I Tricked Stockfish From This Position 😏

[@ChessVibesOfficial]

Happy New Year guys! The first 10 people to use code 2023 will get 30% off my Mega Bundle course pack (5 courses): 👉🏽chessvibescourses.thinkific.com/bundles/mega-bundle

[@ajb129]

I like this format a lot for "busting stockfish" type puzzles. For regular puzzles where stockfish is in agreement with the solution, I think the old format is still good

[@rycegabrielmajor9847]

one of the few occasions where human ingenuity beats "perfect games"

[@Samarth916]

but its only relevant if a human is exposed to the puzzle right when stockfish is. What's happening here is, the human knows the solution beforehand. Its simply not an equal match. And I don't think stockfish will come in real world to oppose this.

[@mmfreshmeat2860]

@@Samarth916 stockfish can compute moves way faster and better than humans, yet it thought it was a stalemate when it was possible to get a checkmate

[@Samarth916]

@@mmfreshmeat2860 My point is, can a human being see this puzzle for first time, and get the best conclusion?

[@mmfreshmeat2860]

@@Samarth916 my point is that even with stockfish having this much power, SOMEONE still managed to see a checkmate, unlike stockfish. Doesnt matter who, a human managed to beat stockfish

[@linuslucke3838]

@@Samarth916 Yes, you don't have to calculate all of the moves like Stockfish, you just have to see the idea, which Stockfish can't because he's just bruteforcing every move possible, but that takes too long to find the solution in the puzzle

[@Pedoraruba304]

Yes this format is definitely better as it compares human moves to stock fish moves

[@lxrose]

That would take so many moves for Stockfish to calculate the winning line. That is where human thinking beats a computer algorytm, because we think in patterns no matter how many moves the end result takes.

[@christopherheckman7957]

That's also the key to solving Blathy's monsters (mate in N, where N is in the hundreds); he has a lot of these where Black will cream White if White gives Black some breathing room. But there's a sequence that allows White to make one move without releasing Black. Then the sequence repeats, White gets another move, etc.

[@Trelmayas21]

Personally, I loved the new format. Seeing you work through the puzzle rather than just present the solution was entertaining. Also, it's always nice to see the oddball position that Stockfish can't correctly evaluate. Happy New Year, I hope the flood repair is going well, and thanks for the video

[@menoob4603]

Even Hans Niemann cannot solve 💀

[@davidking4838]

Yeah, for some reason he thought it was a draw....😁

[@namehandle198]

💀

[@stellarsub-orbital9922]

He can solve it with some smooth vibrations

[@caudravan]

💀

[@Hauserstjepan557]

Vibration Hans = stockfish 15 😅

[@amaarquadri]

I think trying to win against stockfish is a good video approach for positions like this where there's no single forcing line and you are sort of demonstrating a technique without showing every line.

[@unclegreyball]

Correct me if I got it wrong but I believe Nelson's knight check is a blunder. Unlike other corners, once the black king was allowed to move into the a1-c2 rectangle, there is no way for white to chase it out because the very important d1 square is covered by the e2 pawn. The white queen would not be able to check/do a waste move on d1. You can try to chase the black king out with the following FEN, I can't: 8/8/8/4p3/4p3/1Q2pppp/4prpr/k3Kbnq w - - 0 1 Going back to the very beginning of the puzzle, I was wondering what is the purpose of setting a white knight in b5. I believe it is forcing the black king to lose a tempo. At 0:48 I believe it should be like: 1. d7 Kxb5 2. d8=Q Kb4 3. Qd3 e4 4. Qc2 e3 5. Qb2+ and the no matter which way the king moves, the white queen just staying a knight-move distance with the king to prevent it coming down again. And then I suppose the number of pawns in the e file is carefully set as well, if there was one less pawn, I believe it would not be enough for white to chase the black king into the g file "well". That part I didn't careful think it through, but just believe a chess composer would not set useless pieces on the board.

[@unclegreyball]

NVM from 6:40 suppose all pawns are advanced, then it's 1. Qa3+ Kb1 2. Qc3 Ka2 3. Qa3 Kb3 4. Qb1+ Ka3 5. Qc2 So it takes some time to force the black king out of a corner but seems can still finish off black within 50 moves

[@vivaankohli9448]

I really enjoyed this format a lot because I like how you work through the solution to the puzzle. I like how this also shows how a computer can be wrong sometimes in evaluating the solution. Happy New Year!!

