No one is actually, publically, even trying to solve chess. There is a whole professional industry that does not want chess to be solved. That is why all of these engines start from pre-determined, human positions. If they started from actual scratch, they might play something so extreme as to be forcing, that humans would find absurd. They might (almost definitely) forfeit meta objectives like control of the center. Who knows? The point is - all public engines nowadays are starting from human standpoints. Their progress toward solving the game is ("coincidentally") limited. I expect this insight to be the basis of future videos and topics. I look forward to watching my content be said by more popular voices, to their benefit.
@@michaelfetter5413 isn't Leela reasoning the same as AlphaZero, as in, there's no opening theory input, and the engine arrives at opening theory from scratch? AlphaZero and Leela have yet to provide a completely novel opening, or at least play a lesser used one.
@@michaelfetter5413 they actually do include games with no open book in TCEC. I'm not sure what you are trying to predict as a reason why they always start from human positions. The real reason is purely for entertainment. Non open book games are almost always boring draws. TCEC purposely chooses open books that are imbalanced to see how both engines defend or attack. This has its problems when determining which engine is stronger but it does add entertainment value since nobody wants to watch thousands of draws.
I don’t comment often on anything, but I had to pause the video 1/4 of the way through to be certain to commend you on what has quickly become my favorite video of yours. Your passion and joy of the game is apparent, infectious, and scholarly. Thank you for all you do.
I feel like I benefitted not just from the concrete analysis of this specific game, but also from Jerry’s example of what it means to understand a game more generally. Clearly, it’s about much more than just memorizing all the moves. It’s about asking questions, comparing alternative lines, identifying strategic focal points, and weighing different plans against one another. Thanks, Jerry. Now I feel like I could look at a few games on my own and have at least a rough idea of what it would mean to understand them fruitfully.
Huuge amount of work you've done, Jerry. I do appreciate this kind of videos. Thank you very much indeed to share with us your thoughts and ideas. Looking forward to watching more.
Never change your voice. The calm soothing sounds of your analysis reminds me that not everything is tone dependant. Makes you appreciate the game and life more. thankyou sir!
Thanks Jerry. Very instructive! Qe1 move is super interesting indeed! Interesting how Leela transformed bad knight on the rim into even worse rock on the rim.
Get Jerry thank you for the content. I only realized today something very important about your commentary. I really appreciate your discussion of themes rather than simply showing tactics. It really helps me understand the game in a broader sense. Thank you
Another awesome analysis, thank you for sharing this. This is not the first time Leela manages to isolate Stockfish's pieces and this one in particular by leaving the c5 pawn around is very instructive. Kind of reminds me of different positions where you make use of your opponent's blocked pawns to shield your own pieces.
I love how passionate Jerry is about this game, his excitement at the pawn sac was contagious! It's incredible that he spent 10 hours looking through this one, and it shows; the ideas are beautiful, thorough, and well-presented. Jerry, if you ever get so engrossed in a game again, please share it with us!
I almost never like watching long videos on youtube, but yours feel like 15m or less, even though they are few times longer! One gets immersed into the chessboard and loses the track of the time! Thank you for all these enjoyable and insightful videos! ♥
Jerry I absolutely love your in-depth analysis videos like this one. Your videos alone, above any other content available online have taught me 90% of what I know about opening theory. I can't wait to implement these ideas into my own KID games, especially since I groan every time I see Nf6 g6 after d4 c4. Thanks a lot!
14:45 This position is the closest to "box art" I've seen recently. Queen commanding in the center, all pieces on board, white taking more space, etc. It just struck me as pretty cool as a "chess battle" position.
Hi Jerry! I had this video bookmarked for a few days due to the length and I'm glad that I finally got around to watching it. I always learn something new from your videos. There were several instructive positions in this game but the most interesting to me was 7:42 which helped me further develop my strategic planning skills. Thank you very much! Cheers!
Excellent analysis. I play the KID as black so this was extremely helpful. The in depth variations and well-articulated explanations were fantastic and helpful, because I actually learned things I can use in my own games. Thanks Jerry!
Qe1 was excellent analysis! I can't imagine a world where black gives up the "dragon" bishop for a rook, leaving white with an unopposed dark square bishop and dark square holes all around the king
Thanks Jerry. You study computer matches so we don’t have to! It’s actually your analyses that make these deeply calculated machine games interesting. Did you say some of these sidelines are from your own calculations?
