Using AI to improve human chess intuition!

[M ANSARI]

I was watching Magnus Carlsen yesterday play 1 0 bullet chess and destroy and adopting 3 opponents all over 3000 ELO on Lichess in one sitting. For anyone doesn't know what adoption is ... it is winning 10 games in a row !!!! I mean really the feat is just incredible and it is amazing how he knows where to put the pieces. MC is well known to have probably the best chess intuition in a chess player ever and at 1 0 he will instantly play moves that he feels are best without really calculating. It seems crazy but he really does it so well! This reminded me a lot of how NNUE works and how just pattern recognition is so important. Maybe it is possible to create a training app for humans where a position is flashed on the screen and you have 5 or 10 seconds to play the best moves with minimal calculation ... just to play a move where your intuition tells you is the best. It could be a defensive move that gives that one extra tempo to be able to protect a position from collapsing or it could be a simple pawn move that gives the opposition and wins ... or even just a positional move that blocks a piece from being effective and thus giving a huge advantage. A point system can then be calculated depending on the move chosen. The app would have say 100 positions and then you get a score and then you try to improve that score.

I think that would be an interesting idea to improve someones chess. Of course it would not be the only thing required to improve chess but for an advanced player this type of application would improve someones game a lot by just giving him a stronger chess intuition. With better chess intuition you will calculate effective lines more efficiently and thus waste less time on useless moves. Is there such an application out there?

[purechess]

Puzzle Rush?

[M ANSARI]

Puzzle Rush is just a tactical calculation with immediate results ... which I think is also very good. But I was thinking more of say ... you have an open file and a rook on hand and you immediately put a rook on the open file. Or say you have a quick way to get a bishop pair ... or say a way to destroy your opponents pawn structure ... or a pawn move that locks in a bishop forever ... a bishop move that constricts the flight squares of a Knight and thus allows a winning endgame. These are not immediate tactical wins, just better moves that have long term benefits and are really not based on immediate tactics.

[Guenther]

I was watching Magnus Carlsen yesterday play 1 0 bullet chess and destroy and adopting 3 opponents all over 3000 ELO on Lichess in one sitting. For anyone doesn't know what adoption is ... it is winning 10 games in a row !!!! I mean really the feat is just incredible and it is amazing how he knows where to put the pieces. MC is well known to have probably the best chess intuition in a chess player ever and at 1 0 he will instantly play moves that he feels are best without really calculating. It seems crazy but he really does it so well! This reminded me a lot of how NNUE works and how just pattern recognition is so important. Maybe it is possible to create a training app for humans where a position is flashed on the screen and you have 5 or 10 seconds to play the best moves with minimal calculation ... just to play a move where your intuition tells you is the best. It could be a defensive move that gives that one extra tempo to be able to protect a position from collapsing or it could be a simple pawn move that gives the opposition and wins ... or even just a positional move that blocks a piece from being effective and thus giving a huge advantage. A point system can then be calculated depending on the move chosen. The app would have say 100 positions and then you get a score and then you try to improve that score.

I think that would be an interesting idea to improve someones chess. Of course it would not be the only thing required to improve chess but for an advanced player this type of application would improve someones game a lot by just giving him a stronger chess intuition. With better chess intuition you will calculate effective lines more efficiently and thus waste less time on useless moves. Is there such an application out there?

Sounds very much like what was possible already with ChessHero?
Was surely mentioned here already a few times, just do a search for it.

I see that it it wasn't updated since version 2.5 and this was already in 2014. Incredible how time goes by...
http://innokuo.altervista.org/chesshero.html

[Ovyron]

The reason MC is so good at bullet is because he doesn't need to calculate because he knows where everything is because of the story of the game and what has happened.

If MC was shown random positions out of context he'd need to calculate, at least a bit, to know what's going on, and wouldn't be as good at picking the move and what other people think is "intuition".

Basically, when a position appears on the board MC had already thought about it several moves ago, and bullet players take great advantage of the opponent's time, that's why MC can't repeat these feats against an engine that replies instantly.

MC is the best at bullet for many reasons but the best moves coming to him purely by looking at the position is a myth, he's been doing a lot of calculations and recalling them or knows them from the out of book opening he'll reach, so basically he's like "10 moves ahead" and when the position is on the board he "already knows", which isn't possible in random puzzles thrown at him.

