LOL just a day ago you say the Rossolimo is a bad opening and the best chess engine plays it. Goes to show how bad you are at chess and how ENORMOUS the difference is.
False equivalency. Stockfish can see far farther and calculate far more than any human can. You wouldn't be able to play any position the way it plays it so what's the point of this comment when it dosent defeat the original sentiment. Pin of shame.
This statement shows that you know nothing about chess lmao. Even levy knows that top gms and engines play the rossolimo. Did you watch he last video on the worst chess openings. Well all the gms play those as well. What levy is talking about is that he doesn't recommend these openings to beginners.
You should do stockfish 15 vs 8 again but have them swap sides after 20 or 30 moves. See if the engine is strong enough to come from behind. Something like that. Be interesting to see how the strong engine plays from behind instead of ahead the whole way
That's a really cool idea. Wait for the evaluation by the stronger engine to be something like +-0.5 and then swap. I think anything higher would probably be impossible to come back from at engine level, but who knows. Might be a fun parameter to play around with.
@@abelgreen5046 30 seconds is wayyyyyy to much for Stockfish. Stockfish will always win given that much time. More like half a second stockfish and up to an hour for Magnus
There should be a “Stockfish vs the World” paying obvious homage to Kasporov. I’d be very interested to see the greatest chess minds collaborate on a large scale like that
Maybe if you disallow Stockfish to calculate on the opponents move and give it 1 second per move, and the World-Team 1 day per move. Even then I don't know if that would make up for 1000 points of ELO difference.
18:31 I love stockfish and its king walks so much lmfao. It deadass goes "If king doesn't move how can his subjects follow" and makes it its main strategy
this definitely is really cool and I hope it's ongoing but with more creative elements, handicaps eras etc. however Magnus played titled Tuesday yesterday and I have been fully expecting a Gotham recap uploaded today and it seems like that's not going to happen in which case he's a lazy POS tbh
@@glajolambokla I don't think Levy will cover Magnus on TT because he has retired from such competitive weekly online chess so it would make him jealous/sad.
@@s-x5373 Let me just assume that you never tried to get some random software from even the late 80s or early 90s running on a modern machine... Even with things like DOSbox and VMWare around this can be a huge pain. And in this case the chance is high, that the software was written to run on architecture that is basically still around (X86 IBM-PC) in a language that is still widely used (e.g. C/C++). If we're talking 70s the hardware might be sth like a PDP-7/9/11 on an operating system before UNIX, written - if you're lucky - in FORTRAN. If you're unlucky in assembler. For example Kaissa ran on an ICL 4/70... good luck finding and emulator for that. So you might be better off getting the papers / source code from back then and reimplement everything from scratch hoping to approximate at least half way decently the strength of the machine back then.
Crazy to think that Stockfish 8 is 509 elo points higher than Magnus atm , and Stockfish 15 is 330 points higher than THAT The level that these engines play at is absolutely mind-blowing
Those things play on an entirely different level. I once played against stockfish for fun. I progressed a pawn, but in return, it pushes a pawn elsewhere. I was like “huh, it hung it’s knight?” I took it. Then it proceeded to tear my kingside a new asshole
It is important to note these are engine ratings. Rating is relative to the pool of players you are playing. For example, imagine there was a GM-only rating and GMs started at 1500 like on lichess, then magnus carlsen would be around 1800-1900. Stockfish got to 3700 crushing other engines, not humans
@@pawncube2050 yeah, but it doesn't actually matter what Stockfish's exact rating against humans would be. Given a half way decent desktop computer it will beat every human likely 100% of the time in a classical game. I guess the ELO curve can not accurately measure how much better stockfish really is - so calling it 3500 is as good as calling it over 9000 from the human standpoint. Even if we consider that Stockfish's strength also depends on the hardware used etc.
"Three Queens!!" Engines are just so sick. Great content with some teaching points for us carbon based units. Taking space and patience are key lessons. Thanks!
"Only 96.7% (accuracy) for stockfish 15" had me rolling. Never in my life would I have thought I would would heard the words "only 96.7% accuracy" unironically.
Some random info about chess engines 1ms is quite a lot of computation time, a good computer can get many millions of variations calculated in that time(also it will skip stuff on the way that seems bad). Still when I do a basic alpha beta pruning which is the old way of doi g chess engines it can easily take 50ms to run 5-7 moves deep. So 1 ms is pretty damn impressive for how the amount of moves scales. Edit oh 0.1 yeah that's stupidly low :^)
Man I remember back in 2018 when A0 was smashing SF8 which was by far the best for the time. I even read their book/paper on the matches. The most memorable game has to be the one where A0 locked up SF8's queen in the corner of a fianchetto structure. I mean, SF8 got beat so bad it looked like it just learned chess the day before. I think it would be a great video idea to analyse those games with SF15 and see how it disses its past self, and also to check if A0 made uncaught mistakes.
speaking that andrew tang beat stockfish on equal time at 30s (stockfish 8) i think classical player hikaru could beat a 1ms using stockfish 15 at least 3/10 times and get like 4-5 draws
@@autosemimatic6071 Lichess's "play against the computer" level 8 *is not* Stockfish 8. It's Fairy-Stockfish 14 with a custom level 8 made by Lichess limited in thinking time and depth so humans can still sometimes win.