Excellent analysis, but I’d love to see the evaluation graph for this game: at what point did stockfish realise it was doomed? where did Leela spot the opportunity? I think these would be helpful cues.
Thx Jerry! This chess channel is really good! For some reason the algorithm always gives me annoying content like Checkmate with a king and rook for 1 second, Fastest 1 min game, 17 knights against 16 bishops--who would win, the best 30 sec game ever. None of this crap here.
It's been fascinating watching Lc0 vs Stockfish over the years. It seems Stockfish is still stronger through sheer calculation, but there have been a number of games now where Leela has demonstrated a deeper positional understanding that exploits the piece evaluation. Leela seems happy to say "you can keep your rook because is not worth it's full value in this position."
This is hugely understating, if not completely misunderstanding how Stockfish works. No modern engine evaluates positions by assigning "values" to pieces without considering the position like an intermediate player would, otherwise they would not be able to easily beat the best human players. Stockfish crushed Leela in this match, by the way. It was not even close in the end: 19 wins to 7.
Lc0's approach to developing a chess engine also seems to have plateaued. Despite being around for a while now and tons of donated GPU training hours, it hasn't been able to take the crown from Stockfish
I did some deeper (like depth 60 stockfish master) engine analysis on this game, and it's likely that the point the game turns around is 24. ...Bxf2 At that point the eval jumps to +4 pawns, and despite black moving optimally for quite a few moves from that point on there's just no chance of survival. The best move (and the only one not instantly losing) from that position seems to be Bd4, although the position is already very imbalanced at that point, with the eval at roughly +1.6 at depth 60. At that kind of eval it's a coin toss whether with insane thinking time you get a draw or a win for white, although perhaps the win is more likely. Interestingly, the move that was played in the game is tied with Rxc4 for the worst possible move in that position.
I love this style of content. Very calm and collected thorough analysis with all the sidelines. Similar with the Shogi content from Hidetchi, very appreciate it. Also, is there a video on the Nimzo-Larsen opening, starting with the Knight? I mainly play this :)
Great job Jerry. Your trademark passion absolutely shines through, and I LOVE IT! It’s why I subscribed to your channel in the first place all those years ago - your offbeat vocabulary and love of the game. Why else? It’s just a standard chessbase-esque board and a disembodied voice called Jerry without those USPS. Keep doing what ur doing, I’ll keep watching! Huge THUMBS UP.
At 4:45 you talk about the move bd2 creating a double attack against the knight on a5 and the bishop on a1. You may have just not gone into it for sake of their being more interesting moves later that are more deserving of the time in a long video, but I noticed it’s not so simple after black plays c6. I initially thought then white can still play bxa5 and with tempo on the black queen, black must play qxa5 but then white has the nice intermiso nxe7 followed by capturing the bishop on a1. But I then realized whites knight is not escaping from e7 so easily. At this point I did not want to miss anything so I put it into Stockfish. May not be the most powerful version so I may be wrong but it seemed to agree with the line I gave above yet also after Qxa1 black plays Qe5 and after white trades queens it recommends sacrificing a piece on c6 due to the trapped state of the knight which created a minor +2 pawn vs a rook imbalance that favors white but isn’t the easiest win I’ve seen. Just thought it was something interesting and thought I’d share.
4:45 I'm glad Jerry discussed some tactical considerations for whether the Nxd5 discovered attack is good for black or not, but I feel like the concluding remark from the time stamp leaves a lot to be desired. White's two minor pieces for a rook and a pawn would usually be only be slightly advantageous. Furthermore, black may gain another pawn after desperado move Nxc4. In a lot of other positions the material imbalance of R+2P vs. B+N would be equal to slightly favorable for the side with R+2P. It would be nice to know what is so drastically different about this position that the side with two minor pieces is winning.
