A tale of two bishops | DeepMind's AlphaZero Game Changer 8 | AlphaZero vs Stockfish 8

[@PowerPlayChess]

If you like my videos do *subscribe* bit.ly/powerplaysubscription and do checkout the supporting options through *Patreon* : bit.ly/patreondanielking or through *PayPal* (links in the description)

[@bornon221269]

Daniel, your video is the best to start the day

[@lemoncake7565]

I'm more excited watching these games than i was at WCC

[@zaksmith1035]

These two last videos have been my ABSOLUTE favorites in your series. Thank you Daniel King for your continued wonderful teaching, and awesome hair. I've been following AlphaZero from the beginning, and so the humanity of its play always struck me, however the way you explained it revealed wondrous new secrets to me, not only of the way AlphaZero plays, but just good chess itself. I find that AlphaZero is always coordinating along lines, building up its pieces to cover more squares, which leads to positions that are often like endgame problems, combined with a Karpovian positional grip. Its control of every important square and desire for activity above all else is a lesson to all of us in our games. I feel like AlphaZero is a Calrsenarpov, married to a Kasparovian 2 point ovian, whose brainchild then took Omega 3 Fish Oil from birth. I love how you contrasted it with Stockfish. That revealed the depth of the difference between the NN and Alpha-Beta style of play with such clarity. I thought I knew a bunch about AlphaZero, and could describe it better than most, however you went to a level I'd never imagined. So, again, thank you. I also learned a TON about Stockfish's playing style, which I'd never studied much about. I always considered the computers to be boring. Accurate, but boring. Unable to teach. But I guess they have a lot in common with Korchnoi, so perhaps their style could be described with some Legendary Chess Monikers as well. That is beyond me though, as I feel I just met Stockfish for the first time. And I wasn't bored at all, by the way. Cheers.

[@Brenden97]

The thumbnail is excellent. The quality of these videos is excellent aswell.

[@tasdourian]

Another outstanding commentary. What a great selection for a pair of games to analyze. I learn a lot about chess style from watching these videos.

[@dustinbachstein]

I love this series, thank you!

[@leedsfanalways]

Great video and analysis as always Daniel, it's always a pleasure to watch. Keep them coming!

[@howardpearse]

What an amazing game and instructive video from Mr King! I'm fascinated about these AI games- and these two games really pose some quite interesting and even philosophical questions. One thing I noticed about AlphaZero's strategy of initiative is that it's always attacking. While this makes for a dynamic and entertaining game it should be considered that a big part of chess is playing for the draw. I suppose an interesting question might be to ask whether it is possible to program AlphaZero to do just this- the subtlety being that the decision to play for a draw vs a win is not only one that chess players make when playing tournaments (such as in the recent WCC) but one that requires an great sense of intelligence. I wonder if any of the games played demonstrate AlphaZero's capacity to judge the value of a strong draw over a probable win.. fascinating stuff! :0)

[@Guitare_picking]

Thank you so much for this Daniel !

[@martinet1985]

I'm new to the channel and I really enjoy your commentary style!

[@nilsp9426]

Maybe with colours reversed, A0 was panicking a bit, because the position is just better for white?

[@KolinKrake]

This game reminds me, in particular the Bc2 maneuver, of Gusev's Immortal. Black's pieces were just paralyzed by the bishops and the open files. Very interesting!

[@atrichatterjee2299]

Great content

[@yorick021]

Somehow A0 seems to have a more human style than SF. Very interesting to see how different a computer can approach a position based on experience versus raw calculation power. What is especially notable for me from all these games is how SF virtually never has a safe king in the middlegame while king safety is number 1 priority for chess engines usually. Somehow A0 seems to find a way to upset SF every game. Is A0 still learning every game and can it thus adjust to its opponent or has it only learned from playing itself?

[@sumitstir]

Mr King, it would great if you could cover the reception of AlphaZero among the chess elites, specifically the old guard like Karpov, Kasparov, Anand. It would be interesting to know their thoughts, thank you !

[@PK-hs7up]

Thanks for the video, but at about 12:40 there's no sound anymore :-(

[@PowerPlayChess]

Sound works for me in the whole video, you might want to check your audio settings.

[@PK-hs7up]

Strange, I double-checked it but now it works also for me... The last time, sound stopped suddenly. Thanks for your reply. Now I'll enjoy your analysis in full length ;-)))

[@melrakan]

How should Stockfish have responded to the poisoned pawn AlphaZero offered it?

[@Nick-nb8kf]

Hi Mr. King, unrelated but before I start this. If wanted to say you have to check out Nakamura’s video of doing chess puzzle rushes. It’s crazy!

[@rbr4784]

A0 reminds me of a Morphy-esk style of playing; peice development and material sacrifices for an attack on the king.

[@proflaxis6968]

Its like watching Brazil v Italy in football. Alpha with its attack at all costs philosophy and sf with its solid stubborn defence. Bananarama were right. It ain't what you do its the way that you do it. That's what gets results. Ah Alpha, a breath of fresh air. Making chess great again.

[@BboyDschafar]

Turns out, Danny is a fan of Korean Horror movies.

[@FlaminalLow]

...”to human eyes”, but most importantly, to AlphaZero’s eyes (!), grabbing material and holding on to it looks ugly and inappropriate, nevertheless Stockfish won! So my question is, at what point in the evaluation did AlphaZero go wrong in the game it lost? Also, it now bugs me no to know which of the b5 or b6 approaches is stronger; they both lost but one move has to be better than the other, right? :)

[@Joshman345]

I knew that alphazero would destory stockfish 8

[@burt591]

Tal once said "There are two types of sacrifices: correct ones, and mine." but I'm starting to think that his sacrifices were actually correct

[@zastava9412]

At 1:55 real sf doesnt take Nxg5. It plays Nef6. Which is 50CP difference.

[@fixpontt]

i rechecked this one with newest sf10, after 31 minutes on my i7 machine the eval fluctuates every 2 minutes between: ne4-f6, ne4xg5, when i turned off the engine after 31 minutes it was: 1. ne4xg5 +0.50 2. ne4-f6 +0.62 the difference is negligable and i can imagine on a 44 core machine (with different hashtable size) the eval was slightly different, but very very important to run the engine in multi PV 4-16 search not a single PV, because it is misleading, in a single PV mode it drops candidates very fast so even the very best and newest version takes it, of course ne4-f6 was the first candidate after 10 minutes, but not after 31 i.imgur.com/vkRkblA.jpg

[@walkerstreet4323]

Alphazero is playing like topalov Danny, topalov will come again with alphazero's concepts maybe.

[@zastava9412]

In MANY cases this "stockfish" is choosing moves that are not real SF first choices. For instance 9:13 Nxg4 - there is no such a move in SF movelist. Over 100 CP drop after that 1 move.

[@PuzzleQodec]

Good to know that Stockfish is still the strongest engine in the world. Phew.

[@fixpontt]

there is real and not real SF, this is the 2018 jan version, the best was available at that time, SF9 was released in feb 2018

[@zastava9412]

@@fixpontt im just using sf9 android on my phone. Not fake 44 core server with 3 hours of time. Sf is relesing new versions like 5 times per month. So its not true that sf9 was the last available version

[@fixpontt]

it is easy the reproduce the circumstances that DM used during the test yet nobody has done it, i read a lot of criticism on the net about the match, none of which can be taken seriously including yours, nobody has done the effort to really prove their point, running newer version of SF on a phone means absolutely nothing

[@SoldierOfTwilight]

@@zastava9412 Stockfish fanboy triggered