@@autosemimatic6071 Still, good content, insane moves. I wouldn't be surprised if something like a rook and a knight sacriface for 8 pawns somehow happened. It's fun to watch, especially with Levy's commentary.
@@pay-g2163 The issue is that with stockfish 15 playing against itself it would see all the lines every time it offered itself a sacrifice so just wouldn't take. It would just be a display of perfect positional solidity from both sides leading to an inevitable draw
Video idea: you play hikaru or other GM buuut you can consult stockfish 5 times, this would be reallo cool because there are gonna be lots of mindgames involved, he will doubt if you blundered or if that is one of the stocfish moves (he wont know when you use it)
i once let stockfish play against itself, with 1 second per move vs 1 minute per move, and white won with 3 disgusting brilliant sacrifices on empty squares. it was the best combination i have ever seen
Please use increment type time control tho. All engines are built for that. Also, CCCC and TCEC run these games on much bigger hardware, sometimes against 'weaker' engines, especially the start position event earlier.
"Nobody is sacrificing this rook for content. The game is over." Lmaoooo, that's actually hilarious because its so true. A computer is not playing for fun or content, just wins. 😂
Not sure if this is really gonna be seen much - but would absolutely watch Stockfish vs Stockfish in many different openings. Would be very fascinating.
better analogy for 0.1 millisecond is to say that if they played through the whole tournament that just passed it would still be too fast for a human to react before it finished
Would love to see a Stockfish 15 vs Stockfish 15 video! Of course, I actually would love to see a video where you stream a 15 vs 15, but instead of the engines going up against each other directly, you play both sides for them. That way, when the engine gives you a result of multiple different best moves, chat can vote for which one that side of the board actually goes with.
I love your engine vs engine vids. It’s so cool to watch the ice cold ruthlessness of a machine crush chess. Even thought it might have ruined the game overall
I just checked out Gotham's "The Most Famous Chess Game of All Time" and was astonished at how much healthier he looks now. Back then he looked pale and drawn, but now he looks robust with a ruddy, healthy complexion. His muscle shirts are emblematic of his newfound health and thus should not be criticized.
This is honestly up there as one of your best videos, incredibly fast analysis which is very insightful, with stories which put into context what these bots are doing. Awesome stuff.
There should be a whole new series of this engine vs engine You might consider giving different odds to the weaker engines, would be a great content i suppose
I find it funny that stock fish doesn’t need 3 queens in the final game and you would only see like beginners like me make a few queens to do ladder mate. But stockfish deemed it the best way to play, so folks make a lot of queens to disrespect your opponent no matter any level
Stockfish is the god of chess engines? I haven't been following the engine side for like a year or two but I thought Leela had taken the edge? Stockfish uses brute force calculation and Leela uses machine learning (AI) to improve. If the scales have not already tipped in the favor of the AI, they will very soon.
Stockfish now also uses neural networks for its engine (since iteration 12 or 13 I believe), it even selects the best performing architectures from a platform called Fishtest!
Levy, you don't need to become a GM. Your teaching skills are much more appreciated than your title. Thanks to you I could beat a random person 3-0 because I knew what I was trying to do the entire game and they only did during the opening.
Levy + 500 elo = Magnus Magnus + 500 elo = Stockfish 8 Stockfish 8 +500 elo = Stockfish 15 In human terms this video is basically Levy playing a chess 960 match (because these engines don't know theory) against Magnus and it went absolutely how you would expect
Stockfish 15 vs Stockfish 15 would be hilarious I'm sure. Also it would be interesting to see how much of a time advantage you'd need to give Stockfish 8 for it to beat Stockfish 15.
20:33 The French defense is not refuted by engines or ghosts like A0. The Winawer in fact is still drawn on ICCF, where Stockfish 15 gets HOURS to DAYS to play a single move.
@@maximussaktish On the contrary, I pointed out the engines have hours to days to ENSURE the opening is refuted as Levy claims, and quite the opposite is happening: The Winawer becomes a draw in the main lines with frustrating regularity. I've faced it many times myself on ICCF, and it's just rock solid there. I also want to point out 13:36 that again, Levy has no clue about the current state of chess engine behavior. Stockfish will NEVER play d3 anti-Berlin unless forced to with an opening book, because it immediately considers the move losing ALL advantage (humans have a tougher time, but the engines laugh at d3 as black). As a result, Stockfish only plays the best evaluated move, which is purely a horizon effect, as after about 12 moves later, it starts to drop to 0.00 just as it does with d3.