I analyzed it because I was curious, sorry for being long winded. You’re right - if after Nxc4 white plays Qxa1 it is almost even. If bxc4, Stockfish 14 says white is +1.3 on depth 31, which for these computers is probably winning but isn’t super clear. However, black has Bg5 after Nxc4, threatening the pawn to fork the queen and rook while still attacking the bishop and knight. Best way for black to defend is Bf6, after which white exchanges bishops then takes the knight. This position is ~+2.2, and after playing through some lines here are the main issues with black’s position I noticed: black essentially has to play c6 as the knight on d5 is way too powerful, and exchanging the light square bishop for it will never be good as after cxd5 the c7 pawn will never be able to move and white gets a huge space advantage and initiative and eventually wins all the pawns on the queen side, and also white’s light square bishop will be unopposed and wreaks havoc. So after c6, the d6 pawn becomes weak, and if black tries to defend it (best is to give it up), white can completely paralyze the queenside pawns with a4 a5, leaving black with absolutely no moves to make any progress in the position. Eventually after coordinating the knights, rook, bishop and queen and using whites space advantage, the queenside pawns can be won as black hopelessly shuffles. Additionally, black doesn’t have its dark square bishop protecting, so the whole game it will have to be careful of tactics on the f6 pawn and around its king. If black chooses to give up the d6 pawn instead, it is able to mobilize the queenside pawns, however the combined factors of weak dark squares around the king that constantly have to be watched after and the two knights being able to coordinate better than the rook + one pawn make it winning for white. I will be very surprised if anyone reads all this, but that’s my explanation.
TLDR: blacks two pawns up are on the queenside and white can paralyze the queenside while blacks dark squares are weak around the king. The two pieces coordinate better than the rook and eventually win pawns back
One variation that I am curious about is if the black rook on a5 moves to b5 before white forms the battery and if pawn takes then the b file would be blocked off, and if rook takes then black pawn can take back. Since the rook is pretty much useless anyway may as well sacrifice it.
How far away do people believe chess is from being completely solved? Is it even feasible?
No one is actually, publically, even trying to solve chess. There is a whole professional industry that does not want chess to be solved.
That is why all of these engines start from pre-determined, human positions. If they started from actual scratch, they might play something so extreme as to be forcing, that humans would find absurd. They might (almost definitely) forfeit meta objectives like control of the center.
Who knows? The point is - all public engines nowadays are starting from human standpoints. Their progress toward solving the game is ("coincidentally") limited.
I expect this insight to be the basis of future videos and topics. I look forward to watching my content be said by more popular voices, to their benefit.
@@michaelfetter5413 isn't Leela reasoning the same as AlphaZero, as in, there's no opening theory input, and the engine arrives at opening theory from scratch?
AlphaZero and Leela have yet to provide a completely novel opening, or at least play a lesser used one.
If processing power continues to advance in accordance with Moore's Law, it will be about 200 years before chess is strongly solved.
@@michaelfetter5413 they actually do include games with no open book in TCEC. I'm not sure what you are trying to predict as a reason why they always start from human positions. The real reason is purely for entertainment. Non open book games are almost always boring draws. TCEC purposely chooses open books that are imbalanced to see how both engines defend or attack. This has its problems when determining which engine is stronger but it does add entertainment value since nobody wants to watch thousands of draws.
To be sure the game is solved, all games possible will need to be looked at up until table bases. That will never happen
I don’t comment often on anything, but I had to pause the video 1/4 of the way through to be certain to commend you on what has quickly become my favorite video of yours. Your passion and joy of the game is apparent, infectious, and scholarly. Thank you for all you do.
Thank you very much Joseph. 👍🏼
@Fluffy Maximus Maximus and Network sittin' in a tree
Not bad, but I need to put playback on Giri speed (1.5x)
@@Horzinicla F-I-A-N-C-H-E-T-T-O-I-N-G!
@@ChessNetwork I'm bad at chess but your vids make it easy and fun, where have you been all my life? Take a quick sub.
Thanks Jerry you are so generous with your time and energy. So many people appreciate your work. Have a great holiday season!
I feel like I benefitted not just from the concrete analysis of this specific game, but also from Jerry’s example of what it means to understand a game more generally. Clearly, it’s about much more than just memorizing all the moves. It’s about asking questions, comparing alternative lines, identifying strategic focal points, and weighing different plans against one another. Thanks, Jerry. Now I feel like I could look at a few games on my own and have at least a rough idea of what it would mean to understand them fruitfully.
Huuge amount of work you've done, Jerry. I do appreciate this kind of videos.
Thank you very much indeed to share with us your thoughts and ideas.
Looking forward to watching more.
Never change your voice. The calm soothing sounds of your analysis reminds me that not everything is tone dependant. Makes you appreciate the game and life more. thankyou sir!
