Chessbase's Albert Silver on LC0 & Fat Fritz

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.
Post Reply
supersharp77
Posts: 753
Joined: Sat Jul 05, 2014 5:54 am
Location: Southwest USA

Chessbase's Albert Silver on LC0 & Fat Fritz

Post by supersharp77 » Mon Aug 26, 2019 6:40 pm

https://en.chessbase.com/post/fat-fritz-and-the-kid

I just read a interesting and profound statement by Chessbase's Albert Silver originating from a discussion on the double edged 'Kings Indian Defence' (A Fischer and Kasparov Favorite, currently under a cloud) many modern GM's are veering into other indian lines at the moment...

"While it enjoyed its spot in the sun with its last great proponent in Garry Kasparov, the latter’s inability to overcome his rival Vladimir Kramnik with it, leading him to actually stop playing it, was a hammer blow. Almost as if dictated by fate, the rise of the chess engines was around this time, with Fritz at the top, later followed by Rybka, Houdini and more, and they all seemed to universally condemn the opening with extremely positive evaluations for White, which were only reversed by a mistake or blunder. The conclusion one might draw from this is that while the KID might offer plenty of practical chances, it is fundamentally suspect." :wink:

........"Or is it?".........

"Fat Fritz offers a very different perspective on this, and it is not just because it is a neural network. It is also because of how it was trained and developed. When I began training it, the idea was not only to offer it the bulk of human genius from Mega Database, but to see it learn about all the openings that exist. This does not mean it will necessary like them, just that if it doesn’t, its opinion will have come from the practical study of millions of positions. This highlights one of the major differences compared to the ‘zero’ approach used by AlphaZero. The ‘zero’ approach accurately claims to remove all human bias, since in effect it is learning solely from games played against itself, but its proponents fail to highlight that instead this introduces significant biases of its own.

Reinforcement Learning Paper (Alpha Zero)
https://kstatic.googleusercontent.com/f ... 45072dc7bd

"The published article on AlphaZero, and even the excellent book on it by GM Matthew Sadler and Natasha Regan, clearly show that while AlphaZero may be incredibly strong, it is also a super specialist in very specific openings. When allowed to steer the games towards openings of its choice, it did unusually well, but if greater variety were enforced the results were much less impressive. Was this because some of these other openings were inherently flawed or because its skills were not as universally strong in all of them? I suspect a bit of both."

This fact was noticed from the beginning of the AlphaZero-LC0 Era...."Playing a extremely limited number of openings
(Ex.. D4 or C4 with white or white side of queens indian defense) but playing them extremely well'...Is this some "fatal flaw?' Unknown..."Time will tell"...... :) :wink:

carldaman
Posts: 1720
Joined: Sat Jun 02, 2012 12:13 am

Re: Chessbase's Albert Silver on LC0 & Fat Fritz

Post by carldaman » Mon Aug 26, 2019 10:32 pm

No real surprise here, but it's a welcome development.

I've argued in favor the viability of the Classical KID lines (from Black's perspective) in this forum, as did Lyudmil Tsvetkov to an even larger extent, for quite some time, and way before AlphaZero and Leela, or "Fat Fritz", came around. Most people tended to believe the top engines evaluations, which wrongly claimed a large edge for White, of at least 0.70.

jdart
Posts: 3824
Joined: Fri Mar 10, 2006 4:23 am
Location: http://www.arasanchess.org

Re: Chessbase's Albert Silver on LC0 & Fat Fritz

Post by jdart » Tue Aug 27, 2019 12:55 am

I believe that most openings are playable by very, very strong engines. You can see this in venues like TCEC where there is a fixed set of openings (by no means always good ones) and the draw rate is quite high. With enough search strength and processing power, engines can at least find a way to equalize. There are still some lines it would be best to avoid, but only the most dubious (Grob for example).

--Jon

Dokterchen
Posts: 104
Joined: Wed Aug 15, 2007 10:18 am
Location: Munich

Re: Chessbase's Albert Silver on LC0 & Fat Fritz

Post by Dokterchen » Tue Aug 27, 2019 5:48 am



Nxe4 is nothing special for NN programs:

Code: Select all

ScorpioNN 3.0-n_maddex:
00:06	        119k	20k	+0,10	Nf6xe4

AllieStein 0.4-n10.1:
00:06	        269k	41k	+0,28	Nf6xe4

Lc0 0.21.1-nT40.T8.610:
00:11	        104k	9k	+0,32	Nf6xe4 
Torsten

Post Reply