[@jimvinci8295]

I enjoyed this very much and believe you can really learn much about mating techniques from studying these types of puzzles. I used the Komodo Dragon engine, and it too indicated an initial draw. It was entertaining to play this one out on my own. One thing that is important to note is that you do have to make forcing moves or this will be a draw by the 50 move rule. There is a definite method needed to force the king to the other side of the board where it can be mated. Thanks for finding these interesting positions!

[@stanleydodds9]

The thing is, stockfish can't solve puzzles like this (quickly) because it's simply not designed to be optimal for this. In order to be good at chess that happens in real games, it doesn't waste time by being good at solving puzzles. The simple way to put it is that each puzzle of this type is essentially a special endgame which requires its own endgame table; there is a certain relatively small set of states that if you stay within (forced by one player) then you can find a forced checkmate at a relatively high depth, where a normal game tree search isn't optimal. For instance, in this puzzle, if you tell white to never move the king and to never capture any piece other than the pawn that ends up on g3, then a computer can easily and automatically generate every possible state, verifying that black cannot escape this state space, and make a small endgame table that finds the forced checkmate for white, in the same way that normal endgame tables are made and used for real endgames with a small number of pieces (and therefore you don't need to follow any special rules to stay within the "small" search space). In principal it wouldn't be too difficult to automate this type of puzzle solving process; look for "rules" that one player can follow which massively restrict the space of board states, and then for each of these, make the endgame table, and see if the current state has a forced checkmate, forced stalemate, or nothing. The problem is that this isn't practically useful, because this will almost never successfully find anything in a real game, so almost always just wastes time when you could be doing a more shallow but wider search, looking for "good" positions at a lower depth.

[@trdi]

I had this exact position last week. Wish I'd seen this before.

[@wiscorpio72]

I've tried this puzzle over and over again playing myself and I actually found, this is black to play and draw to prevent white to play and win. I wish I could talk to this composer about this but the composer may be passed on. We were both able to beat Stockfish because of the horizon effect. Maximum just responds right away. I get the idea of this puzzle and I learned how you can corner a king with just a queen, but all the black king has to do is run to the bottom corner. The queen doesn't have diagonal leeway on the left of the a file or under the 1st rank to force the king anywhere with only 3 files, and the queen can't force the king up even of the queen is on the 1st rank, the queen can't step on the d file because it can be captured unless the square doesn't have a threat, the queen can't force the king up from the 3rd rank and black can even draw by 3 fold repetiton. The pawns can move and cause zugzwang to white. When the pawn wall isn't solid, the king can escape through it, when the wall is solid, the queen still can't force the king up. Do try playing against yourself, you'll see black can draw. You can set up this position with the king on other squares, the white bishop and knight already gone, say if you set this up with the black king on or close to a8, then you can play this idea. Stockfish will still think it's a draw but it will get smarter as the king is diagonally forced to the pawn wall. The queen can get the king to a8, go back and forth between b6 and c7 and force the pawns until they can't move, then the queen has diagonal leeway down the board to force the king against the solid pawn wall, and to the h file, capture that pawn, diagonally force the king in that hole, make sure the pawn wall is tall enough, you can do it with 1 pawn on the wall captured, but you would have to ride the king down the diagonal below it. So you can see, I studied the hell out of this puzzle for weeks.

[@BobChess]

"Stockfish is the best computer" Human who can trick the Stockfish: 😂

[@bobbymcjoey9432]

But- it is the best chess computer?

[@idkman4748]

@@bobbymcjoey9432 so far

[@wariolandgoldpiramid]

I actually really loved the new method of playing against Stockfish. Please do this more often.

[@GeoStreber]

Isn't it much faster to use the queen to kill the pawn that ends up on G3 first before giving it the chance to plug that position up?

[@pimhuisman646]

And even the E pawn is not needed for the mate I think. In the before last move E5 is covered by the Queen anyway

[@choco1101]

Maybe it would be less dramatic that way because stockfish would “understand” the position earlier

[@sollevi9846]

50 move rule ?

[@AbouTaim-Lille]

That looks like a similar situation in Omega Chess where the queen is the only piece that can force checkmate without any support.

[@MrCegeb]

Annotation of King Move is probably ????? (Massive Blunder). Just look at the evaluation score after King Move, it's like over 30.

[@Austin101123]

I think when stockfish didnt take the knight at the start then your knight can escape and its an easy win.

