Page 1 of 1

Split off from 'Last good Stockfish'

Posted: Mon Oct 07, 2019 8:57 pm
by Eelco de Groot
Ovyron wrote:
Mon Oct 07, 2019 1:19 am
Ovyron wrote:
Fri Sep 27, 2019 7:35 pm
Anyway, I'm looking forward to a release of Bluefish in an official fork to check it out :)
So, did this happen and I missed it?
I thought that you were going to do it, take up programming I mean. :) I believe that you already managed to get Large Pages working in Stockfish Ovyron, a good while back in Rybka forum so I don't quite understand why you find it such a hassle to compile?

I don't think there is much added value in putting Bluefish on GitHub as a fork. Would require continuous updates whenever Stockfish is updated or there is not much point to it. And it can't always be done automatically, and I would have to make a totally new fork as my official one is on Windows XP now. Can't update that one anymore. Most pople are only interested in something that is compiled. Sure I can make some updates but I don't think daily new versions are needed for this one. Anyway I am not planning on that. For one thing it would be totally confusing for most and the differences between versions just with new Stockfish commits would be minimal I think.

Re: Split off from 'Last good Stockfish'

Posted: Tue Oct 08, 2019 3:09 am
by MikeB
Eelco de Groot wrote:
Mon Oct 07, 2019 8:57 pm
Ovyron wrote:
Mon Oct 07, 2019 1:19 am
Ovyron wrote:
Fri Sep 27, 2019 7:35 pm
Anyway, I'm looking forward to a release of Bluefish in an official fork to check it out :)
So, did this happen and I missed it?
I thought that you were going to do it, take up programming I mean. :) I believe that you already managed to get Large Pages working in Stockfish Ovyron, a good while back in Rybka forum so I don't quite understand why you find it such a hassle to compile?

I don't think there is much added value in putting Bluefish on GitHub as a fork. Would require continuous updates whenever Stockfish is updated or there is not much point to it. And it can't always be done automatically, and I would have to make a totally new fork as my official one is on Windows XP now. Can't update that one anymore. Most pople are only interested in something that is compiled. Sure I can make some updates but I don't think daily new versions are needed for this one. Anyway I am not planning on that. For one thing it would be totally confusing for most and the differences between versions just with new Stockfish commits would be minimal I think.

I can do it for a while, not that it will be forever. One comment, folks will have to compile their own .

Re: Split off from 'Last good Stockfish'

Posted: Tue Oct 08, 2019 4:55 am
by carldaman
Eelco de Groot wrote:
Mon Oct 07, 2019 8:57 pm
Ovyron wrote:
Mon Oct 07, 2019 1:19 am
Ovyron wrote:
Fri Sep 27, 2019 7:35 pm
Anyway, I'm looking forward to a release of Bluefish in an official fork to check it out :)
So, did this happen and I missed it?
I thought that you were going to do it, take up programming I mean. :) I believe that you already managed to get Large Pages working in Stockfish Ovyron, a good while back in Rybka forum so I don't quite understand why you find it such a hassle to compile?

I don't think there is much added value in putting Bluefish on GitHub as a fork. Would require continuous updates whenever Stockfish is updated or there is not much point to it. And it can't always be done automatically, and I would have to make a totally new fork as my official one is on Windows XP now. Can't update that one anymore. Most pople are only interested in something that is compiled. Sure I can make some updates but I don't think daily new versions are needed for this one. Anyway I am not planning on that. For one thing it would be totally confusing for most and the differences between versions just with new Stockfish commits would be minimal I think.
Continuous or frequent updates would be overkill. The idea is not to extract every bit of Elo from all the SF updates, but to have something (like Bluefish) that's capable of creative and adventurous chess of very high quality (meaning very high Elo, as well, despite the changes).

The 'problem' with the SF team is that while they're doing a wonderful job of improving the engine, Elo gains dominate the thinking behind all the improvements, creating a form of 'tunnel vision'. A stylistic enhancement that loses some Elo would never make it into the real SF. That is a pity, unless someone steps up and decides to put style before strength as a priority. Eelco's Bluefish belongs to this latter category, breaking away from the dark old 'tunnel' into a new world of 'bright daylight' and exciting possibilities, if you will. :)

Re: Split off from 'Last good Stockfish'

Posted: Tue Oct 08, 2019 8:20 am
by Ovyron
Eelco de Groot wrote:
Mon Oct 07, 2019 8:57 pm
I thought that you were going to do it, take up programming I mean. :) I believe that you already managed to get Large Pages working in Stockfish Ovyron, a good while back in Rybka forum so I don't quite understand why you find it such a hassle to compile?
I guess that, as embarrassing as it is, it's time to confess... it's because of my laziness :oops: People have actually made compiles for me, because they thought I couldn't make them myself, and I never told them I could compile them myself, but I was lazy.

At least, I promise to be less lazy about it. My experience about code tweaking and doing my own compiles is that, whatever I think about doing, there's someone in the world that is doing it as well, and it's doing it better than me. So at some point I was doing my own optimized compiles, then Jim Ablett appeared and made compiles that were running faster than my own compiles on my system.

At some point I just got fed up with using Stockfish TCEC by Jeremy Bernstein because it had learning, because it was a very old engine and I wanted a more modern version of it, so I started working into porting that learning to Stockfish 8. A few days later a programmer (that wished to remain anonymous) told me he had came up with a learning algorithm for Stockfish, and what he had was better than anything I could have done myself (because it had the brilliant concept of storing white's learning and black's learning separately.)

I started working on an opening book that would bridge the hardware gap between me and my opponents on InfinityChess, and a few days later someone sent me their private book that was better than what I could have achieved after months or years.

What about something that would allow people to access my database of analyzed positions online? noobpwnftw already has that, with a system much better than what I could do, and his database is 1200 times larger than what I have.

I was thinking about examining Stockfish 5 code and check what was allowing it to have "the sparkle", to make code changes in official Stockfish to allow it to play in a much better style, but you could have done it already better than what I could have done, with Bluefish.

And that's why I've become lazy, because anything I could be doing programming-wise someone else is potentially doing it already, potentially doing it better, potentially getting results faster, or perhaps instead of working on it it'd be easier to request someone like Ferdy to do it (Aquiri was a success.)

The irony is that I studied programming, that's supposedly my career, and I don't think it has had any use.