Re: when a daily tester should review his tests

[deeds]

Thank you all for warning me about this other Sedat's lie.

When Sedat thinked he created an exact copy of my EXP files without my Eman's learning/tourney games, without the games/openings i reinforced at D40, without the experience data i manually updated, without the key position i analyzed at D70, etc :

...
In Deeds test: Depth4 exp managed to win all His opponents (opening books)
...
Under SCCT: almost all books are performed better than Depth4...
...

...
One thing more, it looks likes under SCCT conditions:
DeeDs techniques are NOT best techniques for sure !!
...

And when Sedat confessed his lies above :

...
your work is never played in SCCT
...

A little bonus, here are some examples of the disastrous SCCT conditions :
https://outskirts.altervista.org/forum/ ... ost#p66382
https://outskirts.altervista.org/forum/ ... ost#p66383
You are not dreaming, Sedat's settings are so bad that it even disrupts the engine's evaluations !

[deeds]

...
Since he is not so much experienced (maybe 3-4 years in testings)
...

Oops, twice as experienced

...
Deeds target is now become more clear for sure, He tries to sell Exp
...
it seems He doesn't want to gain much Euro
...

Nonsense If you don't want to gain much money, you don't try to sell something.

Reminder : I discovered the Sedat's unfair behavior on July, 2022 and I opened the sales of my first EXP file on April, 2023. So what Sedat considered an attack could not have been an attempt to sell something.

[deeds]

For those who want to have a laugh :

unfair tourney

[sarona]

Well. . . that data is not news to me.

He clearly did not understand the error of his ways when running that test (although, it appears he FINALLY did realize his oversight later on)


You want to have another laugh?
You will be pleased to know that AlexChess's ProteusSF is second-last (17th place) in his latest Championship posted here:
viewtopic.php?f=4&p=38235#p38235

Why would any experienced tester (and, no doubt, Sedat is a veteran in computer chess) include that thing in a serious tournament? Why not ShashChess?

P.S.
And no, I am not deeds using another fake account. Although I wish I was in France right now. Pretty cold here with snow. . .

[deeds]

He clearly did not understand the error of his ways when running that test

In any case, the ground is starting to shift under Sedat's feet.
Previously his own supporters/participants denounced the conditions of his tournaments, that his behavior no longer suited them and that his messages were polluting.
Sedat either banned them or added them to his foes list.
And now his comments are censored :
https://chess.massimilianogoi.com/forum ... 3200#p3200
His thread was even locked :
https://chess.massimilianogoi.com/forum ... 3206#p3206

(although, it appears he FINALLY did realize his oversight later on)

After Rulleeeee, you are the 2nd person to think that !?
For me, Sedat is still the least enlightened TD in terms of EXP files / learning engines.
But maybe you or Rulleeeee have found something in all Sedat's messages, because I haven't...

Why would any experienced tester (and, no doubt, Sedat is a veteran in computer chess) include that thing in a serious tournament ?

I have already explained several times why specialized TDs in ENG+BOOK tournaments sabotage the performance of learning engines...

Why not ShashChess ?

https://chess.massimilianogoi.com/forum ... 3208#p3208

[deeds]

...
although, it appears he FINALLY did realize his oversight later on
...

I doubt that Sedat understands that he is ranking nets rather than learning methods.

How long will it take Sedat to understand that Brainlearn 30 is embedding one of the very latest 2024 nets when Eman 9.51 is embedding one of the first 2023 sfnnv7_compressed nets ? So we don't know if the engine won thanks to its net or thanks to its learning method...

if there are 60 openings, it would take at least 30k games/engine for each engine to start learning something in each opening. But since there are 18 engines in a round-robin, it would take 540k games. And this is where we understand that Sedat has again used a concurrency > 1 with learning engines

[deeds]

...
if there are 60 openings, it would take at least 30k games/engine for each engine to start learning something in each opening. But since there are 18 engines in a round-robin, it would take 540k games. And this is where we understand that Sedat has again used a concurrency > 1 with learning engines

In 2024, after years of explanations, proofs, tests, experiments, who still tests the learning feature of an engine with a concurrency > 1 ?
Only Sedat :

source

In 2024, after years of explanations, proofs, tests, experiments, who still tests the learning feature of an engine with only 50 games per opening ?
Only Sedat :

source

[sarona]

I knew about the issue with ShashChess.

I actually wrote that post in order to see if I could get a rise out of him and that he would then (predictably) praddle on in future posts about trolls and how terrible he is being treated on the fora. I see he is doing that on another forum with 62 members. Thanks for not letting me down.

I also did not want to be mean and mention the fact that such an experienced engine tester was scammed by including the (fake) Eman 10 in his test. I am sure he will eventually figure that one out, too.

After Rulleeeee, you are the 2nd person to think that !?
For me, Sedat is still the least enlightened TD in terms of EXP files / learning engines.
But maybe you or Rulleeeee have found something in all Sedat's messages, because I haven't...

Chris, I seriously doubt he has learned anything about experience files. That post was pure sarcasm in order to provoke a reaction. People can test engines for 20 years and still be ignorant of the proper methods required. He doesn't even comprehend that either.

[deeds]

At first, I didn't know Sedat so my first feeling was of pain for him as a TD to see his ignorance in absolutely all areas of learning engines / EXP files.

Since none of his supporters explained his mistakes to him, I understood that in fact they were all afraid that learning engines / EXP files would make all their work on BIN books obsolete.

So, I was interested in the openings most covered by BIN books and I trained Eman :


After each learning session, I update a small BIN book with the most effective/played moves up to the 60th ply so not only when I use my EXP file in tournament mode (experience book = true), it is less and less of an "experimental" feature (as Khalid said) but I also learn about their world of BIN books.

When I see that they are still at the starting point with < 500 games per opening, engines configured in training mode (experience book = false) during their tournament, EXP files that do not match the engines, concurrency > 1, hash size < 2 GB, untrained openings, I enjoy waking them up from time to time because it gives me ideas for coding new tools.

[deeds]

I was told that here is the new Sedat's lie :


source