[@AyaanMansuri-gn2bg]

Yeah

[@wiscorpio72]

One of the times I played this against Stockfish, it made a big time blunder. After 1. d6+, Stockfish replied with Kxb5, I responded with 2. Bxd7+ so black can't take my pawn, then I can play d6xe7 and I saved both my bishop and promoting queen, and it is an easy win for white. Really other than Stockfish giving it to me, I have been practicing this puzzle. 2. Na3+ is necessary for decoying the king to trap it against only 2 files, if black lets the knight escape, it's an easier win. If black really only moves the pawns on zugzwang, a cornered king, it is the best defense, maneuvering the queen just right to corner black against the a file, then the 8th rank, then the h file is very hard to learn, it takes practice, and it is possible, the queen can't always force the king against the a file and black can draw by 3 fold repetition. This is one puzzle, it will take me a while to get it down pact.

[@BigDBrian]

why not take the pawn first and then hunt the king?

[@mbmillermo]

After seeing it all -- it looks like you could speed it up a bit by capturing the g pawn almost immediately after getting the queen.

[@ДенисИванов-э9у]

It might have some connection with 50 move rule, not sure how long would the whole king chase take from 0 to mate, might be > 50 moves

[@andy02q]

@@ДенисИванов-э9у it's not that. He takes 44 moves total including all of the pawn moves at the start. Edit: No, more than 44 because you don't see the actual move number, but like 6 moves above, but it should still be well below 50 with pawn moves resetting the timer.

[@johnb6723]

Wow, a smothered mate with the queen.

[@peterwolf4230]

Surely when you check the king in the top left and top right corner, each time it could stay on the side/top rather than continuing round clockwise?

[@thezetes]

Nope, it cannot because of the horsey placement of the queen!

[@jimmymcfoss2059]

Qc6+ and Qf6+ don't seem accurate, since you can go back to the a file/8th rank. You have to check on the a-file, then go way back (i.e. 1st rank) and then go back on the a-file and restrict the king a knight move away. i.e. 7:51 Qa6+ Kb8 Qa1 Kb7 Qa5 Kb8 Qa6 just to show the pattern

[@Malefic_Roar]

@@thezetes It can bro. The queen CANT cover knight move.

[@thezetes]

@@Malefic_Roar I am not talking about that. Horsey placement is a technique to push the king in a smaller and smaller box with the queen, and you can push a king from the back rank, making sure it cannot return to the back rank. @Chess Vibes, can you join in?

[@Malefic_Roar]

@@thezetes And also im not talking about that. What the comment said is that the king can go back when the king is at a8 and the queen check in c6.

[@ciantheoneandonly7857]

You can checkmate with the queen and Knight, I'm confused why stockist didn't take it and why you didn't move it

[@demboyz808]

Yea there must be a reason can someone enlighten?

[@ДенисИванов-э9у]

It just lags with it, if you plug the wrong moves and return it then finds it must take the horse

[@maxlepocher2627]

At 8:05 Q is on a5 , K is on b7 , black should play K c6 and I do not see how white makes progress.

[@lockaltube]

Yes, I tried to play with myself and indeed it looks like draw due to the fifty-move rule or repetition...

[@NRC_Flare]

I really liked this video, but I don't understand why you didn't save the knight.

[@PhilipPlaysGames]

Stockfish doesn't compare with a human brain💀

[@hudumannaia]

Wait, what are u doing if king dont go to the right (in left top corner)?

[@tykemorris]

Stockfish has proven humans wrong many times. I love it when our species returns the favor.

[@Grason20]

I remember winning against stockfish with stockfish having 32 same coloured bishops and I have only 1 opposite coloured bishop (stockfish thinks it's a draw)

[@justfrankjustdank2538]

this makes sense imo since stockfish looking into the future takes into account material differences and this is such a weird foreign position to it i can see why it was fooled

[@edsimnett]

More efficient to take the G pawn before stating to corral the king?

[@scaramocha]

hey this was insane, I'd have never figured this out on my own, thanks for the insight and keep up the good work!

[@squealer42]

Good luck convincing the arbiter it's neither draw nor helper-mate when black's clock flags!

[@nickdsp8089]

We need a proof that the king is forced to leave the a8 corner by force and not by mistake. After Qc6+ the king can go back to a7 for example.

[@firiasu]

Feel free to experiment with your format

[@iknowyourerightbut2448]

Loved it. Happy New Year!

[@tobiasreineke761]

Correct me if I'm wrong, 0:54 Nb5, kb3, Nc7 is winning faster. Can someone explain?

[@PanterD2S]

Exactly what I thought! 3...e4 seems to be weak try from stockfish not taking the knight.

