Cerebellum, correspondence chess and me

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Rodolfo Leoni
Posts: 544
Joined: Tue Jun 06, 2017 2:49 pm
Location: Italy

Re: Cerebellum, correspondence chess and me

Post by Rodolfo Leoni » Tue Jan 30, 2018 12:03 pm

MikeB wrote:
Rodolfo Leoni wrote:
pijl wrote:
Jouni wrote:What do You think about this message in SF forum:

"sign up with ICCF and let stockfish make ALL the moves with no intervention on human part - in all probability your rating will not go over 2200"
This statement is probably true, although it may depend a little on the amount of time/processors you have at your disposal
Jouni wrote:I don't believe this with 3300 engine!
A matter of comparing apples and oranges. The rating scales are not comparable.
I guess truth is halfway. I won some games because opponent "centaur" couldn't see beyond horizon effects. A monster hardware helps, but it's often not enough. Human skills still play a role, for now. But if someone finds a way to get an automatic analysis with a persistent hash Stockfish, then we could guess that Stockfish would be an extremely strong opponent, even without human intervention. Score propagation heavily impacts on game quality and that's what I'm manually using, thanks to Daniel Jose code.
"But if someone finds a way to get an automatic analysis with a persistent hash Stockfish, then we could guess that Stockfish would be an extremely strong opponent,..." but that is the rub, if you have it, so does everybody else , so yes, you will be an extremely strong opponent, but so would your opponent. The best analogy is that when one engine had EGTB and the other engines did not, it produce Elo, but when all engines have EGTB, you have to have it just to stay even. You might get a lot of draws, , and you might have very few losses, but without consistent wins, the Elo will never get to 2600. This is what ICCF has become, but you still have players like Harvey that maintain a 2500 rating despite of all these challenges because they know chess better than any program in certain positions, Positions that are posted here all the time where a top engine does not see the critical drawing or winning move.
But consider this. Few people know that Daniel made a special SF version with PH feature, so I've got an advantage...

And if you talk about Harvey (and Uri as well) I think at a perfect interaction. Do you remember the position we were analyzing as a possible continuation of a Deep Blue-Kasparov game? You ran a game McBrain-AsmfishW with a draw start position, but AsmfishW heavily blundered to avoid this position and it was an easy win for McBrain:

[D]8/8/1P1k3p/3P2pP/5pP1/5K2/8/8 w - - 0 1

To know how an engine works, what static eval is, and what happens when engine returns a wrong static eval, all these things make a difference too. You can then conpare me (good player, some engine knowledge, crappy hardware, great engine) and Harvey (excellent with all above) and you get some hundreds ELO difference. :wink:
mroh wrote:...............................................................

Also, as Rudolfo said, good tools and analysis methods help a lot, ways to propagate an eval up the tree, and I am working on such a thing ^^
Interesting. So we'll have another 2600+ ELO player soon. 8-)
F.S.I. Chess Teacher

mroh
Posts: 29
Joined: Thu Nov 02, 2017 5:51 am
Location: Germany
Contact:

Re: Cerebellum, correspondence chess and me

Post by mroh » Tue Jan 30, 2018 1:32 pm

Rodolfo Leoni wrote: So we'll have another 2600+ ELO player soon. 8-)
No, of course not.
With the draw rate at corr. chess (>90%) I clearly dont play enough to make any progress... (and I even dont care that much)
And, I think my next tourney will be a FRC one, so lets start with 1800 (again) ^^

corres
Posts: 1578
Joined: Wed Nov 18, 2015 10:41 am
Location: hungary

Re: Cerebellum, correspondence chess and me

Post by corres » Tue Jan 30, 2018 1:44 pm

[quote="Jouni"]

What do You think about this message in SF forum:
"sign up with ICCF and let stockfish make ALL the moves with no intervention on human part - in all probability your rating will not go over 2200"
I don't believe this with 3300 engine!

[/quote]

If one have a rather moderate (~1700) OT (FIDE) Elo and he does nothing than he uses a modern PC of four cores around 4000 MHz clock speed to run Stockfish in Infinite mode and he also uses a common opening book, in that case he can hope to reach an ICCF(!) Elo around 2200 .
But if that one uses more PCs or a PC with much more cores AND
he uses specialized opening book(s) + analytical tools like Fritz or Aquarium, in this case - with patience and playing lot of games - he can reaches around 2400 ICCF Elo too.