@@user-df4zw7yb4v And why would having more processing power matter? The neural network will saturate either way And if you are talking about more processing power to play the game then that makes it literally not a fair. Because otherwise you could take Leela in your computer with a bunch of graphics cards, make it play against Stockfish on your phone and when Leela wins claim that she is better.
What if you make Stockfish constantly play with itself analyzing every single match and doing some conclusions? It already probably works like that, but this kind of idea still feels so crazy and unfair
Is the discontinued AlphaZero engine still accessible? If so, could you pair Stockfish 15 against it to see if it can avenge it's older counterpart? Awesome video, fascinated by these engine games!
It was never accessible, but estimates say Alphazeros's rating was around 3450, both Stockfish and Leela are approximately 300 points higher rated than that.
At 10:30 you said this about Stockfish15, "It doesn't care!" That immediately reminded me of a line from the movie Terminator. "It can't be reasoned with, it can't be bargained with, it doesn't feel pity of remorse or fear, and it absolutely will not stop. Ever. Until you are dead."
5:43 the stopping a car comparison doesn't even get close. Your brain can't even process a visual signal in 0.1 milliseconds let alone react to it. By the time you stop your car these engines will have completed hundreds of games at that speed.
He says later it’s 0.1 milliseconds, so that’s _2500 moves_ in the time it takes you to blink. By the way, it would still win against a grandmaster with as much time as they want. Fascinating.
that 4th game is why the french or caro kann work against humans but not engines. no human can create an absolute boa-constrictor of a position where black has 15 possible moves and they all lose. no human can keep up with the 24 remaining pieces and making sure that every single one of them is locked into a position where they cant move without crumbling to ash
This was a fun video. Here's a possible follow up idea: first, pair Stockfish 8 against Stockfish 12, let's say. Once Stockfish 12 acquires a significant advantage in the game (assuming it can defeat Stockfish 8), you then let it swap places with Stockfish 15 and see if 15 can make a comeback against 12. Or something along those lines...
For context on how damn fast a millisecond is: human reaction time is ~230 milliseconds. For immediate, direct reactions that take 0 thinking or decision making. These engines would be able to checkmate you in the time it took you to see they had played 1.e4
Stockfish 15 holding off on promotion in game 3 until it walked it's king to knock out the knight is equivalent to a bully holding a kid down and slowly dripping a loogie onto his face
I wonder how insane kids will be in the future. Playing against the strongest opponent to ever live? A player who's also super helpful in explaining your inaccuracies? That has to be a game changer.
so the whole first game was basically levy pressing play and getting a supersonic checkmate served in under a second, still hot to the touch? That's like turning a light switch to instantly feel bad about your chess
LOL just a day ago you say the Rossolimo is a bad opening and the best chess engine plays it. Goes to show how bad you are at chess and how ENORMOUS the difference is.
Shin of Pame
Haha fast pin of shame
Shit of lame
False equivalency. Stockfish can see far farther and calculate far more than any human can. You wouldn't be able to play any position the way it plays it so what's the point of this comment when it dosent defeat the original sentiment. Pin of shame.
This statement shows that you know nothing about chess lmao. Even levy knows that top gms and engines play the rossolimo. Did you watch he last video on the worst chess openings. Well all the gms play those as well. What levy is talking about is that he doesn't recommend these openings to beginners.
Stockfish crushes itself, then analyzes its mistakes. What a gigachad.
That's called self introspection...absolute chad trait
@@jyotiradityashukla4569 i was going to say that, this machine is a step forward to become a super human
So one omnipotent engine in 3 persons... Feels suspiciously familiar.
I just cringed so hard.
@@MirioLOL how that’s it not even cringe
You should do stockfish 15 vs 8 again but have them swap sides after 20 or 30 moves. See if the engine is strong enough to come from behind. Something like that. Be interesting to see how the strong engine plays from behind instead of ahead the whole way
That's a really cool idea. Wait for the evaluation by the stronger engine to be something like +-0.5 and then swap. I think anything higher would probably be impossible to come back from at engine level, but who knows. Might be a fun parameter to play around with.
Come from behind💀💀
a possible case might be sf15 gets to winning positions, switches, but then sf8 has no idea its winning and throws lol
@@vanshjain3428 yeah funny..
Gotham should really consider this
17:22 : stockfish 15 awarding stockfish 15 a brilliant move be like : yes incredible i definitely would not have seen it
game review isnt sf 15
*obama medal meme*
@@_wetmath_ it's an equivalent implementation
Techno?
@@robinhood5448 RIP :(
I personally love watching Stockfish, Magnus, Leela and any other chess engines at your channel!
I chuckled lol
Can confirm Magnus Carlsen is a chess engine.
