Fixing New Analysis

Sort:
Avatar of erik

BTW, any pasted in examples (actual URLs and screenshots) of on-going issues would be appreciated!

Avatar of flashlight002

@Eric noted. If I see any wobblies I will be sure to post a visual. I personally haven't hit any snags lately.happy.png happy.png. Some in this forum are very experienced and can spot anomalies quite fast.

My main gripe right now is the inclusion of all my "in game" saved self analysis moves and variations (from my daily games) into the analyzed game list, and that new "post game" engine moves & variations, retries, MultiPV lines etc also now get included (in the same font and style too) into the same list. It becomes incredibly messy and unmanageable quite fast depending on how many "during game" self analyses one saved. A solution to this needs to be addressed. I gave my ideas...I hope they were of use.

Avatar of flashlight002

@PawnstormPossie my understanding of unlimited engine time is that it's free reins to the engine to go as deep as it feels it can go given the trees and nodes representing the position being analysed. With alpha beta pruning it probably is quite agressive in creating a viable tree. The question as to when it is finished? Shu. Good one. When it all calms down and nothing is moving anymore is my take on it! There certainly ain't any "finished" notification or indication of any kind in this system. 5 lines of MultiPV sure may take a while though if its a tricky comolex position. I see it took 8 or so minutes to climb one depth... 26 to 27. There is no "engine has finished" notification of any kind on MultiPV on the system. All visual guestimation. 

Avatar of flashlight002

@Eric @dallin I have just noticed this now and it has already been picked up on in another Forum post thread, but the small icons denoting Brilliant, Best, Excellent, Good move, Inaccuracy, Bad move and Blunder that were placed at the bottom of the move bar charts are now all GONE!! 

Apparently one can only now see these icons if one hovers over each individual bar with one's mouse or stylus as these move quality icons have now been moved to the tool tip that only appears if one hovers one's mouse or stylus over that particular bar. This is really a terrible idea/change guys! First off anyone viewing on a mobile device cannot access pop up/hover over tool tips!! So all mobile phone users without a built in stylus on their devise cannot see these icons anymore. Second of all, these icons along the bottom were a GREAT IDEA in my opinion, as at a glance I could quickly see where the concentration of blunders were, where were the great moves etc, and it gave a great visual mapping of the game's blueprint in terms of quality of moves and how they were distributed throughout the game! It was easy to see a turning point move....a blunder icon at the bottom of a long histogram bar. It was quick and easy to find a move to look at in more detail by having these icons at the bottom. Now one has to go through each bar, hovering one''s mouse or stylus over the bar to figure out what the move icon is associated with that move. If you are going to switch off such a key feature can't we at least have an option to decide if we want these icons showing on or off the graph?  

As the user who first reported this said " this is now super annoying" and I agree!!

You had a great idea there and now it's gone!  Now it's just another plain vanilla flavoured histogram. And there is already one of those on the report tab!

Please can you rethink this move. I for one would greatly appreciate it if you brought them back on to the graphs.

Here is the original post thread:

https://bit.ly/30uzLW8

Avatar of flashlight002

@Eric and @dallin I found a perfect example to show you from an actual game how those little icons at the bottom added amazing visual cues that one cannot necessarily pick up from the length of the bar in the chart alone and add an additional dimension if permanently visible.

First move: the histogram shows a small negative shift...but it's a mistake! If I had not hovered over it I would not have necessarily thought so!  

Next move by white: there is still a negative scoring...but look...white made a "best move". If i had not hovered over the bar I would NEVER have guessed.

And thus just from these 2 examples I hope I have proved how one can see NEW and additional data one cannot just pick up from the shape of the graph. Only when you see them TOGETHER do they work synergistically to show the dynamics of what was happening and one sees an even better "x ray" of the game.

By removing these icons to a HIDDEN tool tip you have diminished the utility of this graph and broken your central idea you were building that pulls through in other areas of the analysis ecosystem: the use of clearly identifiable iconography and colors to help easily decode a game and its structure and meaning!!

Avatar of flashlight002

I see there is a new "criticality" graph...but where is an article or release explaining how it works, what the values mean etc? 

Avatar of notmtwain
flashlight002 wrote:

I see there is a new "criticality" graph...but where is an article or release explaining how it works, what the values mean etc? 

https://www.chess.com/forum/view/game-analysis/criticality-graph

Avatar of flashlight002

@notmtwain thank you!! happy.png Very kind of you.

I wonder when the videos and articles are coming. But at least this does certainly help me!

@Eric there really should be a better synchronization between releasing a new feature and an article or PR/Training release news article explaining it to the broader chess playing community. New things just "happen"!...and one gets this "huh" moment. 