Rodolfo Leoni
Posts: 544
Joined: Tue Jun 06, 2017 2:49 pm
Location: Italy

Re: Cerebellum, correspondence chess and me

Post by Rodolfo Leoni » Tue Jan 30, 2018 1:52 pm

corres wrote:
Jouni wrote:
What do You think about this message in SF forum:
"sign up with ICCF and let stockfish make ALL the moves with no intervention on human part - in all probability your rating will not go over 2200"
I don't believe this with 3300 engine!
If one have a rather moderate (~1700) OT (FIDE) Elo and he does nothing than he uses a modern PC of four cores around 4000 MHz clock speed to run Stockfish in Infinite mode and he also uses a common opening book, in that case he can hope to reach an ICCF(!) Elo around 2200 .
But if that one uses more PCs or a PC with much more cores AND
he uses specialized opening book(s) + analytical tools like Fritz or Aquarium, in this case - with patience and playing lot of games - he can reaches around 2400 ICCF Elo too.
And LSS server is harder too. ELO is deflacted because a new entry starts with ELO 800. Many ICCF IMs and SIMs have lower ELO than me....
F.S.I. Chess Teacher

MikeB
Posts: 3391
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: Cerebellum, correspondence chess and me

Post by MikeB » Tue Jan 30, 2018 3:08 pm

mroh wrote:
Jouni wrote:What do You think about this message in SF forum:

"sign up with ICCF and let stockfish make ALL the moves with no intervention on human part - in all probability your rating will not go over 2200"
I (as a ~2350 ICCF player) would guess this to be true.
Milos wrote:Time control is known, 864000+259200s. Processors if you have at least 16 at your disposal mean very little.
At that TC and with 16 cores Xeon server with 128GB of RAM and Cerebelum limited to 8 moves for example, newest SFdev alone would have ICCF rating of 2600+ and probably not lose a single game.
This is nonsense. Try it and find out!
I am not sure, if you would even pass 2000 with this approach.

Engines are still too stupid, too often to let them run alone. Want some examples?

The people who think that you can win with bigger hardware or newer engines dont have a clue how corr. chess works and more often than not, havent even played a single game.

Sure, better hardware and better engines help, but both are not the main factor. Time and patience is, especially patience. If you analyzed a position for days and you think you have it right, make a break for a day or two and look at it again with a different angle, different engine etc. You will nearly always find something new and this is a (nearly) endless process, chess is that deep, but at some day you have to make a move, or you dont have that much time, play too many games etc...

Just talk to people who are doing that for years, you will nearly always hear the same answers...


Also, as Rudolfo said, good tools and analysis methods help a lot, ways to propagate an eval up the tree, and I am working on such a thing ^^
Somebody who knows what he is talking about. Thank you Michael - hope someday we might get matched up in a game.

corres
Posts: 1578
Joined: Wed Nov 18, 2015 10:41 am
Location: hungary

Re: Cerebellum, correspondence chess and me

Post by corres » Tue Jan 30, 2018 3:56 pm

[quote="Rodolfo Leoni"]

And LSS server is harder too. ELO is deflacted because a new entry starts with ELO 800. Many ICCF IMs and SIMs have lower ELO than me....

[/quote]

Elo of LSS server is a very inconsistent thing.
Members with high LSS Elo have gotten their Elo at the time of IECG.
Much of them are inactive. Because of this it is rather difficult to manege a competition for participant with high IECG/LSS Elo. Moreover the most of competitions are Open Tournament and almost every player use the same engines so the newer LSS Elos are very equalized. This is good for beginners and bad for older ICCF players.

Rodolfo Leoni
Posts: 544
Joined: Tue Jun 06, 2017 2:49 pm
Location: Italy

Re: Cerebellum, correspondence chess and me

Post by Rodolfo Leoni » Tue Jan 30, 2018 5:35 pm

corres wrote:
Rodolfo Leoni wrote:
And LSS server is harder too. ELO is deflacted because a new entry starts with ELO 800. Many ICCF IMs and SIMs have lower ELO than me....
Elo of LSS server is a very inconsistent thing.
Members with high LSS Elo have gotten their Elo at the time of IECG.
Much of them are inactive. Because of this it is rather difficult to manege a competition for participant with high IECG/LSS Elo. Moreover the most of competitions are Open Tournament and almost every player use the same engines so the newer LSS Elos are very equalized. This is good for beginners and bad for older ICCF players.
This is a real problem. My LSS ELO is 2128 now, so I play in the range of 2000 -> 2199 and it's not a problem, there's a lot of opponents. It doesn't make any sense to register for a 2200+ ELO tournament if nobody plays. BTW, I won 3 tournaments and I gained 3 tickets to participate to those tournaments but I can see none. Maybe that range has been deleted, so players with higher ELO could participate to 2000+ ELO range tourneys if they want.

In any case, as everybody uses engines it becomes harder and harder to compete with anybody and best players get "equalized" to around 2100. FIDE GM Kovalenko is 32 ELO points stronger than me... LOL
F.S.I. Chess Teacher

Uri Blass
Posts: 8557
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: Cerebellum, correspondence chess and me

Post by Uri Blass » Tue Jan 30, 2018 6:00 pm

Milos wrote:
pijl wrote:
Jouni wrote:What do You think about this message in SF forum:

"sign up with ICCF and let stockfish make ALL the moves with no intervention on human part - in all probability your rating will not go over 2200"
This statement is probably true, although it may depend a little on the amount of time/processors you have at your disposal
Time control is known, 864000+259200s. Processors if you have at least 16 at your disposal mean very little.
At that TC and with 16 cores Xeon server with 128GB of RAM and Cerebelum limited to 8 moves for example, newest SFdev alone would have ICCF rating of 2600+ and probably not lose a single game.
I totally agree.