[@Grivian]

It just shows that the future of chess engines is the combination of stockfish and AI. Here it doesn't matter how many different moves you can predict and their outcomes. The human mind sees the specific moves 10, 20, 30 moves ahead as irrelevant. What the creative human mind can find is that the queen can force the king anywhere it wants. It understand a scenario where the king can be mated with just a queen, how to trap a queen with pawns. With that information alone, knowing where we start, the goal and the concept of the actions to take each move can be deduced on the fly. In fact anyone can achieve this result after you explain the goal and the general concept of how to achieve this. Super easy for a human but not possible for stockfish because it is analyzing a trillion irrelevant moves.

[@awang_ir]

Happy new year, everyone Edited: congratulations Nelson for having more than 2M views at top 35 chess principles video

[@atharvajadhav9471]

U membershi_ff

[@RikMaxSpeed]

I wonder what it would take to help StockFish understand that the position is not a draw, one side has mobility the other hasn’t, and then there’s a mate in 25+

[@dentonyoung4314]

Kind of reminds me of that puzzle by Oscar Blathy where White wins with just a knight.

[@dantemalick2872]

When he didn't take the knight couldn't you have saved it and won alot faster

[@RedToilets]

Then his pawn would’ve been captured and he wouldn’t be able to make a queen.

[@dr_gamess]

Not too hard, Experienced players would know the general idea! Anyways good video and explanation.

[@Player-eu3ru]

I feel like you should put the king near enough to your king to deliver the checkmate

[@ralkadde]

Happy New Year, Nelson! Hopefully you have overcome the damage after the flood. It is very much appreciated that you are making chess videos for us on the side.

[@benngo7408]

I once made a position where i filled up the board with bishops on the same color and stockfish said it was a win.

[@chessartist]

Fascinating puzzle!

[@johnkesich8696]

Isn't Na3+ a wasted move? The knight is doomed, what does moving it accomplish? Why d7 rather than dxe7? Leaving the e pawn gives black extra stalling moves? And, why not pick off the other pawns before going after the king? It would take 4 moves, which you spent pointlessly shuffling your queen around, and would again eliminate stalling moves. --- I prefer straight up analysis. What's the point of watching someone making mistakes against stockfish? While some people seem to think you "tricked" stockfish, the position is a win for white, stockfish was simply not clever enough to see that white must create a cul-de-sac to force black into.

[@adrianancheta-bw5fm]

what if he dont move to b8 hat if he stays at a file

[@jasmint6703]

Chess that is nearly not chess! Outrageous! Inconceivable! Confuse stockfish whenever we can.

[@a.jz19]

What if Ka7 after you check him at the top left corner

[@michaelb2938]

Could this be drawn by the 50 move rule if they move the pawns first?

[@pheon_ix4267]

can someone tell me why king a7 doesn't work when the king is in the top left (after being checked)

[@johnkesich8696]

with the queen on b5 and king on a8 1. Qc6+? Kb8? at which point white should play Qb5 better was 1. Qa4+ Kb7 2. Qa5 Kb8 and white can now drive the king rightward 3. Qa6 Kc7 4. Qb5 Kd6 etc 4. ... Kc8 5. Qb6 etc 4. ... Kd8 5. Qc6 etc

[@andy02q]

Are there other engine which are worse than Stockfish generally, but better in these kinds of positions? Like Lc0 for example.

[@zobayerzahid3368]

format is great

[@Iblis2Lakon]

After Stockfish refused to take your knight on a3 the easiest win would be 4. Nb1 Kb2 5. Nd2 or 4.. e3 5. d8Q Kb2 6. Qb6 1-0.

[@denisrho1019]

Happy New Year 2023! I like time to time to see such bizarre situation.

[@danlee9049]

Amazing puzzle!

[@SmerteLunefull]

Why sacrifice the knight at the beginning please?

[@ranchoabilities7928]

Happy New Year Nelson! Anyway, I just came across this interesting puzzle rated 2458, and it didn't really seem like it was that high rated. I am still a beginner (around 800 rated) and solved this puzzle easily. It comes out of this FEN: (8/7p/1p1p1k2/3Pp1N1/1P2Ppp1/7P/4QPPK/2q5 b - - 0 1) with Black to Move. Why not test the viewers on it?

[@roderickmain9697]

I guess you could have eliminated some of the pawns to speed it up. Particularly the one that you need to take anyway. Certainly an interesting technique to force the king into a hole.

[@ДенисИванов-э9у]

You can kill a G pawn, other 2 are required for checkmate

[@alexrobinet7576]

Moar puzzles plz :)

[@hubertznajomski3546]

Amazing position. I enjoyed it ❤

[@shrutiranjansahoo652]

can you make a video on playing against stockfish with stockfish down a queen😁

[@odjidja-ofoe7880]

7:57 why doesnt the king just run back to a7 ?