Lol
I wanna see magnus vs stockfish, where stockfish gets 30 seconds and magnus 3h
@@abelgreen5046 30 seconds is wayyyyyy to much for Stockfish. Stockfish will always win given that much time. More like half a second stockfish and up to an hour for Magnus
Stockfish 8: makes brilliant move in 0.2 seconds
Stockfish 15: makes brilliant move in 0.1 seconds
Me: gambits queen in 10 moves
They actually were playing moves in 0.0001 seconds in the first game
@@Samuelissad 0.001s
@@henryk7033 A second has 1000 milliseconds and the engines played in 0.1 millisecond. So...
@@Samuelissad At 0:54 he says they had 1 millisecond per move
@@Samuelissad or did they play faster than the time they had?
I love how Stockfish casualty saw mate in 12 in 10,000th of a second
i think it already had calcualated that it will be mate in 12 bevore the move happend
@@imran34 then it wouldve calculated mate in 13 or 14 or 15 and so on
@@AquaficMC No, it calculated all the possible moves after the move it did and saw that if the opponent would play that move it would be m12
wrong. it calculated checkmate in 53 after the very first move of the game.
@GothamChess Why can't you let them think longer, that would replicate the alpha zero vs stockfish 8 match much better....
Stockfish 15: *rates itself to have had a practically perfect game.*
There should be a “Stockfish vs the World” paying obvious homage to Kasporov. I’d be very interested to see the greatest chess minds collaborate on a large scale like that
Maybe if you disallow Stockfish to calculate on the opponents move and give it 1 second per move, and the World-Team 1 day per move. Even then I don't know if that would make up for 1000 points of ELO difference.
@@Nighty93_ i think stockfish is much weaker with only 1 sec per move and could probs be defeated my the world
I know this is a comment 1 year ago but that would be sick an engine VS the greatest chess minds in a collective to beat it that would be insane
Stockfish 8: it’s gonna be great!
Gets checkmated with 3 queens
18:31 I love stockfish and its king walks so much lmfao. It deadass goes "If king doesn't move how can his subjects follow" and makes it its main strategy
I like this, this should be a new series, Computer Chess War or something.
this definitely is really cool and I hope it's ongoing but with more creative elements, handicaps eras etc. however Magnus played titled Tuesday yesterday and I have been fully expecting a Gotham recap uploaded today and it seems like that's not going to happen in which case he's a lazy POS tbh
@@glajolambokla I don't think Levy will cover Magnus on TT because he has retired from such competitive weekly online chess so it would make him jealous/sad.
GM Matthew Sadler at his TH-cam channel Silicon Road covers many more engine games if you're interested.
Isn't there already a chess engine World championship??
it's already a thing.
Might be interesting to pit older engines against each other with a view to see how engines have improved, like engines from the 70s, 80s and 90s.
Agadmator did that not so long ago.
Engines from that long ago are actual trash against anything modern.
It's very likely a huge pain in the behind to get a chess engine from the 70/80s (if you can even find the source for it) to run on a modern computer.
@@tetsi0815 I dont think that much, it's just not conventionnal, but these are machines easy to emulate on current hardware
@@s-x5373 Let me just assume that you never tried to get some random software from even the late 80s or early 90s running on a modern machine... Even with things like DOSbox and VMWare around this can be a huge pain. And in this case the chance is high, that the software was written to run on architecture that is basically still around (X86 IBM-PC) in a language that is still widely used (e.g. C/C++). If we're talking 70s the hardware might be sth like a PDP-7/9/11 on an operating system before UNIX, written - if you're lucky - in FORTRAN. If you're unlucky in assembler. For example Kaissa ran on an ICL 4/70... good luck finding and emulator for that. So you might be better off getting the papers / source code from back then and reimplement everything from scratch hoping to approximate at least half way decently the strength of the machine back then.
Crazy to think that Stockfish 8 is 509 elo points higher than Magnus atm , and Stockfish 15 is 330 points higher than THAT
The level that these engines play at is absolutely mind-blowing
Those things play on an entirely different level. I once played against stockfish for fun. I progressed a pawn, but in return, it pushes a pawn elsewhere. I was like “huh, it hung it’s knight?” I took it.
Then it proceeded to tear my kingside a new asshole
@@tp7033 lulz
It is important to note these are engine ratings. Rating is relative to the pool of players you are playing. For example, imagine there was a GM-only rating and GMs started at 1500 like on lichess, then magnus carlsen would be around 1800-1900. Stockfish got to 3700 crushing other engines, not humans
@@pawncube2050 yeah, but it doesn't actually matter what Stockfish's exact rating against humans would be. Given a half way decent desktop computer it will beat every human likely 100% of the time in a classical game. I guess the ELO curve can not accurately measure how much better stockfish really is - so calling it 3500 is as good as calling it over 9000 from the human standpoint. Even if we consider that Stockfish's strength also depends on the hardware used etc.