PS. @JDCannon said there was also an article in the pipeline to demystify CAPS and give greater insight into how accuracy scores are statistically arrived at. That was quite a while ago. I see constant talk on this topic. Right now it is a bit of a "black box" feature. Maybe @DanielRensch can bring us up to speed on this? 

Avatar of notmtwain

It's @Erik

Avatar of flashlight002

Ooops. Thanks for spotting that!! I went and corrected my last few posts where I had it wrong as well! 

Avatar of flashlight002

@Erik @Dallin: Erik you asked for some screenshots of problems and bugs...here is a new one hot off the press per se...affecting the new idea to put the move quality icons together with the engine eval in a pop up tool tip on each bar. (You already know from my views that I think it's a terrible idea. I attempted to use it on mobile today and it was a frustrating laborious experience...I literally had to try hover carefully my stylus from bar to bar....took forever!!!).  

Ok to the bug: if the engine score on a move is 0.00 no tool tip is shown and therefore no eval data or move quality icon....but these moves could be good or best moves! This has just cropped up in my first analysis using the new idea. Here are the screenshots of this bug:


And lastly, here is some further food for thought on why we HATE the removal of these icons:


 

Avatar of flashlight002
  • @Erik....the bug I have identified in post #327...do you want me to send a bug report or will @dallin show his team. Please let me know.
  • Also what is your opinion of the move quality icons being left out of the score graph now? I really believe this is a functional step back. I was taught as a marketing manager to make things easier for one's customer...not more difficult. When I had to be the lead on building a few big retail websites I read a few web design books and they echoed the same principle of making the visual journey easy for the customer with as few clicks as possible. More actions with one's mouse etc is more work for one. Your idea was great. Now that graph is essentially just a repeat of the one on the report page. So a waste now really. Please please reconsider this. I have seen this so often. A design team come up with a sterling idea. It gets implemented. A few customers (not representing the majority) complain about something. The team starts to second guess themselves. They have committee meetings...and then things are changed for the worse. In my 20 years as a Director I cannot tell you how many times I have seen this happen. 
  • Then an observation of what I call the new product and marketing disconnect. The criticality graph is launched....but it is a soft launch. No articles or help files are made available on how it functions. One member was kind enough to send me a screenshot of a comment made by @DanielRensch from the June in review article where in the comments section someone asked what is this criticality graph. He explained very briefly, but said videos on it are "coming". But it's launched! Those videos and articles should be available now! I asked Customer Support what the value in the tool tip of a criticality bar means. They don't know! They have to find out from the dev dept! @Erik don't you think it would have been a good idea to launch the criticality graph together with a news article on how it works? Plus ensure the amazing guys in Customer Support also knew what it was about? 
Avatar of FiddlerCrabSeason

Hello -

I recently got some wacky results in my post-game analysis of game #230408288 (https://www.chess.com/daily/game/230408288).  If there's a more appropriate thread for me to post this issue, please let me know...

 

A) When I click the "analyze" button from the game page, and am redirected to the analysis page, I get the following bizarre results:

A1) the actual move order from the game:  [thematic game staring with 1. e4 e6...  actual play started with move 2]  ...2. d4 d5 3. Nc3 Nf6 4. Bg5 Be7 5. e5 Nfd7 6. h4 Nc6 7. Nf3 h6 8. Bxe7 Qxe7 9. Qd2 a6 10. O-O-O Nb6 11. Rh3 Bd7 12. Rg3 f6 13. Re1 O-O-O 14. exf6 gxf6 15. Bd3 Rdf8 16. Rg6 h5 17. Kb1 Kb8 18. a3 Rfg8 19. Rxg8+ Rxg8 20. g3... 1/2-1/2.

A2) a screen shot of the post-game analysis "report" tab (https://www.chess.com/analysis/game/daily/230408288):

A3) screen shot of the "analysis tab" (same url as in #A2 above):  **compare the bizarre moves listed below with the actual move order listed above**

B)  When I download the pgn from the game page, then open a virgin analysis page in a new tab, then load that pgn into the virgin analysis page I get the following (presumably correct?) results:

B1)  "report tab" (https://www.chess.com/a/2JLQfQ3SEap7g):

B3)  "analysis" tab (URL same as in B2 above):

So, clearly loading the game pgn into a virgin analysis page seems to work, but clicking over from the actual game does not.

- Mitch

Avatar of flashlight002

Wow! Crazy. I can't even come close to understanding what happened here! This is the second report I have heard of this happening (I saw someone reporting the exact same thing in another forum). @Erik @dallin @JDcannon what's going on?