Your videos are like curling up with a warm blanket. Thanks for continuing to upload, really excited to watch this game.
Thanks Jerry. Very instructive! Qe1 move is super interesting indeed! Interesting how Leela transformed bad knight on the rim into even worse rock on the rim.
Get Jerry thank you for the content. I only realized today something very important about your commentary. I really appreciate your discussion of themes rather than simply showing tactics. It really helps me understand the game in a broader sense. Thank you
Love these engine game analyses! Such an interesting form of the game
Always a joy and pleasure to see your insights/commentary on these chess engine games
Love your Leela videos; there is so much depth to these games, and your analysis always helps me understand them better!
Another awesome analysis, thank you for sharing this. This is not the first time Leela manages to isolate Stockfish's pieces and this one in particular by leaving the c5 pawn around is very instructive. Kind of reminds me of different positions where you make use of your opponent's blocked pawns to shield your own pieces.
Your videos about Engine chess are my favorite Jerry, keep 'em coming!
Like the way you break it down. Great commentary and analysis
I LIVE for these long videos with deep explanation , perfect thing to watch
That was top 3 of the most instructive chess videos I had ever seen. That's what genius engines combined with top commentary can do.
Yeeeeeeeessss! These engine games are my absolute favorite series of yours. So glad you are still covering them
I love how passionate Jerry is about this game, his excitement at the pawn sac was contagious!
It's incredible that he spent 10 hours looking through this one, and it shows; the ideas are beautiful, thorough, and well-presented. Jerry, if you ever get so engrossed in a game again, please share it with us!
Oh, how often have my anthropomorphized rooks stood around staring at each other, asking themselves: "what now?" 10:29
I almost never like watching long videos on youtube, but yours feel like 15m or less, even though they are few times longer!
One gets immersed into the chessboard and loses the track of the time! Thank you for all these enjoyable and insightful videos! ♥
Jerry I absolutely love your in-depth analysis videos like this one. Your videos alone, above any other content available online have taught me 90% of what I know about opening theory. I can't wait to implement these ideas into my own KID games, especially since I groan every time I see Nf6 g6 after d4 c4. Thanks a lot!
Very good analysis, all my questions on the many variations have been answered in this video, great job!
I can't express how helpful and instructive you are. A real inspiration. Both in chess and general demeanor. You've helped me tremendously.
I believe many would apprechiate more of these kind of videos, which is to say, engine games, from you.
14:45 This position is the closest to "box art" I've seen recently. Queen commanding in the center, all pieces on board, white taking more space, etc. It just struck me as pretty cool as a "chess battle" position.
Incredible analysis. By far the best video on your channel. Keep up the great work Jerry 👍
Extraordinary. Thank you for spending the time on this!
oh my god thanks for all that work, such well done
Hi Jerry! I had this video bookmarked for a few days due to the length and I'm glad that I finally got around to watching it. I always learn something new from your videos. There were several instructive positions in this game but the most interesting to me was 7:42 which helped me further develop my strategic planning skills. Thank you very much! Cheers!
Very instructive game! Love your analysis!
what a beautiful game ...Queen e1 wow ... thanks a lot Jerry ... always been a big fan of your work on AlphaZero and his descendant
Thank you Damon
Thank you for posting this and thank you for the, as always, thoughtful analysis, Jerry. This video with your analysis is gold. Thanks, again.
Excellent analysis. I play the KID as black so this was extremely helpful. The in depth variations and well-articulated explanations were fantastic and helpful, because I actually learned things I can use in my own games. Thanks Jerry!
Qe1 was excellent analysis! I can't imagine a world where black gives up the "dragon" bishop for a rook, leaving white with an unopposed dark square bishop and dark square holes all around the king
So much tactical subtlety in this game 99% of chessplayers simply ain't gonna see it. Goes to show how deep this incredible game is.
Thanks for your analysis Jerry ! Always appreciated !
Excellent analysis Jerry.
Yes, more Jerry covering engine games. This is my favorite kind of video from this channel.
Amazingly explained, bro. You're a Capo!
Wait, 40mins video for a single game?
Great!!!
Epic analysis Jerry. I thought while watching, "this is what it would look like in a magnus v magnus match" lol TY
This was so enjoyable and informative, thanks a lot.