Brilliant pfp brother
"Three Queens!!" Engines are just so sick. Great content with some teaching points for us carbon based units. Taking space and patience are key lessons. Thanks!
Only reason to make three queens is to flex on humanity.
Actually it just finds the fastest way to mate
Making three queens is actually a pretty human thing to do
This could means that Stockfish is selfaware...nah
Wow i didn't expect stockfish to beat stockfish
Poor stockfish has gone so depressed that is now have gone to self harm
His little brother
I would like to see this game but giving time odds or processor odds to Stockfish 8 until it wins, just to compare
I agree would really display how much of a gap there is more
That's a really great idea!
I want to see stockfish 15 destroy
Yeah and imagine if you gave stockfish 16 the best processors and sf 8 the worst would be fun to see how it crushes his counter part
Please keep these engine games coming Levy! (Like more from TCEC). Engine games are always so fascinating because of how high level they are.
Agree
I was going to comment basically the same thing, but yeah, analyzing engine games are fascinating. It's like walking through an art gallery for me.
TCEC is way crazier than these 0.1/5/20 second games b/c they full-on play classical chess
The games are just unbelievably superhuman in those formats
If u want more engines games, check out the silicon road channel run by GM Matthew Sadler th-cam.com/users/SiliconRoadChessvideos
Awesome content
@@webbowser8834 I completely agree
For a month I thought Stockfish was a type of fish
"I beat my identical twin. We think everything the same. Here is every mistake he made." -Stockfish.
"Only 96.7% (accuracy) for stockfish 15" had me rolling. Never in my life would I have thought I would would heard the words "only 96.7% accuracy" unironically.
Some random info about chess engines 1ms is quite a lot of computation time, a good computer can get many millions of variations calculated in that time(also it will skip stuff on the way that seems bad). Still when I do a basic alpha beta pruning which is the old way of doi g chess engines it can easily take 50ms to run 5-7 moves deep. So 1 ms is pretty damn impressive for how the amount of moves scales.
Edit oh 0.1 yeah that's stupidly low :^)
I CAN CLOCK 0.1 WITH MIN MAX HEHEHE, I ALSO USE BOGOSORT FOR SORTING ELEMENTS. YOU SHOULD WORK FOR ME TBH.
Depends on how strong your cpu is the main thing on the topic of calculation
have u tried using transposition to reduce time?
0.1 isn’t a millisecond 0.1 is 100 milliseconds 0.001 is 1 millisecond.
@@youtubeshypocrisy 0.1 milliseconds. how many people don't watch the whole video lol
Man I remember back in 2018 when A0 was smashing SF8 which was by far the best for the time. I even read their book/paper on the matches. The most memorable game has to be the one where A0 locked up SF8's queen in the corner of a fianchetto structure. I mean, SF8 got beat so bad it looked like it just learned chess the day before.
I think it would be a great video idea to analyse those games with SF15 and see how it disses its past self, and also to check if A0 made uncaught mistakes.
Now I wanna see Stockfish 15 with .1 milliseconds per move vs Hikaru with like as much time as he wants
Hikaru won't waste his time playing stockfish for nothing
@@chessisfun1592 Not nothing, views
@@-zelda- true
speaking that andrew tang beat stockfish on equal time at 30s (stockfish 8) i think classical player hikaru could beat a 1ms using stockfish 15 at least 3/10 times and get like 4-5 draws
@@autosemimatic6071 Lichess's "play against the computer" level 8 *is not* Stockfish 8. It's Fairy-Stockfish 14 with a custom level 8 made by Lichess limited in thinking time and depth so humans can still sometimes win.
Stockfish 15 vs Stockfish 15, yes, please. This content is amazing.
like 9 million draws prob no wins/losses
@@autosemimatic6071 Still, good content, insane moves. I wouldn't be surprised if something like a rook and a knight sacriface for 8 pawns somehow happened. It's fun to watch, especially with Levy's commentary.
@@pay-g2163 The issue is that with stockfish 15 playing against itself it would see all the lines every time it offered itself a sacrifice so just wouldn't take. It would just be a display of perfect positional solidity from both sides leading to an inevitable draw
Stockfish 15 mirror match, but with a material imbalance, so maybe a rook and a pawn for two pieces, who would win, that could be fun
Video idea: you play hikaru or other GM buuut you can consult stockfish 5 times, this would be reallo cool because there are gonna be lots of mindgames involved, he will doubt if you blundered or if that is one of the stocfish moves (he wont know when you use it)
Like Deep Blue v Kasparov when it ran onto a loop and just picked a random move.
