forecast EmptyRe: stockfish forecast

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