AWESOME 👍👍👍 miss coverage like this. MORE PLEASE
i cant believed I watched all of that XD. 40 minutes of my life gone. Thanks Jerry. You're the best
Awesome analysis, thank you! Definitely saving this game
I love these engine series! By far my favorite Chess video on TH-cam. Keep it up!
What a lucid analysis of such a complex game. Bravo!
A jerry notification, at 4am, with leela chess! Absolutely fantastic! Hype it up yo
A beautiful smooth game, thank you for the excellent analysis.
Thanks Jerry. You study computer matches so we don’t have to! It’s actually your analyses that make these deeply calculated machine games interesting.
Did you say some of these sidelines are from your own calculations?
I used an engine throughout and tested the stuff I was curious about.
This is an awesome analysis
Great video Jerry! Qe1 is a super interesting idea.
Great analysis! Thank you.
This is the 4th time going through this video, and I've learned a new lesson each watch through. What a game, and what a great analysis of it.
These engine games with some master analysis are always so entertaining and inspiring!
Excellent analysis, but I’d love to see the evaluation graph for this game: at what point did stockfish realise it was doomed? where did Leela spot the opportunity? I think these would be helpful cues.
Amazing Analysis by a World Class Mentor, Thanks Jerry
Always love to watch your videos where you deeply analyse games. Lots to learn and understand the "why" of moves.
Thank you!. For your wonderful,entertaining lecture.
You are a saint, Jerry!
Thank you for the lesson Jerry! Best of the season to you and Merry Christmas from me.
👍🏼
Great job with a very interesting game!
Greetings from Warsaw! Love your work.
Excellent analysis, pls keep making these videos
Absolutely fascinating Jerry! Awesome video.
Hi jerry glad to see you doing great
I love your analysis
Please do more chess computers games analysis 🙏
Amazing work Jerry
Joyeuses Fêtes de fin d'année, Jerry. Keep on the good work!
Thx Jerry! This chess channel is really good!
For some reason the algorithm always gives me annoying content like Checkmate with a king and rook for 1 second, Fastest 1 min game, 17 knights against 16 bishops--who would win, the best 30 sec game ever. None of this crap here.
Thank you Peter
wow i didnt realize video is 40 minutes long, very nice analysis
Me: being a 900 pawngrabber
Computer: suggests to capture a pawn 12:40
Me: Good boi!
Unconventional Thinking and Analysis !!! Thanks Jerry
Very very nice commentary, you covered all the questions I had, very easy to follow! thanks :)
It is unbelievable how you figured out the reasoning behind even the most unnatural looking moves. Keep up the great work.
It's been fascinating watching Lc0 vs Stockfish over the years. It seems Stockfish is still stronger through sheer calculation, but there have been a number of games now where Leela has demonstrated a deeper positional understanding that exploits the piece evaluation. Leela seems happy to say "you can keep your rook because is not worth it's full value in this position."
This is hugely understating, if not completely misunderstanding how Stockfish works. No modern engine evaluates positions by assigning "values" to pieces without considering the position like an intermediate player would, otherwise they would not be able to easily beat the best human players. Stockfish crushed Leela in this match, by the way. It was not even close in the end: 19 wins to 7.
Lc0's approach to developing a chess engine also seems to have plateaued. Despite being around for a while now and tons of donated GPU training hours, it hasn't been able to take the crown from Stockfish
Nice breakdown
Very insightful!
I did some deeper (like depth 60 stockfish master) engine analysis on this game, and it's likely that the point the game turns around is 24. ...Bxf2
At that point the eval jumps to +4 pawns, and despite black moving optimally for quite a few moves from that point on there's just no chance of survival. The best move (and the only one not instantly losing) from that position seems to be Bd4, although the position is already very imbalanced at that point, with the eval at roughly +1.6 at depth 60. At that kind of eval it's a coin toss whether with insane thinking time you get a draw or a win for white, although perhaps the win is more likely.
Interestingly, the move that was played in the game is tied with Rxc4 for the worst possible move in that position.
thank you!
I think 22...Qe8 is the losing move here.
Always fascinating seeing how chess can be played beyond the limitations of human thought
Everything above 1700 level is well over my human thoughts
Nice work, Jerry.
I love this style of content. Very calm and collected thorough analysis with all the sidelines. Similar with the Shogi content from Hidetchi, very appreciate it. Also, is there a video on the Nimzo-Larsen opening, starting with the Knight? I mainly play this :)
awesome as always
Thanks for such a great analisis
🎩🎯
@@shantoreywilkins651 hat target?