@mgt3 have you logged a formal e mail bug report via membersupport@chess.com ? 

Avatar of FiddlerCrabSeason

@mgt3 have you logged a formal e mail bug report via membersupport@chess.com ? 

 

No sir.  I figured I'd contribute to an existing thread, as Dallin, et al were already in the loop on this. 

F(all)YI - I've got another anomaly I'm compiling and will post that here as well...  assuming I can re-create it...

- M

Avatar of FiddlerCrabSeason
PawnstormPossie wrote:

@mgt3 I tried, didn’t reproduce results as you did with A2/A3.

Did you save any analyses during the game? Perhaps the report/analysis is from a retry or something...IDK

 

I was able to re-create by going to the game, clicking on the "analyze" button, and when re-directed to the analysis page got the following results (https://www.chess.com/analysis/game/daily/230408288):

Same incorrect move order,  but different evaluations..

And yes, I save analyses during every game.  But I usually clear them out when I do the post-game engine eval.

- M

Avatar of FiddlerCrabSeason

Here's another weird one...

Of the following position John Nunn (Understanding Chess Move by Move) call 22...a5 an "error..  White wins by force with 23 e5!... etc..."  I wanted to see what S-fish thought of that, so I plugged it into the analysis page (https://www.chess.com/a/2NL7YHY4Wap7g):

1) the initial eval:

2) the eval after 22...a5...

3) and then after 23. e5...

Seems that Nunn is right, and that analysis page's initial eval of 22...Nd7 vs 22...a5 is incorrect.

- M

Avatar of FiddlerCrabSeason
PawnstormPossie wrote:

@mgt3 the Nunn evaluation is just a good example of depth=20 not being strong enough. Most engines will do this. It may find Nd7 at depth=21, it may take more. Some positions may take depth=30+ or even 40+ to properly evaluate. They're not perfect and it's not a puzzle with a solution.

 

Just so I understand this correctly... 

1) When I finish a game and click the "analyze" button, that initial analysis - including the "inacc/mistakes/blunders/etc." - that's all based on depth=20. 

2) And if I let it run for 5+ minutes, I may or may not get a different result based on a deeper analysis.

3) But that initial analysis, based on depth=20, will always remain in place, if/when I should return to the analysis page for that game,

4) and if I load that game's pgn into the analysis page, that analysis will be the same as the previous initial analysis based on depth=20 (this seems to be the case in a couple tries, but it might be coincidental).  

Have I got that right?

- M

Avatar of flashlight002

@mgt3 correct. An entire game analysis.....when you click "analyse" and you see that green bar on the left of the board moving up...when it is finished it is to a depth of 20. There is nothing stopping you however from analysing a particular move to a higher depth. Just click the tick boxes  for "show lines" and "self analysis". Then tap the cog wheel (engine settings) and choose a highish time limit like say 1 or 5 min if you want to analyse a particular move to greater depths than 20. You will see as you wait that depth indicator in the right corner slowly moving up and as it does obviously the variations in the show lines are (the MultiPV section) will change. Showing 3 lines is sufficient.

Regarding the bizarre scenario of the wrong moves showing once an analysis is done in the game list I do suggest you formally log it @ membersupport@chess.com as I don't see Dallin or Johnathan or anyone from chess.com commenting on the bug I found or what you have shown. 

Avatar of FiddlerCrabSeason
flashlight002 wrote:

@mgt3 correct. An entire game analysis.....when you click "analyse" and you see that green bar on the left of the board moving up...when it is finished it is to a depth of 20. There is nothing stopping you however from analysing a particular move to a higher depth. Just click the tick boxes  for "show lines" and "self analysis". Then tap the cog wheel (engine settings) and choose a highish time limit like say 1 or 5 min if you want to analyse a particular move to greater depths than 20. You will see as you wait that depth indicator in the right corner slowly moving up and as it does obviously the variations in the show lines are (the MultiPV section) will change. Showing 3 lines is sufficient.

Regarding the bizarre scenario of the wrong moves showing once an analysis is done in the game list I do suggest you formally log it @ membersupport@chess.com as I don't see Dallin or Johnathan or anyone from chess.com commenting on the bug I found or what you have shown. 

 

Thanks, that all makes sense.  But if I return to a game's analysis at a later date, is the analysis I see at that later date the same analysis I saw the first time around?  In other words, is that initial analysis saved on a server somewhere and pinged when I come back to it?  This seems to be the case, but maybe that depth=20 analysis is just freakishly consistent...

Re: bug report - Will do.  I will link them to my post #329 on this thread.

- M