[Uri Blass]

The reason MC is so good at bullet is because he doesn't need to calculate because he knows where everything is because of the story of the game and what has happened.

If MC was shown random positions out of context he'd need to calculate, at least a bit, to know what's going on, and wouldn't be as good at picking the move and what other people think is "intuition".

Basically, when a position appears on the board MC had already thought about it several moves ago, and bullet players take great advantage of the opponent's time, that's why MC can't repeat these feats against an engine that replies instantly.

MC is the best at bullet for many reasons but the best moves coming to him purely by looking at the position is a myth, he's been doing a lot of calculations and recalling them or knows them from the out of book opening he'll reach, so basically he's like "10 moves ahead" and when the position is on the board he "already knows", which isn't possible in random puzzles thrown at him.

When you play bullet you do not have much time to think on the opponent moves because the opponent play fast so I disagree MC is strong because of doing a lot of calculations.

I guess that MC probably know the best move and the evaluation and the reasons in many positions so
maybe what can help humans is simply memorizing good moves in many positions together with the evaluation and words (like I play a3 in this position to prevent Nb4 or to prepare Bb3-a2-b1 or I play move 1.x1 because I already memorize the lines 1.x1 y1 2.x2 y2 3.x3 y3 and 1.x1 z1 2.x2 z2 3.x3 z3 that are justifications for it).

I say maybe because I do not know about chess players who do spaced repetition about things like this and how much it help them.

[Werewolf]

On the Magnus app it claims (apparently written by Carlsen himself) he chooses the best move by intuition and then uses calculation to verify whether that move is good or bad. Perhaps, quite the opposite of Kasparov's play where calculation was how he "felt around" in a position.

If this is true I think it would explain the incredibly high level of play Magnus has at fast time controls and the (relatively) small gains he makes as as the TC gets longer compared with other players.

[Ovyron]

When you play bullet you do not have much time to think on the opponent moves because the opponent play fast so I disagree MC is strong because of doing a lot of calculations.

"A lot" relative to the time control, after watching hours of bullet footage of Carlsen on Youtube I concluded that he's mainly out-calculating his opponents, not relying on intuition.

I guess that MC probably know the best move and the evaluation and the reasons in many positions

This isn't a big factor because bullet games play a novelty and get on a position MC had never seen before, I believe the big factor is that while he's playing opening moves from memory he's already calculating all out of book positions and deviations that might happen, and when they do he already knows what to play (some sort of pre-calculation.)

Pay attention to the moments when he takes longer to move, and his face of surprise, this is when the opponent makes a move he didn't pre-calculate, and he has to rely on intuition and calculate on the spot. He can still play the greatest move in the situation, but he can't play it as fast as the other ones.

[Uri Blass]

When you play bullet you do not have much time to think on the opponent moves because the opponent play fast so I disagree MC is strong because of doing a lot of calculations.

"A lot" relative to the time control, after watching hours of bullet footage of Carlsen on Youtube I concluded that he's mainly out-calculating his opponents, not relying on intuition.

I guess that MC probably know the best move and the evaluation and the reasons in many positions

This isn't a big factor because bullet games play a novelty and get on a position MC had never seen before, I believe the big factor is that while he's playing opening moves from memory he's already calculating all out of book positions and deviations that might happen, and when they do he already knows what to play (some sort of pre-calculation.)

Pay attention to the moments when he takes longer to move, and his face of surprise, this is when the opponent makes a move he didn't pre-calculate, and he has to rely on intuition and calculate on the spot. He can still play the greatest move in the situation, but he can't play it as fast as the other ones.

There is not much time to calculate in bullet.
Of course MC is calculating and the same for other players but you need to know to calculate the correct lines in order to find the best move and it means that even if sometimes you do not know the best move and you need to calculate then you already know important information from memory based on similiar positions(for example that you are almost sure that the best move is A or B and you need to calculate to decide).

A player who does not know that the best move is A or B will not be able to play as well as MC even with the same abilty to calculate because
that player may spend time on calculating C and D that are bad moves.

I do not see the basis for the claim that the ability of Carlsen to win at bullet is based on calculating more lines than the opponent and not based on superior knowledge.

I do not have a proof that superior knowledge is what helps MC relative to top players but it is my opinion based on what I read in the past(I remember reading that MC's evaluation of lines is better than Kasoarov when Kasparov calculated more lines)