Great study. Thank you, Jerry!
It didn't feel like 40 minutes just went by.
Great job Jerry. Your trademark passion absolutely shines through, and I LOVE IT! It’s why I subscribed to your channel in the first place all those years ago - your offbeat vocabulary and love of the game. Why else? It’s just a standard chessbase-esque board and a disembodied voice called Jerry without those USPS. Keep doing what ur doing, I’ll keep watching! Huge THUMBS UP.
Thank you Brabham. 👍
Wow xcellent breakdowns good job
Tanks for sharing. Excellent video
That was great, thank you!
Great stuff, thank you!
Awesome video!
I'm too tired to even grasp this. Crazy, crazy game.... I'll probably have to watch again tomorrow.
This channel is like a talk show and chess tut combined.
Very instructive! Well spent time!
Very instructive, thanks Jerry.
At 4:45 you talk about the move bd2 creating a double attack against the knight on a5 and the bishop on a1. You may have just not gone into it for sake of their being more interesting moves later that are more deserving of the time in a long video, but I noticed it’s not so simple after black plays c6. I initially thought then white can still play bxa5 and with tempo on the black queen, black must play qxa5 but then white has the nice intermiso nxe7 followed by capturing the bishop on a1. But I then realized whites knight is not escaping from e7 so easily. At this point I did not want to miss anything so I put it into Stockfish. May not be the most powerful version so I may be wrong but it seemed to agree with the line I gave above yet also after Qxa1 black plays Qe5 and after white trades queens it recommends sacrificing a piece on c6 due to the trapped state of the knight which created a minor +2 pawn vs a rook imbalance that favors white but isn’t the easiest win I’ve seen. Just thought it was something interesting and thought I’d share.
4:45 I'm glad Jerry discussed some tactical considerations for whether the Nxd5 discovered attack is good for black or not, but I feel like the concluding remark from the time stamp leaves a lot to be desired. White's two minor pieces for a rook and a pawn would usually be only be slightly advantageous. Furthermore, black may gain another pawn after desperado move Nxc4. In a lot of other positions the material imbalance of R+2P vs. B+N would be equal to slightly favorable for the side with R+2P. It would be nice to know what is so drastically different about this position that the side with two minor pieces is winning.
I analyzed it because I was curious, sorry for being long winded. You’re right - if after Nxc4 white plays Qxa1 it is almost even. If bxc4, Stockfish 14 says white is +1.3 on depth 31, which for these computers is probably winning but isn’t super clear. However, black has Bg5 after Nxc4, threatening the pawn to fork the queen and rook while still attacking the bishop and knight. Best way for black to defend is Bf6, after which white exchanges bishops then takes the knight. This position is ~+2.2, and after playing through some lines here are the main issues with black’s position I noticed: black essentially has to play c6 as the knight on d5 is way too powerful, and exchanging the light square bishop for it will never be good as after cxd5 the c7 pawn will never be able to move and white gets a huge space advantage and initiative and eventually wins all the pawns on the queen side, and also white’s light square bishop will be unopposed and wreaks havoc. So after c6, the d6 pawn becomes weak, and if black tries to defend it (best is to give it up), white can completely paralyze the queenside pawns with a4 a5, leaving black with absolutely no moves to make any progress in the position. Eventually after coordinating the knights, rook, bishop and queen and using whites space advantage, the queenside pawns can be won as black hopelessly shuffles. Additionally, black doesn’t have its dark square bishop protecting, so the whole game it will have to be careful of tactics on the f6 pawn and around its king. If black chooses to give up the d6 pawn instead, it is able to mobilize the queenside pawns, however the combined factors of weak dark squares around the king that constantly have to be watched after and the two knights being able to coordinate better than the rook + one pawn make it winning for white.
I will be very surprised if anyone reads all this, but that’s my explanation.
TLDR: blacks two pawns up are on the queenside and white can paralyze the queenside while blacks dark squares are weak around the king. The two pieces coordinate better than the rook and eventually win pawns back
One variation that I am curious about is if the black rook on a5 moves to b5 before white forms the battery and if pawn takes then the b file would be blocked off, and if rook takes then black pawn can take back. Since the rook is pretty much useless anyway may as well sacrifice it.
Thank you Jerry!