17:51 I couldn't understand what Levy thought was so beautiful about the e6 queen sac. I wasn't even sure white is winning so I checked with stockfish. After Be6+, Rf7 it's a crazy position. At depth 29 the best move is +4 and the next best is -7.3, over 11 points hinging on an only move. White has to take with the rook, let black take the knight, and force a trade that leaves R vs B with all the pawns. Be6+, Rf7, Rxf7, Qxe4, Rf6+, Qxe6, Rxe6... +4 The next best move is Be6+, Rf7, Ng5... -7.3 Be6+, Rf7, Bxf7+, Kh8... -7.6 (depth 25)
The tactical sequence in the 2nd game deserves its own video, I've watched every GTE and it's by far the best tactic I've eve lr seen. Sacrificing a rook and a queen just to win control of a SQUARE for a bishop is utterly insane and is completely winning even after the rook block. For a 900 to see that...wow.
I haven't watched yet, but one way to improve gpt4's performance on this task is to give it a few examples beforehand. In other words, show it a few games and the elo of the players afterwards before showing it the games you are interested in. This is known as few-shot promoting and is used to improve performance on evaluation benchmarks. You generally only need like 5 examples or 5 chess games with elo in your case.
some of the games being 10 minutes versus 30 minutes really explains the differences here. a 1600 player missing mate in 1 makes sense when you consider it was a 10 minute game. and a 900 spotting a brilliant move makes sense in a 30 minute game
Thanks for the videos. You should do one of playing against video game chess games of the 90s and seeing how they played. On a side note, I remember a friend played against the computer in the 80s, but he didn't like playing on the highest difficulty, because the computer would take forever to make a move.
22:51 i think white had the same idea as me when I saw the position, Kf2 (black makes a move) hxg4 qxh1 bb5+ C6 qxh1, but that "black makes a move" is something that is keeping me at 800 elo lmao
Levy to ChatGPT: " - You broke my record! Now I break you as I broke your friend-bot (pointing at his knee where there´s a chip that once met a baseball-bat)
Writing a sequence model to predict Elo of every single move (as a regression task)... Was one of my ideas for the Deep Learning class two year ago. Language models are fragile. They don't do good with numbers, they predict tokens, they don't even know what digits are. It's all learned in a few specific heads.
Levy! minoxidil 5% and finasteride together will work wonders for your hair. Got my hair back with it. You helped me get to 1400 elo, hopefully this’ll help get your hair back legend. Cheers mate
Today's episode is sponsored by me. Enjoy the 40% off holiday courses sale: www.chessly.com/
first 🤓
second 🎉
1 min ago
Man I love your 'Middlegames Masterclass'. I can honestly say, I was once 1500 elo, but now, here I am at 900 elo.
Amusing games as always Levy!! Thanks for sharing!!!
GTE never fails to include chess in its levy.
these comments are getting absurd
I mean 9yos these days are getting creative bro this is what our great grandparents fought for@@pafspaw
I hate that I never fail to laugh at these variations
Don’t forget about the rook boarding the magnusing magnus
absurd never comments to post fail people
That last game though at 27:47. That look of utter disbelief and a solid 6 seconds of silence followed by a "wha...?!", had me laughing :D
I had the same reaction, my jaw dropped. There is no way these two are 1700
Only comment that makes sense
I was also thinking 1k rating. Musta been drunk chess at the bar or at the library.
Same, had me burst out laughing by Levi's reaction haha
He was so disappointed 😂
"Levy always fails" to prove that guy wrong
L0l
Lmao
You lost 😊
But he won 😂
Well if you’re second below him, then you can claim victory on a technicality
lmfao
Those 900's in 2nd game played way better than 1700's in last game.
Today's stare was kinda short for you Levy. Meeting Magnus really changed you.
"Today's _____ was kinda short for you Levy. Meeting Magnus really changed you."
@@zacharymcleod460💀
💀
Levy's guess the elo videos stay timeless classics.
w username
Thanks UwUTheCommenter.
"You're telling me that a 1660 hung mate in one?!"
