Is Stockfish 10+ analysis valid ?
Stockfish 10+ says black make blunder in this position:
This is position after using moves that Stockfish 10+ suggested, prove that it wasn't blunder ?!
second example - black is winning but white doesn't have blunder ?!
In the second example White blundered at move 11. g3??, although you don't need to blunder to be able to lose.
In the first example, what moves did SF actually suggest? At a quick glance that seems like the line of your game. White starts with two blunders in a row and several inaccuracies to follow so they clearly are not SF's suggestions.
look little better at example 1...there is no actual blunder in black move No.14. Then Stockfish10+ suggest moves with no material advantage. So, iff there is no blunder black was cheating !
There are some (admittedly rare) positions where there is a great line that relies on a critical move just out of reach of SF in terms of where its horizon is at a given time that can effectively retrospectively make the move that SF recommends at some point a blunder. This is extremely rare because even if there is some deep line 25 moves away, many of these lines are not forced and therefore can be corrected before that eventuality happens. In addition, there are often lines even deeper that might nullify that line itself.
However, I don't think your examples are really there. I think that as laatikko said there were all sorts of other suspicious moves that took place in that line almost certainly not from SF and therefore SF was not wrong.
The thing about an engine is that the best move is ONLY the best move if you play perfectly. Going down a line where perfection is required is what we call a sharp line. You should only do this if you are an exceptional player who never blunders and the reason you would do it is if you can see you would get a some (small) advantage if you play perfectly 7 moves down say. The computer of course can do that because it never blunders to the limit of its horizon and therefore it is the THEORETICALLY best move for a perfect player but probably not the best for a human that can blunder.
The best grandmasters sometimes try for these sharp lines in order to eek out a small advantage that they otherwise could not get in standard play (due to the consistency of each side and their book knowledge of general chess positions and heuristics) as between many grandmasters in the longer games this might be the only way to win.
I have lost count of the number of times SF has recommended some move in my games that I didn't understand and it turned out to be correct but only if I could see EVERY line just 4 or 5 moves down (which I clearly cannot). I speak as someone who is actually quite good at the puzzles (my current rating for them is 2400 plus). Many of those lines are ones where you give up a piece and there is a forced variation that gets your piece back after 5 moves with a pawn to boot. Well that's great but in a blitz or even rapid game you just don't have the time to get there.
After 14. ...f5 White's advantage jumps from +1.1 to +4.1, so clearly it is a blunder.
Okay I just read your last post with move 14
For that you made the move Nh2 which was the recommended move of SF
Then the recommended move for black was Qa5 or possibly Nd7 as a continuation.
However your opponent blundered by moving his f pawn which was ridiculous - THAT was the blunder
Now you may be getting confused because when you look at the analysis and click on the blunder list at the bottom right SF does NOT take you to the move that was the blunder. Instead it takes you to the move after the blunder and shows with a green arrow what should have been done.
So for example your first blunder was moving the pawn to f3. If you click on blunders for you, you will see the engine go to the position after that move and show you what should have happened with a green arrow.
So you moved f3 but you should have taken en passant. THAT was your blunder.
Hope that helps
One thing to add. Its obviously possible to have a blunder where you gain a massive material advantage. An obvious one is if one opponent puts their queen en prise to distract a defender from guarding the back rank leading to a forced mate. Clearly if you take the queen you gain a massive material advantage but positionally you blundered and as a consequence will lose the game.
Few months ago I uploaded to
lichess.org a game Tarrasch vs Allies (1914) 1-0. Then I press button for analysing. After a while erased that game, and then uploaded game again. In both cases Tarrasch didn't have inaccuracies, mistakes and blunders, but Allies first time had 3 inaccuracies, 1 mistakes and 1 blunder, but second time he had 5 inaccuracies, 1 mistake and 1 blunder. So, Stockfish made 2 different analyses of the same party.
My point was : I think Stockfish 10+ made a mistake in analysing, there was no material advantage for black iff he played something else or move that Stockfish 10+ suggest, white didn't have material advantage after black's move 14., there wasn't blunder for black, that was only one blunder for black in game, and considering his points, he was brilliantly inspired or he was cheating.
Could you please clarify what exactly do you mean with all this, I'm not sure if I understand you correctly:
"there was no material advantage for black iff he played something else or move that Stockfish 10+ suggest, white didn't have material advantage after black's move 14., there wasn't blunder for black, that was only one blunder for black in game"
edit. If I had to guess, I would interpret you meaning that
- Black didn't lose material advantage with the move 14. ...f5
- White didn't gain (immediate) material advantage because of 14. ...f5
-> therefore 14. ...f5 wasn't a blunder.
But this is not the case. White already had an advantage, but after 14. ...f5 it jumped from +1.1 to over +4. That is, White's slight advantage became decisive. Blunder doesn't always mean the opponent getting immediate material advantage. (as spectrox75 above explained with the example of sacrificing a Queen to mate your opponent).
But then again, maybe I read you incorrectly.
Stockfish may make a different analysis depending on the exact amount of analysis it undertook in that particular run. Because this service is SF distributed over the internet, it will dedicate a certain amount of computational power depending on how many users there are etc. So maybe in the first analysis it only looked 22 moves in the horizon, and the second it did 24 or vice versa
The point being that especially for inaccuracies, SF may change its mind a little (its not because it thinks differently, but simply depends on differences in horizon and some computational issues). Inaccuracies themselves are called as such because they are NOT quite mistakes, they are basically moves that change the evaluation a little bit but not much ( not over a point in general i.e. the loss of a pawn) and therefore they may not affect the game at all as your opponent usually has to follow a very sharp line to exploit a single inaccuracy.
So the first run that you had identified 1 mistake and 1 blunder as did the second run. The inaccuracies though went from 3 to 5 probably as a result of a deeper analysis, and it may well be that they went from a change of 0.3 in the evaluation to 0.35 and that was enough to make it an inaccuracy rather than a good move. I am not certain of what the change in evaluation has to be for SF to call it an inaccuracy but hopefully you get my point.
But just to be clear because you are running a system online which will encounter some threading problems, and because the engine must prune to some level, you will get slightly different evaluation analyses each time you use SF. Also, if there are any improvements made in between evaluations via the lichess site you will get a different evaluation.
What is absolutely for certain, however, is at a depth of say 20, it is extremely unlikely for SF to misidentify a blunder, and in fact I would challenge someone to come up with such a line … it would be a very hard computational problem.