back


30338noname
2020-07-10 15:37:32
@how white wins ?, @ queen13. We will block you for life if you continue to publish such crap. Buy dog food like @Ahithopel Ebejer aka WCN (thearbchess), and get full (happy?).
30337how white wins?
2020-07-10 13:30:51
Updated Updated Updated list from 30295:

naughty David top after complaint to arbiter

how white wins? relegated to div 2
freaking corrupt Ingo disqualified for attempted bribery
Annoyed Mrert in prison for taking bribes
strange ideas queen13 disqualified for spreading fake rumors stockfish has opening experience
30336WCN
2020-07-10 11:19:40
THE TCEC SCAMS! NEVER STOP!!! :(

Now let's look at the Second SCAM! game where the Stockfish playing is not the real Stockfish :(

[Event "TCEC Season 18 - Top Of The Top Bonus"]
[Site "https://tcec-chess.com"]
[Date "2020.07.07"]
[Round "14.2"]
[White "Stockfish 202007032109"]
[Black "LCZero v0.26.0-sv-t60-4229-mlh"]
[Result "0-1"]

https://groups.google.com/forum/#!topic/fishcooking/D2DbkWkP3K0




WCN
30335how white wins?
2020-07-10 06:21:49
now everyone counting their comments
30334how white wins?
2020-07-10 06:15:13
it will surprise everyone that stockfish need no opening book
lol
you should give up chess
and stick to counting your comments and mentions :-D
30333how white wins?
2020-07-10 05:56:42
your ideas stockfish has openings knowledge by playing games in fishtest is very strange
you very strange ideas
not right at all
ab engines need opening book
have no opening experience
30332how white wins?
2020-07-09 20:26:26
'>It would be more correct if Stockfish could use their own opening book as Brianfish does it.'
corres say it not me
i quote him
i agree with him

i not agree with anything you say
you completely wrong about all
30331WCN
2020-07-09 19:32:43
THE TCEC SCAMS! NEVER STOP!!! :(

https://groups.google.com/forum/#!topic/fishcooking/D2DbkWkP3K0


OK! So let's look at the first SCAM! game where the Stockfish playing is not the real Stockfish :(

[Event "TCEC Season 18 - Top Of The Top Bonus"]
[Site "https://tcec-chess.com"]
[Date "2020.07.05"]
[Round "3.2"]
[White "AllieStein v0.7_dev2-net_15.0"]
[Black "Stockfish 202007032109"]
[Result "1-0"]

https://groups.google.com/forum/?hl=en#!topic/fishcooking/D2DbkWkP3K0




WCN
30330queen13
2020-07-09 19:25:38
> does stockfish carry information of these millions of games in a multimegabyte weights file?

It carries it in its evaluation and search functions and possibly in many other parameters

which have been tunes for stockfish strength by trial and error.

> stockfish had chance

Stockfish lose the previous 2 finals. There have been discussion in tcec about

Problematic net but I don't remember who said and what exactly happened.

> not surprising stockfish always disadvantage

It's slightly surprising to me. I would expect lc0 to fair much better than what it did

in final but not be better than stockfish ! Unless it is the same net in which case it's

"probably" the conditions. We do have to keep in mind that tcec suffers from

too few games to draw solid conclusions upon. Maybe stockfish was lucky in the final.

The results with 100 games are meaningful in the +-40 or +-50 range(I think it's

something like that, I might be mistaken and it could be +-25 but even this lower

bound does leave a non negligible chance that they are of equal strength)

The results in the current event suffer much more as too little games have been played.

Maybe stockfish was unlucky and maybe given both results that's what one should

conclude(although as I said no solid conclusions can be made.)

> does stockfish remember experience of openings?
you think yes?

Lc0 doesn't remember either then. It just thinks that those moves are good and plays them

"just like stockfish" although it works in different ways.

>then stockfish not need opening book at all like lc0

Stockfish is weaker than lc0 in the opening and stronger in the endgame.

That doesn't make it unfair because no opening was used.

Lc0 plays better with opening book and would be playing a very strong midgame

had it been trained with an opening book. Would you then say that stockfish would

need a middle game book? It's just that the way that lc0 works, it improves the first

stages of the game first/better and then the rest. (at least that's my primitive

understanding of it...)

> It would be more correct if Stockfish could use their own opening book as Brianfish does it.'

To which one may simply reply that it would have been fair if lc0 have been trained with

opening book. Lc0 does not use an opening book. The weights are how it's artificial neurons

will get excited. That's what they determine. It's like the parameters of its eval function

It changes how it will evaluate positions. It would be like saying the stockfish

parameters give it an endgame book. It would only be fair if lc0 had access to 16 piece

tablebases...

I don't think that using openings is unfair, I would prefer it because

chess engines play better when they use those strong moves that we already know are

as good as it gets until someone improves them(which happens all the time) but I

wouldn't call not using them unfair.

Stockfish wasn't created in such a way as to rely on opening books.

