Why in the position



after move 7. ... Kxf7?? the "cloud" evaluation is +6.1 while the engine after recommended 8. Qh5+ gives only +4.1 and making many recommended moves it is hard to get more than +4.0 ?

I was sure "cloud" means "cached from many games and checked with deep analysis", so I trust "cloud" evaluation more than regular one, so why does it fail here?

Black is dropping his queen. What difference do all those little numbers make?

@Motroskin said in #1:
> I was sure "cloud" means "cached from many games and checked with deep analysis", so I trust "cloud" evaluation more than regular one, so why does it fail here?
The cloud analysis will have been cached at some point in the past by someone running the in-browser analysis on their side. It's possible it was done with an older version of the web stockfish, for example, or their computer happened to get to a +6 evaluation. Stockfish's devs do not put very much effort into distinguishing between +4 and +6 - what they care about is finding the best move, and it will be the best move regardless of the immaterial difference between +4 and +6

@MrPushwood said in #2:
> Black is dropping his queen. What difference do all those little numbers make?

Because some players are superstitious and will believe anything an engine will tell them?

@MrPushwood said in #2:
> lack is dropping his queen. What difference do all those little numbers make?

Just curious why this happens for the case this is a mistake in site implementation which potentially could lead to mode serious mistakes. Just want to stop in order to help find potential bug.

But @corvusmellori has covered the topic inside out, so I have the answer here and calm now :)

One point for your question, again, not to argue, but to explain my eagerness here. You always forgetting that I am an amateur and there is a huge gap between you as NM and me. For this evaluations the gap has the following interesting outcome. For you both evaluations are the same "black is dropping his" queen and you in 99% of cases (maybe 100%) will win this for white even against alpha-zero engine. For me this is quite different. The +6.0 evaluation tells me that I have 90% chance to win this in blitz against 1800 rated player; +4.0 tells me that my chance is still 70% or less. You may ask if this is important, "Just play your best and forget about this", but for me it is, since it defines the level of risks I can take, the strategy (positional or combinational approach to use) and how to distribute my time; attack with confidence or rely on dropping the flag.

Of course, this is not for current game, since the evaluation is not known for me when I play, but for the future games, since this is a classical trap and I will do it many times in future.

@corvusmellori said in #3:
> The cloud analysis will have been cached at some point in the past by someone running the in-browser analysis on their side.

Thank you. This makes sense and explains this. I was sure the value is updated every time when the more detailed analysis is done by someone and the position (trap) is classical, but I see the reasoning. Thank you.

@Toadofsky said in #4:
> Because some players are superstitious and will believe anything an engine will tell them?

I want to :) At least I hope that with continuous improvement engines some day will come to this.

I do remember that I owe you a game when you already pointed me that such difference is not definitive. I just didn't want to take up your time to argue when I was agree, but still want to play that game with you. The problem is that I lost that thread and with it the party link, but anyway I still keep hope to find it, learn the position with chess engine and give a fight.

I am always curious how people like you and @MrPushwood (any many others!) find time to develop engines, participate in communities, became NMs and at the same time find time to answer my amateur questions. I can't even imagine such a productivity... that's why I didn't want to take up you time when you proposed a game in that thread many months ago.

@Motroskin said in #7:
> I want to :) At least I hope that with continuous improvement engines some day will come to this.

The other thing most players forget is that almost all displayed evaluations are not "live" from an engine which can answer questions, unless you're running an in-browser (or in-app on Android/iOS) live engine. So even if engines improve to a point where their evaluations somehow mean something, in most cases those evaluations will be stale and live engines may provide contradictory evaluations and variations.

> I can't even imagine such a productivity...

Part of it comes from having clarity of thought about what problem you are trying to solve, and not trying to solve every problem simultaneously.