Forums

# Question regarding the engine

Sort:

Hi guys,

First post; because I don't understand what is going on.

After I play a match I always analyse. Sometimes I encounter something strange. It probably has been asked many times, but I couldn't find it.

Example: Opponents makes a blunder. -7,25 for me. The best move according to engine is Rook B4. I saw that move, because of the check of Bf2, capturing the queen a move later with the rook. However, I thought that white could play Qa6 as the escape square, so I decided to castle and would go on with my life.

During analysis it sais -7,25 for black. Then it stays 7,25 with rook b4, which engine sais is the best move, but drops to -1.5 for black when Qa6 is played (which is the best move).

I guess my question is: how can the engine give me such different results, when the best move is played each time? Not sure If i make myself clear though.

Is it because I currently do not have a paid membership, is it because I am missing something or what is going on?

Mosselmaniac

Mosselmaniac wrote:

Hi guys,

First post; because I don't understand what is going on.

After I play a match I always analyse. Sometimes I encounter something strange. It probably has been asked many times, but I couldn't find it.

Example: Opponents makes a blunder. -7,25 for me. The best move according to engine is Rook B4. I saw that move, because of the check of Bf2, capturing the queen a move later with the rook. However, I thought that white could play Qa6 as the escape square, so I decided to castle and would go on with my life.

During analysis it sais -7,25 for black. Then it stays 7,25 with rook b4, which engine sais is the best move, but drops to -1.5 for black when Qa6 is played (which is the best move).

I guess my question is: how can the engine give me such different results, when the best move is played each time? Not sure If i make myself clear though.

Is it because I currently do not have a paid membership, is it because I am missing something or what is going on?

Mosselmaniac

You could make answering easier if you told us what move you were looking at... But I figure it was black's 11th.

That is correct. Sorry.

This is game URL but that icon didn't work for pasting URL: https://www.chess.com/game/live/33662991519

Can you help?

So I didn't see any move with anything close to a -7.25 advantage for black. It does say 11.. Rb4 was the best move.

The analysis page does show different recommendations for the best move.

The difference would be owed to a different depth of calculation. The Game Review calculation is done on the chess.com server.

It has nothing to do with paid membership.

I don't really see the case you are talking about.  I do get some evaluation jumps in the game you posted, but only when one of you makes a bad move--it's never when you make the suggested move, as far as I can see.

This COULD happen once in a great while, if the engine is missing something.  Engines do miss things, especially very deep things, but not so often.

In particular, it is not completely clear which move you are talking about.  You talk about Qa6 being played, but I don't see that it ever was played.  If you are confusing your squares, that could explain it, because 11.Qa4 does happen.

If that is not what you mean, please clarify and I'll try again.  It's good to use move numbers to show where you mean...

OK...move 11.  So...you are talking about Qa6 being played.  It never was played.  Qa6 was recommended.  Qa4 was played instead.  That's the cause of the immediate jump.  But your problem might be, chess notation can be confusing.  It's pretty easy to make a mistake if you aren't used to it.

Then the engine likes 11...Rb4 for a few seconds before it changes its mind.  This is caused by the engine needing more than a few seconds.

For sure the engine is not playing bad on purpose because you are not premium.  It's just the Stockfish engine, which actually is free to download.

JubilationTCornpone wrote:

OK...move 11.  So...you are talking about Qa6 being played.  It never was played.

He didn’t say Qa6 was ever played, he’s just citing that response as the reason for him not playing …Rb4 and instead choosing castling.

Back to the position, I guess the main advantage of …Rb4 straightaway would be to have the option of …Re4+ next, but simply 0-0 looks like a great move and that’s probably what I’d do as well.

JamesColeman wrote:
JubilationTCornpone wrote:

OK...move 11.  So...you are talking about Qa6 being played.  It never was played.

He didn’t say Qa6 was ever played, he’s just citing that response as the reason for him not playing …Rb4 and instead choosing castling.

Back to the position, I guess the main advantage of …Rb4 straightaway would be to have the option of …Re4+ next, but simply 0-0 looks like a great move and that’s probably what I’d do as well.

Maybe.  But then I'm not quite sure what he's asking.  There's no point where the evaluation jumped by 7.5, and no major jumps at all after playing best moves.

If he means it could go 11. Qa4, 11...Rb4, 12. Qa6, I dunno...that's true, but the computer does show that line for me.

