TalkChess.com

mroh wrote:Yeah, call me naive, but thats how I have always handled correspondence chess since Ive been doing it with postcards as a kid.

Did you? Did you send all the analysis you were making to your opponent to help them out avoid making mistakes so you both achieved the perfect game?

I have never seen anyone do such a thing, you want to play perfect chess for your side, and this implies reaching positions where it's harder and harder for your opponent to avoid playing the mistake.

Sometimes a position with a 0.00 score is no different from a 0.99 score from the theoretical perspective of chess, both moves lead to a draw with perfect play. Suppose you manage to find a draw with the 0.99 move, do you go "oh well, nuts! Looks like I have nothing here, I'll send a draw offer...", or, since you spent days analyzing the positions and know this draw is hard to find, you play the 0.99 move and hope the opponent misses the right continuation because they're not being as thorough as you? Because, most games I've won have been like this, the opponent missed something that allowed me to win, and this isn't different from OTB chess, except it uses a completely different set of skills.

mroh wrote:Sure, its always nice to win, but if I would like to have more of the "fighting" (and winning) aspect of chess, I would play OTB, which is pure fighting and destroying your opponent, as Bobby F. and Garry K. once said.

Yeah, certainly Correspondence Chess is like, excuse the term, "sex without the orgasms", since I have been more frutrated in some of my won games because my opponents clearly had no hope and still drag the game for months... sometimes for years, and sometimes they never return again and I wonder if I should have allowed infinite time controls in the first place.

There's no euphoria in corr chess, because the result is already known weeks in advance, and sometimes your opponent resigns in a game you have not cared about months ago, so wins left a lot to be desired, and the closest thing to satisfaction is finding a continuation your opponent misses that gives you a great edge, but you never know if it's decisive, so nothing like the instant gratification of chess games.

And that's why I quit computer chess.

But what if you really suck at OTB chess but you're good at Corr Chess? I had to come back to it because I found nothing else to do and because apparently I'm still improving my analysis methods, so it's something to do and I have been sad about talented corr chess players wasting their talents by quitting because they'd rather do something else with their lives.

mroh wrote:But maybe, this idea is simply an illusion of an old, naive and stupid man ^^

Chess is an unsolvable problem, what I have seen in the last decade of computer chess is that it's just the openings that change, some appear playable that looked dubious before, others that were mainlines for years are basically solved to a draw with all the live and interesting variations sucked out of them... I still remember when the Sicilian's Poisoned Pawn was interesting!

But we're not closer to finding real truth than we were back when Rybka 1.0 beta was released... And any analysis that you produce today will be obsolete 5 years from now, with Stockfish 11 showing many huge advantage positions being actually 0.00, and finding holes in lines that were supposedly sharp but actually can be refuted, with the side with the supposed advantage being reversed from what it was.

You should analyze chess positions because you enjoy doing so, not because you want so solve a problem or advancing the art. And only if you have nothing better to do. I once played this guy that loved corr chess greatly and was very strong, but had to take a break and tone him down because he was losing his family

mroh wrote:Why should I implement something to help different users to hide their analysis where there are no users and no analysis?

I might have sent something back when you invited me if Private Trees were already implemented, the reason I haven't sent a single position into Free Aquarium is that since it's public, any of my +40 opponents could potentially look into my positions I'm playing with them, and that's a no-no.

So currently the main difference between fA and other alternatives is that those allow the user to keep things private by default and only share them if they wish. Heck, none that I remember have a way to share your analysis with people from inside the GUI

What about this?:

Moves added to Private Trees become public automatically after 1 year of being inserted into fAquarium.

This way more people would be willing to share, hopefully, and then right now you have the main public trees looking das and all but one year later, if people aren't afraid to have their analysis published 1 year from now, thos private positions will start popularing the tree, and you'll have the results you want with a year delay, but better late than never.

mroh wrote: Without going to much in detail (in public), Im currently a bit frustrated, even thinking about continuing this project just for me, as it once was...

But, well maybe I simply need a break and/or have to many expectations ^^

I hope you don't get discouraged. Your idea is really great and you only need to "tune" it a bit. (Hopefully, nobody'll try to give me a lesson on what "tuning" is!)

Ulysses hints are perfect. You'll have lots of clients connecting there if you just separate the "private" and "public" sections, as that would probably solve all the issues about the correct use of the GUI.

Yes! Your are absolutely right with everything you say, especially I like this:

Ovyron wrote: You should analyze chess positions because you enjoy doing so, not because you want so solve a problem or advancing the art.

because its so true, not only for chess...

