But in this new analysis it says f4, then goes on to suggest move 17 is Qg4? This doesn't make sense. Is there a bug in this new analysis window?
f4 is best? Really?
Sort:
I went back to the game to do the analysis again and it suggests f4 both for move 16 AND 17.
DONTcrackMEup wrote:
Can you post a screenshot through the game analysis section without that green star stuff, like in my screenshot? f4 is not suggested, and I don't know how you got that star stuff to appear.
Yes, that's what I was wondering! I've seen this new "game report" option show up before but I don't know why it's shown up a couple times now. I'd like to access the old analysis but I don't know where to find that option now.
Anyway, regarding your comment that you set your depth to 20 (mine is 10, because that's all that's available unless I buy a subscription) vs 10, well, that refers to the number of moves it looks ahead does it not? Whether the depth is 10 or 20, shouldn't it recognize checkmate in 1?
DONTcrackMEup wrote:
So after running the analysis at least a couple times during the day after posting this message and getting the same result, showing f4 as the best move, I just tried it again now and it's showing it as in your screenshot in post #8, with f4 as a blunder and the king move as the best option. Weird.
Apr 12, 2019
0
#7
When reading the title I thought this thread was going to be about the bird opening
PkPum wrote:
Had this come up in the analysis. Computer says f4 is best, even though it leaves a checkmate in 1! Who ever said computers are infallible? As an aside, how do I access the old analysis? This one is nice, but I like the old one that had the black white advantage at the top and list of best, good, mistakes, etc. under it. This new one seems to have some issues with it anyhow!
Maybe the computer is 'alive' and trying to be 'evil' to White.
PawnstormPossie wrote:
Change engine time to unlimited
I can't because I don't have a subscription, but you'd think the engine at a depth of 10 would be able see a checkmate in 1.
PawnstormPossie wrote:
You don't need a subscription.
The engine is limited to a depth of 10 unless you buy a subscription.
PkPum schreef:
PawnstormPossie wrote:
You don't need a subscription.
The engine is limited to a depth of 10 unless you buy a subscription.
Engines are free. You can download a chess gui like arena and install stockfish on it. It works much faster than the engine on chess.com. Stockfish is very unreliable at depth 10. You want to have depth 20+.
I was pretty sure it was how many moves "deep" it's looking. I just searched it and this page seems to describe it in detail, so it is the number of "moves" it's looking. https://uscfsales.wordpress.com/2011/07/08/understanding-chess-engine-evaluations/
So this still doesn't explain, how did the engine miss a checkmate in 1? Nobody has answered this yet, over two pages of replies. Do they just make mistakes? Is there a bug in whatever program chess.com is using? What's the limitation?
Maybe because engine think it is blunder to Rxg2 or maybe don’t calculate enough nodes
Yes, f4 was the last move made by the opponent, and then Q takes g2 was checkmate. Here is the pgn for it.
Apr 13, 2019
0
#15
Yep, 16.f4 is best. Not that hard to understand it's for move 16 not 17.
llamonade wrote:
Yep, 16.f4 is best. Not that hard to understand it's for move 16 not 17.
No, it's not hard to understand for move 16. Now tell me why it was marked as "best" for move 17!
PkPum schreef:
llamonade wrote:
Yep, 16.f4 is best. Not that hard to understand it's for move 16 not 17.
No, it's not hard to understand for move 16. Now tell me why it was marked as "best" for move 17!
Maybe stockfish isn't designed to run at depth 10. It takes like 0.1s to reach that depth and it is very unreliable for sure. You don't need to use chess.com's analysis. Stockfish is free.
DONTcrackMEup schreef:
"You don't need to use chess.com's analysis. Stockfish is free."
Chess.com's analysis is free too. Not sure your point.
I prefer using a program like arena, because it is much faster. I noticed that if I let stockfish run in arena it gives more accurate moves than if I let it run on chess.com for the same time. If you want to do a quick analysis, then don't use chess.com. The difference is really significant. You get different recommendations.
Apr 14, 2019
0
#19
The rook was at f1 when f4 is best. The snap by you and analysis snap is not same - kindly check @PkPum and @DONTcrackMEup. In your snap Rook is at f2 which clearly is mate - engine doesn't refer that at as best move.
PawnstormPossie wrote:
DONTcrackMEup wrote:
What I am trying to figure out is how did the OP get it to change to depth 10? It only shows 16 which corresponds to 5 seconds. If you want a computer to give an accurate readout, shouldn't you wait more than 5 seconds?
OP said 10 was the default depth not using self analysis.
My research (toying around) has found depth of 20 to be inadequate in many cases. I've also found 30 to be unreliable in many positions. 36-38 seems to be pretty good for the majority of positions, while 40+ can be even better.
A problem I've come across with 40+ is the engines sometimes find very exact play (positioning) required by both sides for several moves. This can also take hours to perform.
And of course, by the time I'm close to making any type of elementary conclusions, the engine gets revised and my previous work is now trash. I started on SF8, now on SF10.
You should be able to hit 40+ depth within 10 seconds in most positions in 4 cores desktop.
This is a screenshot from OP position, where my stockfish hit depth 47 within 5 seconds. Of course depth is unreliable indicator of SF strenght. If you dont hit depth 40+ within 10 seconds in most positions, your configuration is wrong.
My system is 2 years old mid range i5-7400, $250 ,4 cores cpu.
In general , if you need serious analysis, just do about 3 mins per position, that hit 1 billion nodes per position. (hit 3400+ strenght, in comparison 0 elo in random play)
Had this come up in the analysis. Computer says f4 is best, even though it leaves a checkmate in 1! Who ever said computers are infallible? As an aside, how do I access the old analysis? This one is nice, but I like the old one that had the black white advantage at the top and list of best, good, mistakes, etc. under it. This new one seems to have some issues with it anyhow!