I got 2600+ ICCF rating and quit correspondence chess and I did not use top hardware.
I did not follow blindly the engine moves but I do not believe that my human input was worth more than running the engine on significantly better hardware with significant more time.

Using better hardware in most games could change nothing and I believe that at correspondence time control stockfish against itself is a draw in most cases even with time handicap of 10:1 but the same is for human input and in most cases my human input did not change the result.

Maybe today it is harder to get 2600+ because people use stronger hardware so you practically cannot get perfomance better than 2400 when you play against 2200 by this way but I guess that when you improve your rating and play in stronger tournaments you may get performance of 2530 when you play against 2400 and performance 2600 when you play against 2500 and continue in this way to get progress.

The numbers that I wrote are not based on calculation and only demonstrate the idea.

corres
Posts: 1578
Joined: Wed Nov 18, 2015 10:41 am
Location: hungary

Re: Cerebellum, correspondence chess and me

Post by corres » Tue Jan 30, 2018 10:35 pm

[quote="Rodolfo Leoni"]
[quote="corres"]

Elo of LSS server is a very inconsistent thing.
Members with high LSS Elo have gotten their Elo at the time of IECG.
Much of them are inactive. Because of this it is rather difficult to manege a competition for participant with high IECG/LSS Elo. Moreover the most of competitions are Open Tournament and almost every player use the same engines so the newer LSS Elos are very equalized. This is good for beginners and bad for older ICCF players.
[/quote]

This is a real problem. My LSS ELO is 2128 now, so I play in the range of 2000 -> 2199 and it's not a problem, there's a lot of opponents. It doesn't make any sense to register for a 2200+ ELO tournament if nobody plays. BTW, I won 3 tournaments and I gained 3 tickets to participate to those tournaments but I can see none. Maybe that range has been deleted, so players with higher ELO could participate to 2000+ ELO range tourneys if they want.
In any case, as everybody uses engines it becomes harder and harder to compete with anybody and best players get "equalized" to around 2100. FIDE GM Kovalenko is 32 ELO points stronger than me... LOL
[/quote]

I made some proposal to Mr.Paetzold to create more higher rated player with the modification of rating categories and the tournament system but
without any effect.
Please read LSS Forum/System Announcement: "Only less than 8 percent of LSS players have 2300 or above Elo. Why?"
If you want to play many chess games in any circumstances but free of charge - LSS is your place. But if you want to see the result of your effort - ICCF is the better site.

Thomas Zipproth
Posts: 56
Joined: Sat Jan 01, 2011 4:33 pm

Re: Cerebellum, correspondence chess and me

Post by Thomas Zipproth » Tue Jan 30, 2018 10:59 pm

Ozymandias wrote:
Rodolfo Leoni wrote:
Rodolfo Leoni wrote:It happens that I'm playing correspondence chess and, with the help of the very special Daniel's code, I'm still undefeated in master tournaments after almost 100 completed games.

Some time ago, I sent a novelty to Stefan Pohl. It was a move uncovered by Cerebellum, much stronger than Cerebellum best move and worth of analysis. Within 24 hours the move was included and Stefan provided me a link to a Cerebellum prerelease. I want to acknowledge Stefan because he's been very helpful and friendly. Cerebellum analyses were kind of a disaster, tough, but I'm not the author and I thought it was better to avoid insisting with my analyses.

Few days ago, I mailed (directly to Thomas Zipproth) another novelty 70 cps stronger than Cerebellum best move. Seventy centipawns/00 :!: No answer. No thanks, no "please avoid sending me moves again", absolute silence.

I found a third very strong novelty yesterday, not included in Cerebellum. But why should I share it? I feel like an idiot, and I probably am!
I just received an e-mail from Thomas Zipproth. He thanks me and both two moves were included in Cerebellum. So... my apologies for my bad post. I sometimes am too much emotive. :oops:
Was his reply timestamped before or after your first post here?
I just read this thread and this whole thing is ridiculous.
Stefan forwarded his first request to me and I integrated the first variation between 24 hours and created a prerelease for him, if this is not fast enough I can't help.

I answered the second request after I integrated his suggestion and released a new Cerebellum containing it.
Chess is not my job, it is a hobby, and Cerebellum is free. Due to my job as a software developper I do not always have time to answer every mail at once.

I'm working on the professional Version of Cerebellum, but the general support of Cerebellum light is sometimes very time consuming .
Maybe I should stop releasing Cerebellum till the professional Version is ready where everyone can expand the book himself which is the correct way to use it for correspondence chess.
This would also avoid such discussions. :)

Post Reply