Stockfish defeats Leela C0 and prooves that AlphaZero was a myth

65 games of the TCEC superfinal and Leela is leading +2. Where are those SF whiners now? Now they can´t complain about unfair hardware, time control or the little SF not having his opening book.

I don't remember any "SF whiners" addressing any LeelaC0 vs. Stockfish hardware configuration, you must be imagining things. The only hardware comparisons have been between the AlphaZero vs. Stockfish matches in 2017 and 2018. In those matches the AlphaZero hardware configuration had an approximate 100X the processing capability as the Stockfish hardware configuration. And none of the "whining" had to do with the superiority or non-superiority of neural network vs. "classical" chess engine implementations, just attempting to reach a conclusion in a match when one of the engines had a 100X processing capability advantage over the other.

In Season 13 the TCEC organizers did a much better job in trying to equalize the hardware capabilities of the neural network and classical chess engine approaches, and it was a hardware problem that prevented LeelaC0 from advancing further. It seems that they have done a similar good job in Season 14.

After 68 games of the Superfinal the score stands LeelaC0 9 wins, Stockfish 8 wins, with 51 draws. Hardly what I could call "crushing". And with 32 games to go the outcome could still go either way.  Which is the way it should be.

70 games now and Leela keeps leading by just one point

final score: SF 50 1/2 Leela 49 1/2 that was a very close match. Leela was leading but after game 63 couldn´t win anymore.  The only black victory was also one of the most spectacular games (remember that the openings were fixed, every contender played them once with white and once with black), an ultra sharp KI line with queen sacrifice included, where SF strenght of millions moves calculated per second beat Leela "weak" thousands per second.

If anything we should be supporting neural nets like Alpha Zero and LC0 over brute force approaches like stockfish as A0 and LC0 have more in common with human thinking processes.

Moreover, neural nets are actually promising rather than brute force computation. Brute force computation cannot work in other areas(like driving a car or predicting the weather) whereas you can do a lot of stuff with intelligent methods such as neural nets.Basically the way alpha zero figures out the next move is much like how a human being would complete a fill in the blank sentence i.e. based on prior experience and training. The neural net is exposed to several games and 'trained' over this 'data set'. It will then use this training to fill in the next move in the sequence of moves just played. This is much how a human thinks.

Yes. I believe that brute force was what gave SF the win in extremely tactical games like the one I showed

Brute force doesn't scale very well at all and is a dead end. On the other hand, if you use probablistic cues and intelligent searching as are inherent in neural nets then you will be able to find the solutions to hard computational problems much much much quicker. 

chess is a great example and thus a great way to prove the superiority of neural networks over blind computation. basically a human being doesn't consider every move when they think over the board. but a blind brute force algorithm does. neural networks don't consider every move either. they are trained over a data set which gives them a probabalistic estimate of which move should be played next or what kind of position should be achieved in the next move. they then move on the basis of these cues and select a move based on this training. this is just how a human being progresses in their thought process.

i won't be surprised to see neural nets solving physics problems in the near future.

In the first competition stokfish wasn't given equal hardware I believe. In the second a year later it was, but it wasn't the most updated version of stokfish and the most up to date version of stokfish has a bigger rating gap compared to it's old version then Alpha Zero did I believe. Meaning Alpha Zero would be weaker then the best version of stokfish.

so you're saying that stockfish is stronger than alpha zero on equal hardware? that's probably demonstrably false.

Well if you look at the results and that do the maths I believe the most recent version of stokfish agaisnt the most recent version of Alpha Zero would end up with stokfish probably winning. I am just basing my self on the maths people did after the  Alpha Zero game. I would love it if you were to do the maths because that is the only way to demonstrate it. It's no my opinion it's based  on somebody's analysis of the results.

This is 100% false. 

Alpha Zero had many millions of times the computing power of Stockfish in the games they played. 

Do not make things up idiot. 

Stupid and wrong. 

Leela Chess Zero still runs on a distributed computer network. 

It's the same bait and switch trick as Alpha Zero, running on better hardware and pretending it's a better engine. It's the exact same thing, it can't run on a normal computer.

Here is the analysis. I do believe that Alpha Zero has played against stofish 9 but not stokfish 10 I don't know what the result was in the stokfish 9 game.

" the results are 155 wins, 839 draws and 6 defeats for Alphazero. This means a score of 57.5% against 42.5%. More or less 70 elo difference, stronger than Stockfish 8."

Based on this website http://ccrl.chessdom.com/ccrl/4040/rating_list_all.html the rating game in between stokfish 10 and 8 is 85 points, meaning stokfish 10 is a better engine the Alpha Zero was when the game was played.

I will have no problem with the DeepMind team, once A0 plays in a real match with a 3rd party organizer.  I've got nothing against A0.  It's the Google team's shenanigans that are the problem.  None of A0's results are official because Google is running the matches behind closed doors.  That's not really A0'sa fault ...Google is just trying to get the free press for "winning" against the top engine without actually taking any risk.  Smart marketing decision, bad chess decision.

100% agree.

Yes, it's true...Star Trek script writers invented the concept of a neural network.../sarcasm.

Spa-ghetto is definitely a troll, though, always has spouted nonsense with no facts.

I have always been a reliable and good poster on this site. I am always right when I say something like that. 

I swear, this site is genuinely one of the DUMBEST places I ever go on the internet. Some people are pretty intelligent of course but lots are like barely functioning. 

So much for the stereotype of chess players being intellectuals. 

I think SF is better until 3rd party match between SF and A0 occurs. Who knows on which computers and what the settings were when SF played against an A0 in google office?

I believe it is actually spelled “proves” not “prooves”. I am referencing the title.