It was created in such a way as to rely on endgame tablebases for simpler endgames

so maybe taking the tablebases out of the equation would be unfair
30329how white wins?
2020-07-09 19:25:15
you me disagree
corres i agree
30328Lalit Mohan Bankoti
2020-07-09 19:20:27
Stockfish NNUE badly heated by Komodo in CCC.
30327how white wins?
2020-07-09 15:21:04
'stockfish does have opening experience of millions of games.'
are this for testing patches?

'Stockfish played millions of games in fishtest in order to determine which changes are good'
does stockfish carry information of these millions of games in a multimegabyte weights file?

some games in sufi had 9 or more book moves
stockfish had chance
topoftop has 1-ply move
not surprising stockfish always disadvantage
does stockfish remember experience of openings?
you think yes?
then stockfish not need opening book at all like lc0

see corres say this talkchess
http://talkchess.com/forum3/viewtopic.php?f=2&t=74421
Thu Jul 09, 2020 11:38 am
'I do not like such competitions in what the main aim to demonstrate only the power of Leela with "Openings book with one move" (?).
It would be more correct if Stockfish could use their own opening book as Brianfish does it.'
30326queen13
2020-07-09 14:11:32
It was discussed in tcec, allie would show high scores when lc0 would show

a very small score. Also, we can now see that stockfish is having trouble

against both Allie and lc0. Clearly if this net of lc0 beats stockfish

then it's relatively safe to say that it is stronger than the net in the final

Stockfish does have opening experience of millions of games.

Stockfish played millions of games in fishtest in order to determine which changes are good

and which are bad. So stockfish has experience of the opening as each time an opening

improvement was made stockfish would gain better opening experience.

Lc0 has opening experience because it did the same but it learned on its own

without humans.

I wonder whether they should use opening books in the training of nets.

It could make the middlegame out of the book very strong.
30325how white wins?
2020-07-09 11:11:21
does stockfish have opening experience of millions of games?
how you know sufi Lc0 net weak/problematic/buggy?
30324queen13
2020-07-09 10:03:32
> stockfish elo would much increase after sufi win
losses in tcec topoftop will falsely reduce elo

I am not sure what you mean.
I could say stockfish elo falsely increased after sufi with weak/problematic/buggy
Lc0 net. We are getting closer to the correct elo now that stockfish loses against
their real strength.
30323queen13
2020-07-09 10:00:39
> i not happy tcec match stockfish without book against 3 nns with opening experience

Stockfish has opening experience. Did the devs not improve stockfish in the opening

through all these years?
30322David
2020-07-09 09:42:52
@ 30317
Stockfish nnue is a different system, similar to Sugar NN.
Instead I say to use the network that they do: beaten with his own weapons.
In theory, a complex system could be made that uses both traditional and neural systems in parallel in the same engine.
30321how white wins?
2020-07-08 22:53:43
stockfish elo would much increase after sufi win
losses in tcec topoftop will falsely reduce elo
30320how white wins?
2020-07-08 20:41:29
is tcec stockfish vs nns match deliberate uneven?
30319how white wins?
2020-07-08 19:37:51
fishtest not saved stockfish games?
what sad waste
make me sad
stockfish devs could easy do stockfish nn if want
do they want?
other stockfish nns now
one analyzes at tcec, stockfish NNUE
there are others
always name stockfish :)

i not happy tcec match stockfish without book against 3 nns with opening experience
no not right
did they announce no book stockfish vs nns match?
if 1-ply first tcec move, should all have cerebellum book
equal
30318queen13
2020-07-08 18:55:55
I feel confident that if the stockfish team want to make NN engine that can compete with the

Strongest NN engines, they can certainly do it. Stockfish is much more difficult to make

It took a lot of effort and a lot of years to create in its current form.

What I would like the stockfish team to do is to save the stockfish games in fishtest.

If they had done so from the beginning they would have had many many games

and the learning algorithms of today need a lot of data. The more the better.

They could prove useful in training NN engines.

I don't to the extent that one could call stockfish, stockfish when it changes so dramatically

from a-b to NN... It seems to me that it would be a different engine.

Maybe call it stockfish NN to keep the name and indicate where it comes from :)
30317how white wins?
2020-07-08 18:10:26
exist now stockfish nns
look at talkchess
30316David
2020-07-08 14:11:24
Obviously it is easy to use a hex editor and change a name, but creating a real neural version is a translation job for competent people.
This would not detract from the classic version which is excellent and fundamental.
30315David
2020-07-08 13:57:15
People love new things.
Can a skilled programmer create a "version of Stockfish" (preserving its fundamental characteristics) capable of using Leela's networks?
It would be very interesting ...
30314WCN
2020-07-08 10:06:23
THE TCEC SCAMS! NEVER STOP!!! :(

https://groups.google.com/forum/#!topic/fishcooking/D2DbkWkP3K0




WCN

next page >
single page