I want to see a game of Stockfish 15 vs Stockfish 15 with maximum time to let it think. It would be interested to have that game.
playing a drawn game from move 1 🤫👀
@@allysson_ exactly
There is no "maximum" time, you can let the engine think for weeks and it won't just stop
So basically, that kind of game would last forever
it would be interesting if one engine has 20 sec and the other engine maybe 5 sec or so and see if it can hold the position
@@itssimplyme1296 Yaa, that's pretty good
Stockfish 8: "When I get bad, I stop being bad and be awesome instead".
Stockfish 15: "please define prehistoric word 'bad', unknown keyword error"
Frankly, no one on the entire platform of TH-cam can beat Levy at thumbnails. He's the Stockfish of that game.
Apart from hikaru
I love these engine vs engine recaps. They're always insane
That's the power of Stockfish family ❤
Your channel is awesome, girl!
I call the 1 millisecond time control "Ultra Hyper Bullet"
i once let stockfish play against itself, with 1 second per move vs 1 minute per move, and white won with 3 disgusting brilliant sacrifices on empty squares. it was the best combination i have ever seen
Move 22 is where the juice is at
1. e4 c5 2. Nf3 e6 3. d4 cxd4 4. Nxd4 a6 5. Nc3 Qc7 6. Bd3 Nc6 7. Be3 Nf6 8. O-O
Bd6 9. Nxc6 dxc6 10. f4 e5 11. f5 Qe7 12. Kh1 b5 13. Qf3 Bc5 14. Bg5 Bd4 15. a4
Bb7 16. Bh4 O-O 17. Ne2 Bxb2 18. Rab1 Ba3 19. Ng3 h6 20. Nh5 Kh7 21. Qh3 Rfd8
22. Bg5 Rd6 23. Nxg7 Rh8 24. Ne6 fxe6 25. Bc4
25... bxc4 26. Bxf6
Qxf6 27. fxe6 Qxf1+ 28. Rxf1 Rf8 29. Rb1 Bc8 30. Qxa3 Rfd8 31. h4 Bxe6 32. Qc5
Bf7 33. Qxe5 Rd1+ 34. Rxd1 Rxd1+ 35. Kh2 Rf1 36. Qd6 Bg6 37. Qxc6 Rf4 38. Qb7+
Kg8 39. Qxa6 Kh7 40. Qb7+ Rf7 41. Qc8 Bxe4 42. Qxc4 Re7 43. a5 Ba8 44. Qd3+ Be4
45. Qe3 Kg7 46. a6 Kf7 47. a7 h5 48. Qf4+ Kg7 49. Qg5+ Kf8 50. Qf6+ Ke8 51. Qh8+
Kd7 52. Qxh5 Kd6 53. Qh8 Ke6 54. a8=Q Bxa8 55. Qxa8 Kf5 56. Qf8+ Ke6 57. h5 Rc7
58. h6 Rd7 59. Qc8 Kd6 60. Qxd7+ Kxd7 61. h7 Ke6 62. h8=Q Kf5 63. Qe8 Kf6 64. g4
Kg7 65. g5 Kh7 66. Qf7+ Kh8 67. Qa7 Kg8 68. g6 Kh8 69. Qh7# 1-0
I personally love the Stockfish and Computer chess in general and you made the break down of the games great and enjoyable.
Wonder if magnus could beat stockfish 15 on .1 milliseconds per move if given like 15-30m
Wonderful question tbh always love interesting questions like these
@@thatloserpig thank u!
You should try stockfish that knows openings vs the one who doesn't but the one who doesn't gets like 5 seconds more
Please use increment type time control tho. All engines are built for that.
Also, CCCC and TCEC run these games on much bigger hardware, sometimes against 'weaker' engines, especially the start position event earlier.
"Nobody is sacrificing this rook for content. The game is over." Lmaoooo, that's actually hilarious because its so true. A computer is not playing for fun or content, just wins. 😂
Not sure if this is really gonna be seen much - but would absolutely watch Stockfish vs Stockfish in many different openings. Would be very fascinating.
"Stockfish 15 vs Stockfish 15, I'll do it."
Let's gooooo
I can beat myself from a few years ago. Stockfish isnt that impressive…
Yo 🔥th-cam.com/video/rdJ9bsN7JAw/w-d-xo.html,.
I can beat myself from 10 minutes ago, given that I remember what moves past me made.
@@rickrolled3666 no :)
Ok? Beat stockfish.
better analogy for 0.1 millisecond is to say that if they played through the whole tournament that just passed it would still be too fast for a human to react before it finished
Would love to see a Stockfish 15 vs Stockfish 15 video! Of course, I actually would love to see a video where you stream a 15 vs 15, but instead of the engines going up against each other directly, you play both sides for them. That way, when the engine gives you a result of multiple different best moves, chat can vote for which one that side of the board actually goes with.