The suggested line does change over time, so maybe that's it.  And if so, it's just the engine needs some time.

Thx for the discussion. I never said Qa6 was played, because that would be the correct response to Rb4. And I didn’t play Rb4. I considered Rb4, to trap the queen, but realized Qa6 was an option as a response so the queen wouldn’t be trapped, so I castled.

Now that is clear I guess…

After the game I analyzed it and saw that Rb4 gave me a 7.25 advantage. So that move was recommended as the top move. I was playing on a laptop if that matters. I was like WOW 7.25! What about Qa6?? So I played (with the self analysis, not in the actual game) Qa6, what happened to be the top move AGAIN, and it dropped to ~ 1.5. I just clicked review, but my free review was done already so I use the “self analysis” option.

That’s my question. I think I gave it enough time. I don’t even know how to stop the engine calculating so it’s still a bit weird for me. To be honest, I encounter this more often than this one incident. But now I decided to make my first post up here…

Thx Jubilation and James!!
Update: on my iPad it doesn’t show 7.25…. I swear it showed after the game on my laptop. Just a bug by to low of a depth or something is the answer?

Generally the results of the self-analysis are not the same as the review results because of different calculation depth.

Mosselmaniac wrote:
Update: on my iPad it doesn’t show 7.25…. I swear it showed after the game on my laptop. Just a bug by to low of a depth or something is the answer?

Yes, if it happened like you said, then it's very likely because of low depth.

Such an easy answer, but people making such long posts... very annoying.

There are multiple factors to keep in mind:

1) The bot that chess.com uses is crap!  It uses Stockfish 10.  Stockfish 14 has been out for a while now.  A full FOUR versions ahead of what Chess.com uses.  Think of a version of a chess program as being equivalent to a generation of humans.  So Stockfish 10 would be Stockfish 14's Great-Great-Grandfather!  Stockfish 10 giving Stockfish 14-level analysis would be the equivalent to you telling your Great-Great-Grandfather to Snapchat with you and use tic toc.

Well, with chess computers, while the latest ones (which Stockfish 10 doesn't fit in this category) are stronger than humans, older ones cannot see proper evaluation.  It will say it's equal, and suggest moves for each player, but then suddenly it jumps up or down by multiple points to a winning position for one side.

2) Even with stronger computers, evaluations are not constant.  If after 25 moves by each player, Stockfish 14 says it's +0.54, and White and Black both play the best move for 5 moves straight, it may not be +0.54.  It could be +0.86 or +0.31 by then.

3) At any one given move, let's say Whites 34th move, if it claims one move is +0.86, another is +0.73, another is +0.64, and the next best is -2.59, then yes, the -2.59 is NOT going to be the best move, but it is not automatic that the +0.86 is White's best move.  It could be one of the other two, because again, computers are not perfect.  Strong, yes, but not perfect!  You have to research far more than just what it claims one move deep!

Mosselmaniac wrote:
Thx for the discussion. I never said Qa6 was played, because that would be the correct response to Rb4. And I didn’t play Rb4. I considered Rb4, to trap the queen, but realized Qa6 was an option as a response so the queen wouldn’t be trapped, so I castled.

Now that is clear I guess…

Nothing is clear. It is impossible to know which move you are talking about. It is even impossible to decide if you are deliberately trolling or not.

It's pretty easy to know what move he's talking about.

I never could have imagined my first post would end up in a discussion about me trolling? LoL I just wondered why the engine would jump so much. I still can't believe that Stockfish 10 is not able to look 1 move ahead though, that's 1 level deep, right? But never mind; the laptop was just plain wrong and I could never be -7.25 up, so I think i'm fine castling there .

Another question regarding the engine.

You see I play the best move here, Rf2. It shows -4.74. However, the evaluation bar on the left shows -2.4. Depth 20 it shows on the right. Isn't that strange? Shouldn't be those numbers be the same?

Friendly bump, now that I noticed this, i see this in every game!

Why isn't this a huge thing? We are evaluating almost every game, but because of depth issues the evaluation is completely different. somethings even -5 points o. In the case of the screenshot -2.3 off. That's huge in this chess game. Can some experienced person point me in the right direction how to interpret this?

I think the number on the right comes from the self-analysis which can't be turned off anymore.