An UNBELIEVABLE MEGA BOOM! || Mighty Stockfish vs Highly Evolved Leela || TCEC Season 18 Rd 60

♚COURSES kingscrusher.tv/chesscourses
📚 kingscrusher.tv/chessopenings
📚 kingscrusher.tv/chesstactics || kingscrusher.tv/londonsystem
FIDE CM Kingscrusher goes over An UNBELIEVABLE MEGA BOOM! || Mighty Stockfish vs Highly Evolved Leela || TCEC Season 18 Rd 60
♚ Play turn style chess at bit.ly/chessworld
♚ kingscrusher.tv - register and I will invite you for a game soon :)
FIDE CM Kingscrusher goes over amazing games of Chess every day, with a focus recently on chess champions such as Magnus Carlsen or even games of Neural Networks which are opening up new concepts for how chess could be played more effectively.
The Game qualities that kingscrusher looks for are generally amazing games with some awesome or astonishing features to them. Many brilliant games are being played every year in Chess and this channel helps to find and explain them in a clear way. There are classic games, crushing and dynamic games. There are exceptionally elegant games. Or games which are excellent in other respects which make them exciting to check out. There are also flashy, important, impressive games. Sometimes games can also be exceptionally instructive and interesting at the same time.
Info about Leela Zero:
en.wikipedia.org/wiki/Leela_Zero
...
Leela Chess Zero (lc0) is a free, open-source, and neural network-based chess engine and distributed computing project.
Leela Zero's algorithm is based on DeepMind's 2017 paper about AlphaGo Zero.[3][6] Unlike the original Leela, which has a lot of human knowledge and heuristics programmed into it, Leela Zero only knows the basic rules and nothing more.[7]
Leela Zero is trained by a distributed effort, which is coordinated at the Leela Zero website. Members of the community provide computing resources by running the client, which generates self-play games and submits them to the server. The self-play games are used to train newer networks. Generally, over 500 clients have connected to the server to contribute resources.[7] The community has provided high quality code contributions as well.[7]
Leela Zero finished third at the BerryGenomics Cup World AI Go Tournament in Fuzhou, Fujian, China on 28 April 2018.[8]
Info about Alphazero:
en.wikipedia.org/wiki/AlphaZero
AlphaZero is a computer program developed by the Alphabet-owned AI research company DeepMind, which uses an approach similar to AlphaGo Zero's to master not just Go, but also chess and shogi. On December 5, 2017 the DeepMind team released a preprint introducing AlphaZero, which, within 24 hours, achieved a superhuman level of play in these three games by defeating world-champion programs, Stockfish, elmo, and the 3-day version of AlphaGo Zero, in each case making use of custom tensor processing units (TPUs) that the Google programs were optimized to make use of.[1] AlphaZero was trained solely via "self-play" using 5,000 first-generation TPUs to generate the games and 64 second-generation TPUs to train the neural networks, all in parallel, with no access to opening books or endgame tables. After just four hours of training, DeepMind estimated AlphaZero was playing at a higher Elo rating than Stockfish; after 9 hours of training, the algorithm decisively defeated Stockfish 8 in a time-controlled 100-game tournament (28 wins, 0 losses, and 72 draws).[1][2][3] The trained algorithm played on a single machine with four TPUs.
...
Relation to AlphaGo Zero
Further information: AlphaGo Zero
AlphaZero (AZ) is a more generalized variant of the AlphaGo Zero (AGZ) algorithm, and is able to play shogi and chess as well as Go. Differences between AZ and AGZ include:[1]
AZ has hard-coded rules for setting search hyperparameters.
The neural network is now updated continually.
Go (unlike Chess) is symmetric under certain reflections and rotations; AlphaGo Zero was programmed to take advantage of these symmetries. AlphaZero is not.
Chess can end in a draw unlike Go; therefore AlphaZero can take into account the possibility of a drawn game.
AlphaZero vs. Stockfish and elmo
...
♚ Register at kingscrusher.tv and KC will invite you for a game!
♚ Or if you prefer bit.ly/chessworld
#KCChess #KCLeelaChess #KCMustSeeChess
♚TEAM kingscrusher.tv/team
♚TWITTER kingscrusher.tv/twitter
♚DISCORD kingscrusher.tv/discord
♚TWITCH / kingscrusher
♚REDDIT kingscrusher.tv/reddit
♚QUORA kingscrusher.tv/quora
♚DONATE paypal.me/kingscrusher
♚COURSES kingscrusher.tv/chesscourses
♚COURSES kingscrusher.tv/chesscourses

[@GnuReligion]

Mega Boom is a good title. They had to call in the POLICE to quell the disturbance in the TCEC lobby due to the discrepancy in the valuations between engines. My head still spins at the depth of Fish's analysis here. How in the world did it see that complicated rundown at the end? That should be impossible even for a computer.