The chat idea seems like it would destroy the whole point of the video
I love your engine vs engine vids. It’s so cool to watch the ice cold ruthlessness of a machine crush chess. Even thought it might have ruined the game overall
I just checked out Gotham's "The Most Famous Chess Game of All Time" and was astonished at how much healthier he looks now. Back then he looked pale and drawn, but now he looks robust with a ruddy, healthy complexion. His muscle shirts are emblematic of his newfound health and thus should not be criticized.
>Levy shows stockfish game to stockfish
"This stockfish 15 guy is quite good, I would play all these moves myself. Perfect game."
~Stockfish
Day 7 of translating Levy’s titles in Roman: Stockfish sdruma Stockfish
Yo 🔥th-cam.com/video/rdJ9bsN7JAw/w-d-xo.html,
Is there a roman word for "stockfish"?
@@SmoggyCold not really, the Italian word would be “stoccafisso” but it’s not Roman
What the hell is roman, yiu mean Romanian?
The rook sac on the last game was crazy. The eval bar slowly crept up to the top as Stockfish 8 realized it was doomed
This is absolutely fascinating content, please make more of this! I love watching engines play
This is honestly up there as one of your best videos, incredibly fast analysis which is very insightful, with stories which put into context what these bots are doing. Awesome stuff.
Saw it, instantly clicked it
Ok
So ok
@@iraniam i do love my chess AI what can I say
@@vatsalgaur3550 O+K
The seconds for moves aren't like the dominant factor of move quality, but combined with your computer's strength it is.
Love the engine battle videos, keep making more of them :)
i find it so infinitely beautiful that i can play the best chess player ever basically anytime i want, those chess engines are a blessing
There should be a whole new series of this engine vs engine
You might consider giving different odds to the weaker engines, would be a great content i suppose
I find it funny that stock fish doesn’t need 3 queens in the final game and you would only see like beginners like me make a few queens to do ladder mate. But stockfish deemed it the best way to play, so folks make a lot of queens to disrespect your opponent no matter any level
Can Stockfish be programmed to draw (not win)? It would be interesting to see it navigate obtaining a draw versus obtaining a win.
So the biggest update stockfish made from 8 to 15, is learning the ability to BM lmao
Stockfish is the god of chess engines? I haven't been following the engine side for like a year or two but I thought Leela had taken the edge? Stockfish uses brute force calculation and Leela uses machine learning (AI) to improve. If the scales have not already tipped in the favor of the AI, they will very soon.
Stockfish now also uses neural networks for its engine (since iteration 12 or 13 I believe), it even selects the best performing architectures from a platform called Fishtest!
@@Nathan-ey6ps oh cool! Thanks!
Levy, you don't need to become a GM. Your teaching skills are much more appreciated than your title. Thanks to you I could beat a random person 3-0 because I knew what I was trying to do the entire game and they only did during the opening.
Levy + 500 elo = Magnus
Magnus + 500 elo = Stockfish 8
Stockfish 8 +500 elo = Stockfish 15
In human terms this video is basically Levy playing a chess 960 match (because these engines don't know theory) against Magnus and it went absolutely how you would expect
What’s next, Stockfish 15 + 500 elo = Stockfish 22?
Levy didn’t put into perspective how quick 0.1 milliseconds is. The average brain takes 100 milliseconds to send signals to your body.
I like the videos where you deep analyze crazy stuff engines do and blow my mind
Definitely enjoyed that. I also like you pointing out the little strategy snippets that could be employed in a game.
I always wondered like what is considered max elo? Would any computer over 4000 rating have it completely solved?
There is no max elo because chess isn't solved, so there can always be a better engine
"The French is garbage, unless your human and humans are garbage" - AlphaZero
Stockfish 15 vs Stockfish 15 would be hilarious I'm sure. Also it would be interesting to see how much of a time advantage you'd need to give Stockfish 8 for it to beat Stockfish 15.
20:33 The French defense is not refuted by engines or ghosts like A0. The Winawer in fact is still drawn on ICCF, where Stockfish 15 gets HOURS to DAYS to play a single move.
To be fair if you need days from an engine just to squeeze a draw it might as well be refuted
@@maximussaktish On the contrary, I pointed out the engines have hours to days to ENSURE the opening is refuted as Levy claims, and quite the opposite is happening: The Winawer becomes a draw in the main lines with frustrating regularity. I've faced it many times myself on ICCF, and it's just rock solid there.
I also want to point out 13:36 that again, Levy has no clue about the current state of chess engine behavior. Stockfish will NEVER play d3 anti-Berlin unless forced to with an opening book, because it immediately considers the move losing ALL advantage (humans have a tougher time, but the engines laugh at d3 as black). As a result, Stockfish only plays the best evaluated move, which is purely a horizon effect, as after about 12 moves later, it starts to drop to 0.00 just as it does with d3.
Levy : stockfish is the strongest computer
Alpha zero: am i joke to you?
