Is EMAN the future number one?

[Dann Corbit]

He's already using the testing system of Stockfish.
He backpatches all the SF improvements every time he does a release.

[corres]

He's already using the testing system of Stockfish.
He backpatches all the SF improvements every time he does a release.

If he backpatches all SF improvement why he needs Stockfish testing system?
As I know, Stockfish testing system used by the developers of Stockfish.

[Dann Corbit]

Yes, of course, but why is he back-patching? Because slowly, reliably, Stockfish gains 50 Elo per year through dozens of programmers and dozens of testers working tirelessly to improve it.
He is back-patching so that he can harvest the effort of the programmers and the testers and the hardware resource of many, many powerful machines.

[Ovyron]

Only in some lucky cases we can hope for some enhancement in Elo.

It's not luck, is adapting the code to the situation.

I already mentioned an easy improvement: add code that detects when the engine is out of book, and for that situation, pretend you have more time than you do to take longer to move and fill up the hash. Once you have this, have in the engine a setting that tells the engine for how much moves this is done.

A code like this will make the engine slightly stronger, though as Stockfish dev keeps improving, it'll catch up quick.

A code like this will never make it to Stockfish master, because if on the testing all engines come up out of book at the same move, then the code is counter-productive. It's only productive in those situations where sometimes the book ends early, and sometimes the book ends late, so if you have a deep book, your side will gain more time on the clock than the opponent, and this code will help. If the opponent outbooked you and your engine is behind on the clock, playing better moves is critical because it's likely you've fallen in some kind of opening trap, so playing quick out of book is a bad idea.

But a Stockfish playing another in identical situations, with colors reversed (like in a laboratory) will just hurt itself by doing this, because it's just going to have less time on the clock for the rest of the game overall, so this patch that benefits real life situations hurts in the laboratory.

MADE UP EXAMPLE:

If that's not convincing, what about this?: The programmer adds code that makes the engine play stronger from the white side for 1.d4, and 1.c4 openings, and worse for the rest. A player that uses this version and plays 1.d4 and 1.c4 exclusively will have an edge over all other Stockfish users, and will want to use this engine, because it's the best thing he can do to improve this performance.

But in the laboratory, 1.e4 and other openings will be played, and with an abysmal performance on those, this engine will appear much weaker than default.

But nothing stops people to stick to the openings that work best for them and ignore the rest, so this is another case where real life scenarios can't be tested in the laboratory, you could make an engine that is stronger than Stockfish default on the laboratory, and it will not have been about luck.

But clearly Khalid hasn't done so, latest Brainfish is performing better than Eman in the real world, so whatever improvements Eman has aren't doing anything significant.

But... But... But...

[carldaman]

Hope to see how it performs on Playchess (we have free accounts, come join us)

I actually had a book that did decently with 1.c3, it was a draw machine. I had to abandon it because there were no wins in sight, I couldn't even beat Fritz 15

http://talkchess.com/forum3/viewtopic.php?f=2&t=74186

SultanKahn 2.0 plays without a book, and is actually offended at the idea of having one imposed on him/it. Just like his human predecessor he's learning the openings from scratch. Playchess may be too much of a future shock for SK2, but not that he couldn't handle it, strength-wise.

[Ovyron]

SultanKahn 2.0 plays without a book, and is actually offended at the idea of having one imposed on him/it.

The idea of the book is to add variety. A human may play 1.e4 some times and 1.d4 some times. An engine might be stuck playing the same moves over and over and we never get to see how it plays other ones.

If the engine has something implemented inside to vary its openings, then it doesn't need a book.

[carldaman]

Right now the variety comes from within, sometimes the pawn goes to c4, other times to c3, even a3 (Anderssen's Opening)! It's an evolving entity, almost wrote "living", so we're letting it evolve for now.

[corres]

Only in some lucky cases we can hope for some enhancement in Elo.

It's not luck, is adapting the code to the situation.
I already mentioned an easy improvement: add code that detects when the engine is out of book, and for that situation, pretend you have more time than you do to take longer to move and fill up the hash. Once you have this, have in the engine a setting that tells the engine for how much moves this is done.
A code like this will make the engine slightly stronger, though as Stockfish dev keeps improving, it'll catch up quick.
A code like this will never make it to Stockfish master, because if on the testing all engines come up out of book at the same move, then the code is counter-productive. It's only productive in those situations where sometimes the book ends early, and sometimes the book ends late, so if you have a deep book, your side will gain more time on the clock than the opponent, and this code will help. If the opponent outbooked you and your engine is behind on the clock, playing better moves is critical because it's likely you've fallen in some kind of opening trap, so playing quick out of book is a bad idea.
But a Stockfish playing another in identical situations, with colors reversed (like in a laboratory) will just hurt itself by doing this, because it's just going to have less time on the clock for the rest of the game overall, so this patch that benefits real life situations hurts in the laboratory.
MADE UP EXAMPLE:
If that's not convincing, what about this?: The programmer adds code that makes the engine play stronger from the white side for 1.d4, and 1.c4 openings, and worse for the rest. A player that uses this version and plays 1.d4 and 1.c4 exclusively will have an edge over all other Stockfish users, and will want to use this engine, because it's the best thing he can do to improve this performance.
But in the laboratory, 1.e4 and other openings will be played, and with an abysmal performance on those, this engine will appear much weaker than default.
But nothing stops people to stick to the openings that work best for them and ignore the rest, so this is another case where real life scenarios can't be tested in the laboratory, you could make an engine that is stronger than Stockfish default on the laboratory, and it will not have been about luck.
But clearly Khalid hasn't done so, latest Brainfish is performing better than Eman in the real world, so whatever improvements Eman has aren't doing anything significant.

So Khalid try to make optimal the playing of Eman for short time control, like Sugar.

[corres]

[quote="Dann Corbit" post_id=847274 time=1592161335 user_id=58]
Yes, of course, but why is he back-patching? Because slowly, reliably, Stockfish gains 50 Elo per year through dozens of programmers and dozens of testers working tirelessly to improve it.
He is back-patching so that he can harvest the effort of the programmers and the testers and the hardware resource of many, many powerful machines.

[corres]

...
He is back-patching so that he can harvest the effort of the programmers and the testers and the hardware resource of many, many powerful machines.

Yes, he (the only one author of Eman... ) does it.