[@Chiller0871]

Looks like it wasn't the end of an Era after all.

[@adamwu4565]

Era's rarely suddenly end, but rather slowly transition and peter out. It's very impressive how Stockfish and its developers have been keeping up with the neural nets, but looking more broadly, lately it's basically Stockfish alone vs a bunch of Neural Net AIs at the top tier right now, as the other AB Engines are falling off. The top 4 at TCEC18 were Stockfish, Leela, Alliestein and Stooflees, while over at the concurrent CCC14 it was Leela, Leelenstein, Stockfish and Stooflees, and it looks like Stockfish is finishing behind both Leela and Leelenstein.

[@utki17]

@@adamwu4565 NN has limits (always !).. so I think if SF keeps winning the way it is, it'll be almost impossible to ever catch up.. Its not like the next net of Leela gains 100 Elo or something..

[@adamwu4565]

@@utki17 There are limits to how quickly and consistently an AB engine like Stockfish can be improved as well. Time will tell.

[@PaulFurber]

A truly beautiful game by Stockfish. Thanks as always for bringing these to us KC. Your analysis is always so easy to understand for us mere mortals.

[@AlayanT]

For reference, the +150 evals by Stockfish correspond to a forced tablebase win eval. It's the equivalent of mate score, but instead of giving a number of moves to mate, Stockfish guarantees that it can force a won tablebase position. Stockfish's output and GUI interpretation are not very helpful to see the actual distance to forced TB win, but the slow creep upwards in the 150 evals correspond to a shorter distance left to TB win.

[@fruiteeplanet3125]

Thanks for detailed analysis ! . This game was amazing .

[@mita4e70]

Thanks for all the content throughout the years, KC!

[@whatpeosaid]

I love your videos. Thanks for the content during this lockdown

[@doncar9]

Wow! What a game. what a finish. Thanks for it KC, my go to man!

[@richardfredlund3802]

the fact leela moved the B on g4 back without it having done anything strongly indicates that the opening it was given was bad, and that bishop was misplaced,meaning leela was two tempos down from the beginning.

[@hw8556]

8:51 min. Here is one more alternative, which Stockfish gives in the analysis. With 32. Qg4 the calculated line at depth 67 ends with a evaluation of only +0,56. 32... Rb8! 33. Qf4 (Here 33. Bxg6 hxg6 34. Qxg6+ Kh8 = would only be a draw) 33... a4 34. Ne3 Tb1+!! 35. Bxb1 Qxb1+ 36. Qf1 Qxf1+ 37. Kxf1 f5! and the position is in balance. So Leelas mistake was 32... Qb3? that allowed Stockfish to play this really amazing endgame.

[@jkonrad]

Love the pompadour! :)

[@dancingcactus5271]

These AI game reviews are just incredible. You really do super impressive work reviewing these AI games and almost no one else really does it.

[@SuperAussi3]

Woow haha indeed SF depth is mega boom

[@orion_nectro.x]

Video titles are getting better by the day

[@joseraulcapablanca8564]

A fine game from Stockfish, played almost like Leela there and seemed to have exploited a horizon problem from Leela. Illuminating analysis, thanks KC keep up the good work.

[@Pioxital]

Also, the 50 moves rules has played against her

[@gmnotyet]

Game 65 would have been the GREATEST GAME OF ALL-TIME if a human had played it.

[@jarvism1449]

Do any other engines beat stockfish? If so what's the lowest rated?

[@shoman7205]

@12:32, instead of Ke3 why not Bf6:+?

[@not2tired]

Wouldn't +150 mean that you have such a huge advantage that the current game, as well as the next two games, are won by force?

[@Kevinrothwell1959]

:-D

[@mikestubbs1708]

So...which version of Stockfish are we talking about here?

[@2phil4u]

I don't know. This 150 evaluations. Without table base normally 149 is the highest so it is a table base win. Mate is normally found fast after having 150 eval. So two options.. 1.without table base the mate is found quick and evals are high 2.all the high evaluation is only founded cause a table base hit. And then lc0 is not this blind. I try it tomorrow. Maybe I find the mate in 25 soner.