Cheers to a lifelong friendship between Gotham and Chat GPT-4
27:45 The look of disappointment on Levy's face when it's actually revelead as a 1600 game lol
You're telling me that a 1600 rated player-
Chat GPT never fails to guess the elo of Magnus putting Levy in this chess videos
It’s so heartwarming that ChatGPT honors someone so small like Levy and allowed him to play guess the elo on GothamChess’s RUclips channel 🥰
Cringe
@@OkartaSamuel what are you yapping about gangy
@@imanerdbroyou are not kai
@@redacted6092 kai cenat ohio rizz duke dennis sigma 🤣🤣🤣🤣🤣
@@imanerdbro Hello
It almost feels like Chat GPT was trained using Levy’s guess the elo series…
Oh wait
😅😂😂
levy never fails to be in a gothamchess video
wait a minute...
I wonder why
Woah
17:51 I couldn't understand what Levy thought was so beautiful about the e6 queen sac. I wasn't even sure white is winning so I checked with stockfish. After Be6+, Rf7 it's a crazy position. At depth 29 the best move is +4 and the next best is -7.3, over 11 points hinging on an only move. White has to take with the rook, let black take the knight, and force a trade that leaves R vs B with all the pawns.
Be6+, Rf7, Rxf7, Qxe4, Rf6+, Qxe6, Rxe6... +4
The next best move is Be6+, Rf7, Ng5... -7.3
Be6+, Rf7, Bxf7+, Kh8... -7.6 (depth 25)
rook vs bishop endgame is winning anyways
not to mention that was my game, although i am 1100 rated currently
The tactical sequence in the 2nd game deserves its own video, I've watched every GTE and it's by far the best tactic I've eve lr seen. Sacrificing a rook and a queen just to win control of a SQUARE for a bishop is utterly insane and is completely winning even after the rook block. For a 900 to see that...wow.
I agree that was a crazy find by me
I haven't watched yet, but one way to improve gpt4's performance on this task is to give it a few examples beforehand. In other words, show it a few games and the elo of the players afterwards before showing it the games you are interested in. This is known as few-shot promoting and is used to improve performance on evaluation benchmarks. You generally only need like 5 examples or 5 chess games with elo in your case.
Chat gtp was trained on Gotham's GTE videos....
This was hilarious. Some of the best Guess the elo games I've ever seen. ChatGPT wrinkle didn't add too much but certainly made me click on the video
I’ve been saying for a while now 800-900 is more of a battleground nowadays than 1100-1300
21:21 Levy pretending to be confused about f4 even tho it's literally what he recommends in his course lol 😂
27:48 1660 meme never gets old😂
Levy never disappoints with guess the elo
Can we all teach ChatGPT to shout "the rook" when it gets sacrificed?
22:00 I usually play this bongcloud vienna. Steinitz has played this line once. Also Hikaru played similar line.
r.i.p levy's keyboard
I heard the keys wobble
THE ROOOOOOOOOOOOK
*Levy realizes that the rook was sacrificed on the next move*
i see chatgpt i click
i see guess the elo I click
i see both, i quickly click
The fact that the players from game one are higher rated than game two is wild
Putting Chatgpt in GTE is like trying to rate a bannana in terms of how well you can drive it
god damn that 900 in the second game would beat my ass
16:22 Bro just hit me with courage the cowardly dog nostalgia. 😊
This is probably the best combination ever 😂
I love how Levy trash talk the bots estimates (when they are different from his), but when they agree he goes all excited and happy 😂
how to beat mittens: set timer 1 min and premove and sacrifice and try to dont get checkmated
The amount of happiness that I felt when I looked at this title...immense, immeasurable
This man is 1660 and he hung a pawn in move 9 161660 hung pawn move 9.
Only just started the vid. Already love that guess the ELO is here.
Levy never fails to make me watch chess videos
Levy never fails to put a AI in a chess video
Just type "Levy never fails" and get the top comment
2nd :/
Top comments almost always can be written without even watching the vid
Levy never fails
Levy must pin this instead then i guess
Levy never fails
Levy screaming THE ROOOOOOOOOOOOOOOOOOOOK is now canon and unreplacable
The best and most iconic receding hair line
Levy's look of horror at 17:48 is priceless.
Do you mean 27:48?
27:57 YOU'RE TELLING ME A 1660 DIDN'T SEE Ra4???
He obviously didn't like it
Chess never fails to include GTE in his Levy.
Ive never seen Levy this nice in Guess The Elo
Gotham never fails to include Levy in his videos
ikr, truly the most of time
I thought this was ChatGPT playing chess, please rename the title to "ChatGPT is INSANE *at guessing the elo of people who play* Chess"
ChatGPT has finally started to take over
16:00 thanks for reviewing my game levy!
