Outsmarting Stockfish

Jollification
 
This game is a beautiful example of how the engine's recommended move may not be the best move.
 
 
In the post-game analysis, Stockfish marks 19. Re3 as a blunder. After 19...axb3 (engine's best move), the analysis gives a score of +0.99. Once I make my next move (20. Bxh7!!), the analysis gives a score of +7.35. In the diagram above, I tried to show all of the variations that could come from the sacrifice.
 
 
My opponent never saw it coming! happy.png
 
 

 

klibuscul

Nice :clap

Lego57man

very nice jollification, still i am struggling to even beat my 850 point ranked opponent, i see you have been getting increasingly better and hope to one day learn ur skills. 

nuclearkid5
No offense,but you didn’t really outsmart Stockfish.After 19.Re3,he could have played 19...g6
Jollification
nuclearkid5 wrote:
No offense,but you didn’t really outsmart Stockfish.After 19.Re3,he could have played 19...g6

The point is that Stockfish said the best move for black was 19...axb3.

Marc101101

21 rg6 and 22 rookh8 seems to save the day for black, which would mean the computer was correct. 

Jollification
Marc101101 wrote:

21 rg6 and 22 rookh8 seems to save the day for black, which would mean the computer was correct. 

Assuming you made a typo and meant 21...Kg6 (not Rg6), 22...Rh8 can be met with the stunning move 23. Nh7!!. I will add this line to the original post.

tacticspotter
Jollification 写道:
nuclearkid5 wrote:
No offense,but you didn’t really outsmart Stockfish.After 19.Re3,he could have played 19...g6

The point is that Stockfish said the best move for black was 19...axb3.

Which type of stock fish you use? mine says white up by 4.8 and best move h7 and nothing about taking the knight, I suppose you use a lower depth of the engine

tacticspotter

stockfish saw it coming but young stockfish(depth 18-20) did not

Jollification
tacticspotter wrote:
Jollification 写道:
nuclearkid5 wrote:
No offense,but you didn’t really outsmart Stockfish.After 19.Re3,he could have played 19...g6

The point is that Stockfish said the best move for black was 19...axb3.

Which type of stock fish you use? mine says white up by 4.8 and best move h7 and nothing about taking the knight, I suppose you use a lower depth of the engine

https://www.chess.com/analysis/game/live/5217476863 The deepest depth chess.com lets me go to is 18.

tacticspotter
Jollification 写道:
tacticspotter wrote:
Jollification 写道:
nuclearkid5 wrote:
No offense,but you didn’t really outsmart Stockfish.After 19.Re3,he could have played 19...g6

The point is that Stockfish said the best move for black was 19...axb3.

Which type of stock fish you use? mine says white up by 4.8 and best move h7 and nothing about taking the knight, I suppose you use a lower depth of the engine

https://www.chess.com/analysis/game/live/5217476863 The deepest depth chess.com lets me go to is 18.

That is exactly what I meant by lower depth

Marc101101
Jollification a écrit :
Marc101101 wrote:

21 rg6 and 22 rookh8 seems to save the day for black, which would mean the computer was correct. 

Assuming you made a typo and meant 21...Kg6 (not Rg6), 22...Rh8 can be met with the stunning move 23. Nh7!!. I will add this line to the original post.

Well seen. The computer was wrong then. I have this feeling too sometimes. The computer can say a move is only good, not excellent or the best, while it allows a better move after and enter into a scheme. But I never caught the computer making such a blunder. Good job 🙂

pfren

Typical case of bad engine usage, and/or a very weak engine.

Of course any properly operating engine evaluates something like +6.0 after Black blundering a piece at move 15, and finds the pretty obvious 20.Bxh7+ instantly.

 

Feeding it to my computer, its only objection was why Bxh7+ was not played at move 18 instead of that Re1/Re3 thingy (which is also good, of course).

pinkblueecho
pfren hat geschrieben:

Typical case of bad engine usage, and/or a very weak engine.

Of course any properly operating engine evaluates something like +6.0 after Black blundering a piece at move 15, and finds the pretty obvious 20.Bxh7+ instantly.

 

Feeding it to my computer, its only objection was why Bxh7+ was not played at move 18 instead of that Re1/Re3 thingy (which is also good, of course).

 

Listen to this man, he know´s what he´s talking about. He is a world champion in correspondence chess which allows the use of computers! His knowledge of the game has helped him absolutely destroy Stockfish users who lack his deeper understanding of the game.

pfren
pinkblueecho έγραψε:
pfren hat geschrieben:

Typical case of bad engine usage, and/or a very weak engine.

Of course any properly operating engine evaluates something like +6.0 after Black blundering a piece at move 15, and finds the pretty obvious 20.Bxh7+ instantly.

 

Feeding it to my computer, its only objection was why Bxh7+ was not played at move 18 instead of that Re1/Re3 thingy (which is also good, of course).

 

Listen to this man, he know´s what he´s talking about. He is a world champion in correspondence chess which allows the use of computers! His knowledge of the game has helped him absolutely destroy Stockfish users who lack his deeper understanding of the game.

 

Nah, it's not a matter of knowledge (although it does play a role), but rather WORKING HARD on finding the appropriate moves over situations that are not always easy, or standard. Pretty much the very same thing one has to do to improve in OTB chess, right? There are things computers hadle perfectly, and others in which they miss something important. here, the case is very simple: A weak engine (as almost all online engines are) evaluating at low depth produces unreliable evaluations most of the time.