I am referring to the engine used in the analysis on chess.com, I don't know what it's called. For this game it failed to find the checkmate, while for another it found a checkmate in 8,.
Because chess.com analysis engine is approx 400 times less powerful than my hand phone.
I don't see search speed but as an estimated guess,
1 sec per position vs 1 min = 60 times
Multi PV 3 lines vs single PV = 2.5 times
Web protocol vs Droidfish GUI = 3 times
Total = 60x2.5×3 = 450
But you neglect the part that it found a checkmate in 8 moves which surely has more possible combinations than a mate in 6.
But you neglect the part that it found a checkmate in 8 moves which surely has more possible combinations than a mate in 6.
Mate in 8 vs mate in 6, mate in 6 is more accurate and powerful.
Engine searching a checkmate is like a man searching an exit of a maze. An engine will search all possible turns, turn left, turn right , turn right , turn right etc.
After certain search, an engine found an exit after 8 turns, and then the algorithm will search again later whether there were two wrong turns.
How come the engine does not find a mate in 6, while in other cases can find a mate in 8?
Do you know how many possible positions in search tree in mate in 6?
If you ask an engine to do brute force search, the number of possible positions in mate in 6 is
30^11= 17,000 trillions. Yes it is 1.7 e+16.
Stockfish can search 5 million positions per second in common 4 cores CPU. 1 min search is 300 million positions only and it will take about 10 years
to find the checkmate.
However, thanks to powerful search priority algorithm in Stockfish 10( checkmate, capture materials etc) , Stockfish 10 found in one minute in my phone. It means it search millions time faster than a brute force in this position.
In case Stockfish search algorithm can be improved with the advance in Neural Network, Stockfish 20 probably find the mate in < 10 seconds.
Alpha-beta pruning is used a lot to wittle down options. For example, Stockfish won't consider a move like 22.Qc2 for much longer. I also doubt Stockfish's heuristics are a simple "priority" thing either; that would be quite a weak heuristic. The science of chess engines is not this simple at all.
I am pointing to the simple fact that mate in 8 is harder to find than mate in 6, yet the engine managed to find the former but not the latter.
I am pointing to the simple fact that mate in 8 is harder to find than mate in 6, yet the engine managed to find the former but not the latter.
I dont know how to explain to you.
There is a mate in 6 in this position. (which has been proven with my smart phone)
It is practically impossible for an engine to find mate in 6 straight, or 1: 1,700,000,000,000,000 lucky chance that stockfish found out mate in 6 straight in her first search. Initially she might see mate in 50, mate in 40 , mate in 8 etc, and then she will exclude multiple wrong search to see mate in 6.
mate in 50, 40 or 8 cannot be seen before mate in 6, it's the other way round.
mate in 50, 40 or 8 cannot be seen before mate in 6, it's the other way round.
When an engine like Stockfish is looking ahead e.g. 20 ply, that does not mean that it has looked at every line 19 ply (or shorter) in length. Due to pruning, Stockfish cuts out many moves. In this way, Stockfish is being like a human in terms of finding what it considers to be the *simplest* mate rather than what is necessarily the shortest mate. For example, the mate in 8 may contain more forcing moves that the mate in 6 and this will make it easier and quicker to find.
22.Nf7+ is a blunder because Black can simply play 22...Rxf7 23.Bxf7 and White's attack is over. White has a material advantage and should win in the end, but there's still plenty of work to do. This is a sad waste of a tremendous attacking position
mate in 50, 40 or 8 cannot be seen before mate in 6, it's the other way round.
When an engine like Stockfish is looking ahead e.g. 20 ply, that does not mean that it has looked at every line 19 ply (or shorter) in length. Due to pruning, Stockfish cuts out many moves. In this way, Stockfish is being like a human in terms of finding what it considers to be the *simplest* mate rather than what is necessarily the shortest mate. For example, the mate in 8 may contain more forcing moves that the mate in 6 and this will make it easier and quicker to find.
Thanks for clearing that up! However, another thing, I analyzed the game on Chessmaster Grandmaster Edition with 10 seconds per move and it found the checkmate. Shouldn't chess.com's engine be stronger than Chessmaster's? I suppose the difference is the 10 seconds per move, which enables searching more positions.
mate in 50, 40 or 8 cannot be seen before mate in 6, it's the other way round.
When an engine like Stockfish is looking ahead e.g. 20 ply, that does not mean that it has looked at every line 19 ply (or shorter) in length. Due to pruning, Stockfish cuts out many moves. In this way, Stockfish is being like a human in terms of finding what it considers to be the *simplest* mate rather than what is necessarily the shortest mate. For example, the mate in 8 may contain more forcing moves that the mate in 6 and this will make it easier and quicker to find.
Thanks for clearing that up! However, another thing, I analyzed the game on Chessmaster Grandmaster Edition with 10 seconds per move and it found the checkmate. Shouldn't chess.com's engine be stronger than Chessmaster's? I suppose the difference is the 10 seconds per move, which enables searching more positions.
Too many unknown variables, (search algorithm, small sample size, hardware configuration, etc, etc...)
you cant compare like this.
Your question is like
A xB x C vs X x Yx Z
Is "A" bigger than X? (without knowing all other informations)
You just need to compare chessmaster 10 vs stockfish in your same computer., According to CCRL rating , Stockfish is about 700 elo stronger than chessmaster. Stockfish 10 can play toe to toe vs CM in 1:60 time odd handicap.
Of course Stockfish has no trouble finding the mate. On my weak laptop (Core i3 3110M using 3 threads, 8 GB DDR3, Manjaro Linux x64 and stockfish-modern 181119 nightly build), it needs no more than 5 seconds to spot a mate in 6 starting with 22.Qh5.
After 22.Nf7+? Rxf7 white should still be winning, but the game isn't over yet.
Stockfish 11 developmental version 1911240 found M6 in 7 seconds in my 4 cores 3 ghz i5 desktop.
(Searching mate also involve luck, she may find M6 in 5 seconds in next run, or 10 seconds in another run).
I am pointing to the simple fact that mate in 8 is harder to find than mate in 6, yet the engine managed to find the former but not the latter.
That's not a fact, but very much depends on your (or your engine's) notion of "hard to find".
It's, e.g, perfectly normal for an engine to find something like a mate in 27 in an endgame and later narrow it down to, say, 22. Chess engines don't just do brute force tree searches.
But this was not an endgame, both games were in the middlegame. I am perfectly aware that it is far easier to search for possible moves in the endgame.
But this was not an endgame, both games were in the middlegame. I am perfectly aware that it is far easier to search for possible moves in the endgame.
That was just an example. I chose it because it's seen very often. Endgame or not, the engine has some heuristics to select certain sub-trees for deeper investigation. It's not guaranteed that the quickest mate is to be found there. Actually, my brain suffers from the same issue.
But brain and engine a two different things 
The most interesting thing about this whole discussion is that the OP seems to have never even considered Black's defensive resort 23..Rxf7+! Even after the computer flags 23.Nf7+ as a blunder. I have made this same amateur mistake many times: Not looking for the opponent's best defense. Instead, we look to calculate lines that "prove" that our combination works. No really strong player would play 23.Nf7+, because strong players look to see what's wrong with their combinations, not to reassure themselves.
If you need help, please contact our Help and Support team.
I am referring to the engine used in the analysis on chess.com, I don't know what it's called. For this game it failed to find the checkmate, while for another it found a checkmate in 8,.