Fixing New Analysis

Hi @Rasta_Jay

To clear things up. If you go to a fresh BLANK analysis board and enter 1.e4 c6 it gives the correct feedback 2.d4 d5.

If however you go and analyze the game first bought to our attention by @UWannaFork in post #242 and made available as an embedded game (for easy inserting into the analysis board) by @PawnstormPossie in post #253 then you get the suspect results.

This part of the engine - the "feedback section" that you get when you have "self analysis" tick box unticked is the part of the analysis system returning the suspect move.

Notice the centipawn score of +0.28 for the move 2.e5. 

Now see below the screenshot with "self analysis" and "show lines" options selected (so essentially using the engine MultiPV function of the engine): 

Line one recommends 2.d4 d5 and it is the book move AND notice its centipawn value of +0.68. That's quite a lot better than the "best" move above of +0.28. that the engine chose in its "feedback mode".

Also I never mentioned that I was looking for best moves. The engine itself classed 2.e5 as a "best move". See 1st picture above.

2.e5 isn't even a book move. It isn't even listed in the "alternative moves" list in the extensive Wikipedia article on this opening. It is not the right move to make.

I would way rather choose a move that gives me an advantage of +0.68 vs +0.28, notwithstanding the fact it is a recognised book move as well!  

Like i said, 2.e5 isnt recommended anywhere..

Post a link to a game with 2.e5 then? You guys are looking at the analysis the wrong way.

I just posted a detailed bug report here, concerning a new issue, involving the client side lines to stop working in certain situations. The repsonse was that the post was rejected because it violated the community guidelines. I can only guess this was because I abbreviated "analysis" (with dot!).

I'm not going to text this all again, but here are the screenshots at least:

I'm definitely losing patience here.

So the last new bug I wanted to report (but instead was threatened to be kicked out, because they can't even get the regex expressions for the site policy check right, see above), does not only manifest itself with TT puzzles, but also with full games:

No need threatening to kick me out for much longer ...

@Rasta_Jay evening. We aren't looking at the analysis the wrong way at all. There most certainly is an error happening. To get the error to appear you must analyse the game that causes the error through the analysis board. To verify the error please  go to post #253 as I mentioned above and analyse that entire game. The error in the analysis only happens when you try analyse that particular game . As I said already above...if you just put 1.e4 c6 into the analysis board as you have done everything works just fine. To review the initial bug report please read through post #242 and posted by @UWannaFork.

chow for now

@Beaker620 yes Lucas Chess is great, but there isn't a mobile version btw. Boy I wish there was! Anyways while we all appreciate there are many other ways to analyse a game, the facts are:

1. As a paying member I/we expect the site features to work properly...and the purpose of this Forum post thread is to bring faulty analysis examples to the dev team's attention so that they can investigate and fix the engine. @dallin the vp of product design for chess.com is quite active in this forum and uses our examples to give to his engine analysis dev team.

2.Novices may not understand or even pick up on an error made by the new analysis system and blindly trust the erroneous engine result. An excellent example is a most recent one ...post #242. A novice learning the Caro Kann would be mighty confused!

But thank you for your contribution

@fschmitz422 Guten Abend! Na abend für mich

Yes I share your frustrations. We really should not be seeing such patent bugs and issues so far down the line. Really weird that your original post was kicked out. As you said...must be a syntax error of sorts. 

But to your problem. Boy this one made my eyes water lol. Holy cow!

Clearly the MultiPV section is mixing up white for black and black for white when it returns its move selection in the first example.

The second example is even more bizzare. Here not only has it mixed up black and white but it has taken a move from move 26. Black and made it a move for white in move 27. So it is a totally scrambled result! Bizzarre!!

The only time I have seen something similar happen was when I picked up that when one set up the analysis board manually with black to move....the analysis started showing a white move (but it was actually the move to be made by black). That has since been fixed. 

Did the above bugs go away after a refresh of any kind at all? Or do they continue to show? 

@dallin 3 bugs for your team to check out.

#268 and #269 from @fschmitz422 as well as the bug in post #242 from UWannaFork that I was able to replicate as well.

Thanks for bringing this to our attention, @uwannafork. Certainly an issue. We are looking into this. 

@dallin hi there. I analysed a game now and got to see the fix relating to the situation where all self analysis moves and annotations done and saved in the self analysis area during a game were being wiped out once a game analysis was done (and clicking the self analysis button after an analysis was done lead to the new analysed  game move list).

Now all one's "prep" work, self analysis moves, variations and annotations done in the "self analysis" area have been added to the new analysed  gamelist. This makes things VERY confusing as the "old" is being mixed with the "new".

In the past one clicked on the self analysis button, entered a "space" where one could experiment, plan and figure out potential moves and strings of moves. If one was a paying member one could save all this.

When one did an old analysis one could still go back and click the self analysis button to view one's saved moves and annotations, BUT they remained totally SEPARATE from the new analysed game move list.

Now all one's old saved self analysis data is amalgamated with the new analysed game move list, which can get very very confusing!

For example if one say does a retry and clicks on the move variation recommended by the engine for this position it can insert it in between and amongst one's old self analysis data for that move and it then becomes VERY confusing what is what as the new engine variations aren't for instance in another colour to distinguish them from the old self analysis moves data. This exact scenario happened to me. As a result I had to go and delete my old self analysis work at that point and then re insert the engine variation again....then I could easily make sense of the list of new moves for that point in the game.