[@johnkesich8696]

Qc6+ was a mistake (as is the Kb8 response) 1. ... Ka7 2. Qb5 Kh8 3. Qa4+ Kb7 2. Qa5 Kb8 and now white drives black rightward although temporarily ceding some space downward.

[@odjidja-ofoe7880]

@@johnkesich8696 thank you

[@Gyro_369]

Actually if you analize it with like dept 25-30 it says that white is winning

[@tottenvillelegend826]

I love these positions ❤

[@taneweliat1608]

I enjoy watching both formats...cool

[@stevejackson2471]

@Chess Vibes I was just wondering, do you know who composed this puzzle?

[@ChessVibesOfficial]

Puzzle details in the description!

[@stevejackson2471]

@@ChessVibesOfficial Thanks I didn't see that!

[@zhongli8819]

I made Stockfish play as white and I still managed to win lol, stockfish couldn't find the mate. few times it was draw but I won once

[@MrTreadfire]

I like this format.

[@ArisKatsaris]

Why in the world not take the black pawn from the very beginning instead of first moving the king over there and then juggling both concerns?

[@RedToilets]

The Bishop is checking the King.

[@gourdy6297]

can you do that with a knight

[@yrsgaming697]

it is because of the depth . if you make depth 100 it will say mate and win for white. depth means the number of moves stockfish will analyse . so you didn't trick stockfish

[@lockaltube]

I won't be so sure about 100. "Depth" in stockfish does not mean that "all N moves were analyzed" - it is impossible for any chess engine. It means that engine analyzed up to N in depth nodes in the tree, trimmed with alpha-beta pruning algorithm (combining with other algorithms in Stockfish, like NNUE). That's why Stockfish can sometimes easily find a mate in 40 (on depth 10) or can't see a mate at all.

[@Malefic_Roar]

@@lockaltube Do you even use stokfish? Stokfish depth 10 can't even find mate in 30.

[@lockaltube]

@@Malefic_Roar yeah... I exaggerated, probably. But it is very common for Stockfish to find a mate in 40 on depth like 25, and then on depth 27 to find a faster mate in 18.

[@sonoan2973]

brilliant puzzle

[@vapormermaid]

I'm curious if AlphaZero could complete this puzzle properly against Stockfish.

[@alcoholrelated4529]

thats what i was wondering about too

[@wiscorpio72]

Black didn't first take the knight, that's an easier win because the knight can escape.

[@Cygis]

Out of curiosity, why didn’t you use the Knight to help?

[@johnkesich8696]

There is nothing the knight can do. If white tries to save the knight, for example 1. d6+ dxe6 2. Na3+ Kb3 3. Nb5 exd6 there will be no queen or win Na3+ is actually a wasted move which accomplishes nothing. 2. dxe7 queens a move sooner and avoids a couple delaying pawn moves. 3. d7 was also a mistake, de7 is better and white should have hunted down the mobile pawns to avoid some of their delaying moves.

[@ItIsYouAreNotYour]

In a tournament, what is a stalemate considered? Draw?

[@Adventurer-te8fl]

Yes

[@ItIsYouAreNotYour]

@@Adventurer-te8fl thx

[@jenm1]

When people say chess is solved:

[@tjfjt]

Hey so I'm looking at this and I'm just curious; could you take the back g pawn before it's pushed and before forcing the king over to the alcove? Like is there a reason that wouldn't work?

[@ДенисИванов-э9у]

Yea i checked it and it still works

[@AyaanMansuri-gn2bg]

Why in the world did he not save his knight when he had the chance? So he would have a queen and knight vs the king.

[@ConnorMooneyhan1]

Beautiful!

[@flyingmonkey3822]

Why not gobble up those pawns on E file to free the king to support?

[@takasaga2395]

Defos a lovely vid

[@meisstupid1831]

why that emoji

[@iplaygamesforfun3371]

How does that even happen?

[@1122445]

God bless all good things for 2023

[@Relkond]

So… stockfish can look ahead some number of moves, but is incapable of having ideas and proving them out?

[@sol_mental]

Both stuff are cool ig.

[@kit.21]

there was a free rook at beginning

[@saurabhbunty123]

fantastic!

[@Bubanicyeet]

Why can't you just take the e3 pawn and get your king out

[@victorfinberg8595]

nice. i got it all wrong

[@kienvanvunguyen4468]

Bruh I new to chest and I found a faster way to win

[@PotatoChess-we7rj]

honestly I hate it when you try to solve it your self I like it when you say white now plays the mind blowing move! ###

[@PotatoChess-we7rj]

@chessvibes1. no stop lie

[@rikkousa]

Very cool