app analysis bugs?

JVprof
I review game analysis after every game, and on the website it works great, but the iOS app gives me some weird results. Sometimes the recommended moves are inconsistent (it recommends a best move, but when you play it, it’s labeled “inaccuracy”), and sometimes it makes clearly bogus claims like “Mate in #1” when clearly there is no mate. You play the move and the score goes back to +2.5 or whatever. Seems like lots of bugs to fix!
RussellHawkinsChess

I am having this same issue.

Example: Nc6 is best.....unless you actually make that move

 



JVprof

Yeah, exactly. The move recommended as best move actually has a lower score than what you played. Also the analysis gives different overall accuracy results every time you run it. Looks like the app is pretty buggy.

TheDishesAreDone
JVprof wrote:

The move recommended as best move actually has a lower score than what you played.

A +0.17 evaluation is better than +0.71 from Black’s perspective.

RussellHawkinsChess
TheDishesAreDone wrote:
JVprof wrote:

The move recommended as best move actually has a lower score than what you played.

A +0.17 evaluation is better than +0.71 from Black’s perspective.

Yes, but as you can see in the 2nd picture, after making the move it shows +1.05, which is worse than +0.71 for black

TheDishesAreDone
RussellHawkinsChess wrote:
TheDishesAreDone wrote:
JVprof wrote:

The move recommended as best move actually has a lower score than what you played.

A +0.17 evaluation is better than +0.71 from Black’s perspective.

Yes, but as you can see in the 2nd picture, after making the move it shows +1.05, which is worse than +0.71 for black

Yeah, I completely realize that is the case. I may have read JVprof's comment incorrectly. It seemed like they were saying +0.17 should be worse than +0.71 in the first screenshot, since getting different results when running the analysis again was added as a second thought.

nklristic

Depth 18 can be unreliable and can act up like that from time to time, and if it is depth 10, then it is even worse.  You are a gold member so try bigger depth. I hope it is different then, as it should be. happy.png

JVprof
TheDishesAreDone wrote:
JVprof wrote:

The move recommended as best move actually has a lower score than what you played.

A +0.17 evaluation is better than +0.71 from Black’s perspective.

Sorry, you're right. I overlooked that you're Black.

Nonetheless, it happens a lot in both ways, classifying a move as a mistake, and then it turns out the "best move" was not that much better or even worse.

I gather that the analysis in the app is computed locally on the iPad / iPhone, whereas the analysis on the website is computed in the cloud? That might be the big difference.

TheDishesAreDone

I do not use the app to analyze games, so I do not know much about it. However, I just tried making alternative moves in an already analyzed game within the app, and the search depth for such moves must be uselessly low, since the app's assessment is displayed instantaneously. Whereas, when making such moves in a browser-run game analysis, the engine actually spends time thinking about the new position until your analysis depth setting has been reached, which can be set significantly higher than 18 for premium members. So I would take the app's assessment of the alternative moves you make within an analysis with a huge grain of salt.

Going into the app's self-analysis mode might give you a better assessment of moves. It looks like you can change the strength of the engine for that analysis in settings. Although, I just tried increasing the engine's thinking time, and it did not seem to make a big difference. That might have more to do with the positions I looked at though.