[@petrabanjarnahor229]

What a mighty mismatch

[@vishalkhombare]

There is sure a major bug in Leela's training...She does not see dangerous king attacks ASAP ... I feel Leela being trained just for winning made her like this... she should have trained for winning in smallest possible move count.. That would have made her train herself for all possible brutal attacks for a quick result and we would have see an entirely different Leela NN personality

[@wondays654]

Vishal Khombare good idea, but won’t that just make her unable to see deep moves. Because if she is trained on winning in the shortest number of moves, she might ignore long term problems in games such as “position” and focus on immediate winning moves like “material gain” which we know is not always winning.

[@plasmaastronaut]

a boom is when the losing player suddenly realises he's lost, while the winning player has already known for some time that he's won

[@TheFulcrum2000]

No not at all, a boom is a sudden big jump in evaluation. Mostly from the *winning* player.

[@plasmaastronaut]

@@TheFulcrum2000 yes i know thats what 99% of kids who are coming to tcec now its got popular think. The reason why the original meaning of boom was lost was because it was a rare and fun event. Bored watchers who watched 16hrs a day, and who didn't fully understand boom started posting boom for less and less dramatic/pronounced changes in evaluation purely to enjoy the hype. Quickly the meaning of boom was lost as dozens of bored watchers all posted boom. it got to the point where boom was posted every single ply, even for a + - 0.01 change in eval.

[@TheFulcrum2000]

@@plasmaastronaut All true, but does not change the fact the way i describe it is how it all started many many seasons ago.

[@plasmaastronaut]

@@TheFulcrum2000 u need on work on your logic, what I say can't be all true AND your description be true at the same time. Or perhaps u didn't read what i wrote.

[@kingscrusher]

Don't think it is that precise. If you do !boom then it indicates: "ModeratorVerifiedMoobot: the concept of the "boom" is a TCEC invention [current general consensus defines BOOM as meaning "eval explodes away from 0.00" - its opposite being a MOOB meaning "eval implodes back in the general direction of 0.00"]" - so just an Eval exploding away from 0

[@kajaskov6270]

How deep did SF calculate?

[@Mikeontube]

around depth 85

[@youneverknow5555]

what about leela dept

[@Mrkontrol007]

Hope Leela will return her lead.

[@Pioxital]

I'm very depressed by this tcec season...leela getting her ass kicked

[@140TrillionSuns]

TCEC gone to far engines are using bluff

[@BeammeupSpotty]

ugh, again, non-stop ads. grrrr

[@Badbentham]

If the best that black can come up with in this (already pretty dubious; Tony Miles liked it...) opening is 5...Bg4-d7, then it is already extremely sad (white is close to positionally winning after it) news...

[@ariadonatellacongedo2727]

Leela seems to have regressed and returned to the slides of a long time ago ... clumsy in the middle game and disastrous in the finals ... if Leela had a merit, in addition to creativity in building fantastic romantic games, it was also creativity in finding personal ways to draw . All this seems lost. Could version 0.25 have any problems? Ever since it came out, she seems unrecognizable.

[@wiltonh]

Alpha Zero crushes Stockfish all the time though

[@balthazarbeutelwolf9097]

Not all the time, at one time. That was about 3 years ago when Alpha Zero played the standard download SF8. A0 has not played since then, and all these speculations sound like the "would Fischer crush Karpov" questions in the 1970s when the answers moved from "probably" to "no chance" as the 1970s progressed and Karpov won every tournament under the sun. We are upto SF11 now, and this was a development version, running on super-strong hardware. Very quickly after the A0-SF8 match, all alpha-beta-engines learned about thorn pawns, and henceforth fishtest had various Leela versions to spar against, and get better positionally as well as tactically.

[@murattuna3524]

Mate just because A0 shadily beat a crippled SF8 doesnt mean current one can rain hellfire on it.Hell,Leela would rape A0 right now.

[@Rick11N]

Who is stockfisch un this game: blacks or whites ? I think that blacks is always to obliged to play defensefely ! If it is Leela, what a bad play in this game against the serious play and attack of stockfish (version ?). I am new in this best chronicle of kingcrusher (Firoujza ?) Thanks and continue like that. !!!👍😄‼️.

[@martinmartin6300]

Yeah Leela is very bad AI. Leela should have beaten Stockfish easily after this period ot development and training.