Yes it is. Current SF absolutely claps A0
Also stolen comment
Yo 🔥th-cam.com/video/rdJ9bsN7JAw/w-d-xo.html,.
@RAISTAR exe And why do you think that? Because Leela is basically like A0 but stronger and even after years of development it can't catch up
@@-zelda- because alphazero is developed by an entire team at google which has access to way more processing power
@@user-df4zw7yb4v And why would having more processing power matter? The neural network will saturate either way
And if you are talking about more processing power to play the game then that makes it literally not a fair.
Because otherwise you could take Leela in your computer with a bunch of graphics cards, make it play against Stockfish on your phone and when Leela wins claim that she is better.
7:52 levi, an inaccuracy means a weak move, not a blunder or mistake. Stockfish 15 basically says that the move was too slow. Henceforth a weak move.
Those who all want Stockfish 15 vs Stockfish 15 🤚🤚.
What if you make Stockfish constantly play with itself analyzing every single match and doing some conclusions? It already probably works like that, but this kind of idea still feels so crazy and unfair
See: AlphaZero
Is the discontinued AlphaZero engine still accessible? If so, could you pair Stockfish 15 against it to see if it can avenge it's older counterpart? Awesome video, fascinated by these engine games!
It was never accessible. The matches were done privately by Deepmind
It was never accessible, but estimates say Alphazeros's rating was around 3450, both Stockfish and Leela are approximately 300 points higher rated than that.
That triple queen promotion in the last game, just a show of pure dominance.
18:30 lmao broke black's ankles with that juke.
At 10:30 you said this about Stockfish15, "It doesn't care!" That immediately reminded me of a line from the movie Terminator.
"It can't be reasoned with, it can't be bargained with, it doesn't feel pity of remorse or fear, and it absolutely will not stop. Ever. Until you are dead."
5:43 the stopping a car comparison doesn't even get close. Your brain can't even process a visual signal in 0.1 milliseconds let alone react to it. By the time you stop your car these engines will have completed hundreds of games at that speed.
16:30 so Stockfish 20 will be able to play 5D Chess With Multiverse Time Travel? Pog
I love watching Stockfish playing with himself
Game opens: SF15, 'aaand I've won'... SF8, 'what are you talking ab...' 3ms later... 'why do I even play this game!?' :[
that king walk was quite something
For human scale, it takes about 250ms to blink. Stockfish was making 250 moves in the time it takes you to _blink_ and still kicks ass.
He says later it’s 0.1 milliseconds, so that’s _2500 moves_ in the time it takes you to blink. By the way, it would still win against a grandmaster with as much time as they want. Fascinating.
This was one of your best videos, these games are fascinating. I hope you continue to show us some of these computer masterpieces
I like how the profile icon for stockfish went from a simple fish on a chessboard to now just the terminator
Literally binged all of your videos about chess engines yesterday, I'd LOVE to see stockfish 15 play itself with your commentary!
that 4th game is why the french or caro kann work against humans but not engines. no human can create an absolute boa-constrictor of a position where black has 15 possible moves and they all lose. no human can keep up with the 24 remaining pieces and making sure that every single one of them is locked into a position where they cant move without crumbling to ash
This was a fun video. Here's a possible follow up idea: first, pair Stockfish 8 against Stockfish 12, let's say. Once Stockfish 12 acquires a significant advantage in the game (assuming it can defeat Stockfish 8), you then let it swap places with Stockfish 15 and see if 15 can make a comeback against 12.
Or something along those lines...
The three brilliancies in the last one must have been the queening of three pawns.
imagine the sheer savagery of a chess bot that completely destroys its opponent but purposefully draws just to mess with the other chess bots
For context on how damn fast a millisecond is: human reaction time is ~230 milliseconds. For immediate, direct reactions that take 0 thinking or decision making.
These engines would be able to checkmate you in the time it took you to see they had played 1.e4
Stockfish 15 holding off on promotion in game 3 until it walked it's king to knock out the knight is equivalent to a bully holding a kid down and slowly dripping a loogie onto his face
At some point there will be a kid grandmaster that learned chess solely by watching the engines play and I'm scared of such kid
0 brilliant moves? We are not that different, Stockfish 8.
I wonder how insane kids will be in the future. Playing against the strongest opponent to ever live? A player who's also super helpful in explaining your inaccuracies? That has to be a game changer.
so the whole first game was basically levy pressing play and getting a supersonic checkmate served in under a second, still hot to the touch?
That's like turning a light switch to instantly feel bad about your chess
Stockfish(s) literally plays the whole game in less than a tenth of a second. What madlads.
(Each move is 1/10000 of a second)
And after the game, stockfish evaluates stockfish for inaccuracies.
Stockfish playing Stockfish: "What a digital dummy."
"It's like an interesting version of the kings indian defense." Just casual shade being thrown at the kings indian defense.