Why have the dev guys mixed up the old saved self analysis data with the new analysed game list? It makes things very messy and confusing when now adding new "best move" variations to the analysed game list as they become mixed up with one's own self analysis data...some of which will be redundant or wrong moves or just practise/experimentation one did at the time.

Is there no way to still keep saved self analysis data done during the game accessible by pressing the same self analysis button as one did during the game, keeping it separate and not getting this mix of old and new data all now mixed together!

I hope I have explained myself clearly enough and you get what I am trying to describe.

@dallin regarding my last post #276 I had an idea. Why don't you guys make the 'self analysis' moves that are now appearing in the analyzed game list a different colour? Or even better make new moves and variations and annotations added to the analysed game list a different colour like this:

Then one could easily identify 'during game' ideas from post game ideas (be these manually entered or even variations pasted from the retry and Multipv engine areas)

What do you think of my idea? I would value your feedback. I have asked some of my other chess.com friends for feedback and they all thought it would greatly improve usability. Otherwise the 'saved self analysis'and they all thought it would greatly improve usability. Otherwise the only logical answer is to have 'saved self analysis' kept seperate from the analysed game list. 

@dallin I really would appreciate a reply regarding how the ' saved self analysis' moves and variations now form part of the postgame analysis list.

Here is an example from a game I just played. if I go and insert a suggested move given by the engine feedback system, say for a blunder I made, this 'post game' addition is not discernible from the 'in game saved self analysis' moves . I have highlighted the new variation added.

Now if the 'new' variation or moves were in a different colour (like the example I showed) this would help differentiate 'during game' self analysis from  'post game' moves and variations added to the analysis move list.

Either that or 'saved self analysis' moves should be kept separate from the postgame analysis list.

I look forward to your input.

We are working on some new styles for our Analysis move list, @flashlight002. We will consider differentiation between computer analysis and self analysis as part of this.

@dallin that's BRILLIANT news

Thank you very much for getting back to me. Looking forward to the new features! Really appreciate the feedback you provide and that you guys listen to our input. It is greatly appreciated!

@dallin while I have you on the line so to speak () can I bring up a possible small enhancement I and other mobile users would dearly love to have implemented on the analysis tab that will greatly improve the user experience. I have seen it raised in another thread by a user. Hopefully it isn't too difficult to do. 

In mobile website view right now the forward and backward move control buttons are waaaaay down at the bottom, at the end of the move list. So one has to keep on scrolling up and down, up and down as one works through a game. If the move list is long it is A LOT of scrolling all the time. 

In a normal game the move control buttons bar detaches from its bottom position and floats to the top to sit under the board when one moves away from the bottom of the web page and has the board on screen e.g. 

While on the analysis tab web page this bar does not float to be closer to the board to make moving and watching the board easier.

Could you get this to work in analysis tab? It would SIGNIFICANTLY improve the user experience.

This has been planned for some time, @flashlight002 — thanks for the nudge.

LOL  

Great to hear!

@dallin I do understand your team sure has had their hands full! But it is all coming together bit by bit. I think sometimes we take fore granted the incredible tools you guys are building for us to use! I wonder what the Chess greats from the 1800's and early 1900's would have thought of all we have achieved with engines and analysis boards etc etc. They would have considered it pure magic! 

Btw (and this has nothing to do with the analysis engine system) I have sent via the 'suggestion' interface an observation that there is nowhere to save an analysis board session to (either for the analysis board found under learning or a live analysis board). My only workaround has been to take note of the unique url that one can see in the 'share' interface and paste that url link in a Word document with a description of what the analysis board session was about, so that I can load it later again. There is definitely room for an analysis board database where users can save their analysis board sessions to. I thought I would share that idea with you. One could also include a 'Notes' section too (and there could be the ability to have subsections in this notes area too), where the analysis board 'author' could write more insights about the analysis board session

Anyway -food for thought . I hope my ideas are of use to you guys.

Now both the client side lines AND the lines computed by the server on the fly server have ceased to work. Seeing only wait cursors when I try to evaluate alternative moves.

The story of the development and deployment of this "analysis board" is a complete disaster. In regards to the chess.com development team, it reminds me of the Voltaire quote "But in this country it is necessary, now and then, to put one admiral to death in order to inspire the others to fight."

Although, strictly speaking, this is not just, since I don't question their motivation, but their competence.

@fschmitz422 I am also experiencing a very slow response on the engine feedback system as well. It's with regards to feedback on new moves. I wait and wait and watch it slowly move up to d=20 when it then stops and gives a feedback analysis. But it often takes quite long to get to the d=20 mark. 

So everyone understands:

1. Initially when a game is analysed the feedback system runs on the server. So when it is finished analysing to d=20 a full result is given for each move that was made during the game. After the report is done, clicking on a move you made during the game will bring up feedback info immediately. There is no additional computations required so it comes up straight away.
2. If one now inserts any new moves or variations the feedback system now operated on the client side (in other words uses the hardware of one's computer or tablet or smart phone). It will also compute an evaluation to d=20.

So it is the process for Point 2 that is taking very long. I see how, when I click on a new move the "d=x" will start say at say 16 (it varies what depth it initally shows) and then it slooooowly moves through 17, 18 and 19 until finally it hits 20 and returns feedback data. It has also baffled me why it is so sluggish. It is not the same for all moves though. Sometimes it is faster sometimes slower. But certainly on average quite slow to return a value. There may be a logical answer to this. Maybe @dallin can explain why the local feedback system on new moves evaluated is so sluggish. It doesn't make sense as the MultiPV section works quickly.