Chat gpt on that zaza on this one
this is not the GTE I signed up for, feels empty without the chat and your interaction with games/players.
No shot game 2 was played by legit 900 players.
Levy's never fail to get his top comment start with "Levy's never fail".
Gotham doesn't know how good a 900 can be if they lock in 🗿
Bro I swear these are the 800-900’s I play everytime
some of the games being 10 minutes versus 30 minutes really explains the differences here.
a 1600 player missing mate in 1 makes sense when you consider it was a 10 minute game. and a 900 spotting a brilliant move makes sense in a 30 minute game
Levy never fails to get a Levy never fails in the top comment
Levy never fails to fortnite battle pass level 10 latina gyatt.
thank you chat gpt for the sponcer
I'm 1423 and my top was 1550. And I really don't remember that I saw such sacrifices even on my level before. Very suspicios game of white.
Gte never fails to be a part of magnus rozman chess
A chatGPT chess video without someone castling capturing his own bishop... 😢
ChatGPT never fails to guess the elo of receding hairline guy
Levy never fails to say shut up to a computer
11:30 Gotham learned Dutch defence using chat gpt
Levy is chatgpt now wonder there guesses were aligned.
Great to see AI like ChatGPT to request there games to Levy to be on GTE
That rook and queen sac is after they turned on an engine to clinch the game. Pretty obv
what is your elo ?
Levy never fails to beat AI children.
Thanks for the videos. You should do one of playing against video game chess games of the 90s and seeing how they played. On a side note, I remember a friend played against the computer in the 80s, but he didn't like playing on the highest difficulty, because the computer would take forever to make a move.
22:51 i think white had the same idea as me when I saw the position, Kf2 (black makes a move) hxg4 qxh1 bb5+ C6 qxh1, but that "black makes a move" is something that is keeping me at 800 elo lmao
I would love to see you playing against chat gpt on an infamous 8D chess match
Magnus never fails to sacrifice his Levy in his videos
Shuutttt uuuuppp
@@JUVARBS I know I should but only if you stop liking ur own comment
I actually didn't I will like it now to prove that it was not me so you will see 2 likes not 1
We want Levy's 2023 Year in Chess♟.
Levy to ChatGPT: " - You broke my record! Now I break you as I broke your friend-bot (pointing at his knee where there´s a chip that once met a baseball-bat)
Petition for Levy to bring back the Gotham City Podcast
The last game definitely felt like the players were both move memorizers and didn't know what to do once things went off script
Of course, ChatGPT destroying "Levi" as always.
ChatGPT never fails to steal jobs.
Chat GPT never fails to include Levy in its videos.
16:22
Neuron activation!
"Chat GPT 4 can't scream 'the rook'.
Chat GPT 5: "...and he sacrifices... THE ROOOK"
An idea for a simmilar video: levy plays chat gpt for 2 games, guesses the elo of chat gpt, and then looks at what stockfish estimates chat gpt's elo
GPT never fails to Levy GTE Chat
Just binge watched Levy’s chatgpt videos. Love it
Writing a sequence model to predict Elo of every single move (as a regression task)... Was one of my ideas for the Deep Learning class two year ago.
Language models are fragile. They don't do good with numbers, they predict tokens, they don't even know what digits are. It's all learned in a few specific heads.
Levy! minoxidil 5% and finasteride together will work wonders for your hair. Got my hair back with it. You helped me get to 1400 elo, hopefully this’ll help get your hair back legend. Cheers mate
the rooook never fails to scream the LEVYYYYY
Levy never fails to recede his hairline.
how tf did you not call this video “GTE vs. GPT”
I think this my first time seeing levy in a short sleeve
11:47 there was opertuniti for the *in back round war image* ICBM
Chat GPT has nothing on Levy's receding hairline
Hey bruda any shot you can do a “best games of 2023” vid. That’d be an awesome way to end the year 💯
Ah yes, the Gotham Personal Trainer chatbot, known as chatGPT for short, truly one of the chess engines of our time.
Bruh literally I beat chess gpt with scholars mate 💀
This entire video was just opposite day for the players and guessed elo
Levy should run one of martins games and a stockfish game through chatGPT.
Another GTE, another time i escaped ending up in the episode and avoiding embarrassment. Lol