Ovyron wrote: What about this?:

Moves added to Private Trees become public automatically after 1 year of being inserted into fAquarium.

This way more people would be willing to share, hopefully, and then right now you have the main public trees looking das and all but one year later, if people aren't afraid to have their analysis published 1 year from now, thos private positions will start popularing the tree, and you'll have the results you want with a year delay, but better late than never.

Hmm, a good idea!

To also solve the problem of "please add support for engine xy (SF clone nr. 339...)", what about this:

I give every user the possibility to use X engines of their choice with options of their choice and the resulting X trees of these are private.

These trees are most commonly private forever, because its unlikely that any of them are interchangeable, because of different versions, options, minimum depth etc...

That way one could also try an idea you had in another thread with different Contempt settings (for the same engine) which results in different trees and so on...

In the long run this might show, that my illusion of "public trees/analysis" would become useless, but I don't know, lets find out.

What I currently don't know is, which value that X should have, lets begin with 2 or so and see if it works...

mroh wrote:I give every user the possibility to use X engines of their choice with options of their choice and the resulting X trees of these are private.

What if you let X be anything? I currently can install any engine of my choice into the Shredder Classic GUI, and add all of its analysis into a Bookup tree. You'd probably want to feature something like that.

So here's an idea:

Transform wsc into an adapter, that sits between the engine and a GUI (it doesn't matter what GUI, wsc would work like WB2UCI or Polyglot, and manage to be installed into any GUI), then it listens to what the GUI says, and reports it to the engine, and listens to what the engine reports, and sends it back to the GUI and into free Aquarium, to be uploaded into some tree online.

This way you don't need to add special support for any engine because as long as it speaks UCI wsc will understand it and will be able to update a tree for any engine. Winboard engines will be able to be used by using some stacking adapter so you don't need to worry about that (I used to use stacked adapters for the InBetween adapter or for Aquarium's Book adapter to allow engines to use ctg books, so it has worked before.)

Then the Tree online just needs to be aware of the engine's name reported by wsc, and think "oh, this engine hasn't been used by the user yet, they probably want me to create a tree for them", so an user can have as many trees for as many engines as they wish, just like with commercial Aquarium.

Ovyron wrote: So here's an idea:

Transform wsc into an adapter, that sits between the engine and a GUI (it doesn't matter what GUI, wsc would work like WB2UCI or Polyglot, and manage to be installed into any GUI), then it listens to what the GUI says, and reports it to the engine, and listens to what the engine reports, and sends it back to the GUI and into free Aquarium, to be uploaded into some tree online.

Yeah, I already had this idea (didn't we talk about it?)

Anyway, with the private trees it would indeed even make more sense, esp. with your idea of some automatic logic for creating them... lots of details and work here...

but let me first implement the private trees than we move on ^^

mroh wrote:Yeah, I already had this idea (didn't we talk about it?)

Hmm, probably not? Last time I talked about adapters it was when I was telling Rodolfo Leoni that we don't need to implement learning to engines or to have a GUI that does it for them because an adapter could handle such a job.

mroh wrote:but let me first implement the private trees than we move on ^^

Looking forward to it! Glad you're not quitting free Aquarium!

mroh wrote: but let me first implement the private trees than we move on ^^

mission accomplished!

Now every (registered) user has the possibility to create (currently up to 5) "private" trees and engines.

These trees are not visible for anyone else and also create new engine endpoints (see doc)

Because people seem to have problems typing anything on a cmdline,
I made a small GUI tool (currently, for windows only) that asks for the parameters in an easy to use (powershell) dialog and starts the wsc/websocket tool.

This tool is available here, source code here

Thanks Michael! I'll be checking it out shortly

Hey, finally got time to try this!

I tried installing an engine by running WSC-GUI, but got this error:

Ovyron wrote: I tried installing an engine by running WSC-GUI, but got this error:

if that is meant to be an user created engine:

An engine name must be alphanum [a-zA-Z0-9] chars, so no " ", "_" or ":" etc. Also you cant name it the same as an "System Engine".
I check all that if you try to create an user engine in fA. Maybe its better to create the engine first, before you try to connect against it...

If you try to connect against the "system" sf8, just remove the " " so its "Stockfish8". With the "system" engines I check the uci response to see if its really a sf8, the tree is public and you have not many options to configure...



If I am here:
I added some new features:
- you can connect your fA account to the one from lichess. fA polls lichess for running corresponding games and displays them.
- as an experiment, I added a webassembly version of sf9 which runs in your browser. Currently, it doesnt write anything to a tree...
- reworked the login mechanism which means no more guest login needed...