Live post game analysis on android app

Sort:
shine5

So for the past few weeks I've been analysing all my finished standard and daily chess games using the quick live 1 min analysis feature and by going through the game myself. As a basic member I only get the 1 min analysis. So anyone know what's the analysing strength of the live 1 minute analyse?

mkkuhner

I do not know the strength of the 1 minute analysis specifically, but in general whatever chess.com is using for its computer is surprisingly weak.  In the endgame "play the computer" exercises I once managed to beat it in a theoretically drawn endgame, and when I ask for analysis of my own games it often gives variations with a very severe horizon effect.  (Horizon effect:  the computer stops looking at the position just before a crushing move that would change its evaluation completely.)  It can be useful to suggest moves but I would not trust it for more than that.  I doubt the 1-minute is any better!

KevinTheSnipe

I only use the engine to find outright blunders or mistakes. I don't trust what it flags as inaccuracies, many times book moves are flagged as inaccurate.

shine5

mkkuhner wrote:

I do not know the strength of the 1 minute analysis specifically, but in general whatever chess.com is using for its computer is surprisingly weak.  In the endgame "play the computer" exercises I once managed to beat it in a theoretically drawn endgame, and when I ask for analysis of my own games it often gives variations with a very severe horizon effect.  (Horizon effect:  the computer stops looking at the position just before a crushing move that would change its evaluation completely.)  It can be useful to suggest moves but I would not trust it for more than that.  I doubt the 1-minute is any better!

Yes, I too felt that way. Maybe I should just analyse the game myself. But yes the 1 minute analysis points out all my mistakes and blunders correctly , but then again I'm not 2000+strenght :). Thanks for the reply.

shine5

KevinLudwig wrote:

I only use the engine to find outright blunders or mistakes. I don't trust what it flags as inaccuracies, many times book moves are flagged as inaccurate.

I too think the moves it labels as inaccurate aren't in many cases inaccurate. Blunders and mistakes are easier to spot than subtle positional inaccuracies. I use the engine for the same reason as you said.

mkkuhner

One place *not* to trust the engine is when you are winning.  I had a position with two continuations:  swap queens into a 2 pawn up K+P endgame that was *very* easy to win, or keep the queens on, make a bunch of very tricky defensive moves to avoid mate, and end up 3 pawns up.  The computer regarded the first line as a blunder, but any strong human would play it instantly.  (In fact when I did, my opponent just resigned.)

shine5

I have the droidfish android app which uses stockfish 6. It's very strong. Why I don't use it to analyse my finished games is because it doesn't categorize my moves as mistakes or inaccuracies. But now I understand I should use droidfish app because it's way stronger and I can trust it 100%.

Thanks for taking your time to reply mkkuhner and KevinLudwig.

mkkuhner

Stockfish is very strong but watch out--it will give you the move that is best for a computer, but that is not necessarily the move that is best for a human.  It has no fear and is always ready to endure a huge attack in order to grab one more pawn, as long as it calculates that it can survive.  We humans should generally not go for that....

Its other problem is that it is playing to maximize its score with no real concept that it's actually supposed to *win*.  I swindled a strong player once and we ended up in a position where each king has to sit in front of some passed pawns.   Absolutely a draw.  No one can budge or they lose instantly.  But Stockfish gave me +0.2 because I had more pawns.  It could not win or make any progress, but it was sure that somehow that one side had an advantage anyway!

That's how GMs draw it--they find positions in which it thinks it's better but will play till doomsday without making any progress.

(These are not criticisms of Stockfish, all strong engines are like this.  They play better than us, but not like us, and sometimes it shows.)

shine5

mkkuhner wrote:

Stockfish is very strong but watch out--it will give you the move that is best for a computer, but that is not necessarily the move that is best for a human.  It has no fear and is always ready to endure a huge attack in order to grab one more pawn, as long as it calculates that it can survive.  We humans should generally not go for that....

Its other problem is that it is playing to maximize its score with no real concept that it's actually supposed to *win*.  I swindled a strong player once and we ended up in a position where each king has to sit in front of some passed pawns.   Absolutely a draw.  No one can budge or they lose instantly.  But Stockfish gave me +0.2 because I had more pawns.  It could not win or make any progress, but it was sure that somehow that one side had an advantage anyway!

That's how GMs draw it--they find positions in which it thinks it's better but will play till doomsday without making any progress.

(These are not criticisms of Stockfish, all strong engines are like this.  They play better than us, but not like us, and sometimes it shows.)

Yes, I'll definitely keep that in mind, while I enjoy the thrill of taking risks, I would never dare enter such position if I can't find a reasonable defense right away( sorry Tal!). And thanks again for the advise.