PROOF That Chess.com's Engine is Weak!

The engine has no definition of what is considered an inaccuracy.  The engine simply spits back evaluations.

Chess.com takes those evaluations and has its own definitions of what is considered brilliant, excellent, good, inaccuracy, or worse.  But that's not the engine, that's chess.com based on the evaluations.

If you want to argue that those definitions could be improved, that's not unreasonable.  They're arbitrary - any definition will be arbitrary.  But that's not a case of the engine itself being weak.

Well, you could argue that Chess.com's engine is weak if Chess.com is using Stockfish 10 and what I used compared to it (for White in the example I gave in post 1) was a DEV build of Stockfish that is newer than Stockfish 11, and Chess.com is using 1 CPU and I'm using 20.

Wouldn't that say that both the Engine and the Hardware are worse for chess.com?

Stockfish 10 is not weak.  It's unimaginably strong compared to human play and compared to virtually any other engine.

If you have a dev build of Stockfish 11, it's not a surprise that it's stronger.

And if you're using 20 cores and chess.com's online analyser is using 1, it's not a surprise that you get deeper analysis as a result.

I don't see why you consider this a shocking revelation.  The online analyser is not intended for competitive play in ICCF events.  It's intended as a quick guide to where you went wrong, primarily targeted towards the casual player in the 800-1800 range, and it's perfectly adequate for that purpose.  You don't need ICCF-level analysis to achieve that goal.

But you do need that level to go tooting your horn about how great you are because you had one game that chess.com's engine gave you a 98+ score for and only 1 "inaccuracy" and to then go around telling others that are chess.com "BLITZ" rated 200 lower that they know nothing when they are simply trying to answer questions that other people ask about things we do know, like in my case, French theory.

I'll let you figure out who that user is since I refuse to direct this at anybody and get my own self in trouble.

How can it be weak. Fritz 5 in a quad core and 8 gig ram could not beat Stockfish 10 in a Pentium 3 and 512 mb ram. 

A handful of engines with a maximum of search depth can beat the chess.com engine with less search depth? 

Breaking news. And still any human wouldnt stand a chance. 

Pretty much anyone will eventually have a game that is considered near perfect.  I have a 70-move blitz game where we both scored 99: https://www.chess.com/live/game/2242614062.

The issue isn't that the analysis depth isn't sufficient.  It also wasn't a particularly brilliant game.  Basically all we did was trade off all our pieces and then play out a drawn endgame where the best move was easy to find.

Deeper analysis would have gotten pretty similar results from that game.  In that endgame, virtually anything was going to lead to a draw, so unless you make a move that actually puts that in jeopardy, it will count your move as "perfect".

So if someone manages a one-off near-perfect game, it doesn't prove that they're brilliant.  But it also doesn't generally hurt to let them have their five minutes of glory if they really want it.

I disagree.  Stockfish 10 is much stronger, but with that kind of a hardware handicap, Fritz would dominate.  Fritz is also unimaginably strong by human standards and pretty strong by engine standards too.  It's in 13th place on the CCRL list, just over 200 points behind Stockfish, and if you give it significantly better hardware that will more than make up for the 200 points difference.

If you want to find an engine that would lose even with that kind of hardware advantage, you're going to need to look at engines rated below 3000 at a minimum, and probably even lower than that.

I'm honestly not sure Fritz 5 on your home computer could beat BrowserFish. I guess it depends on your hardware, but engines have come a long way.

Why don't you try it. I tried it with Stockfish 4. Fritz 5 played like an amateur even with huge hardware advantage.

Hi

Oh, wait.  I just realised you're talking about Fritz 5.  The latest version of Fritz, which is in 12th on the CCRL list, is Fritz 17.  CCRL doesn't have anything older than Fritz 8, which is down at 2700.  According to Wikipedia, Fritz 5 released in 1998.

Yeah, I think it's safe to say that it would take an enormous hardware advantage for Fritz 5 to beat Stockfish.

Modern Fritz 17 would fare much better.

it's not that stockfish 10 is weak, it's just not being let to calculate deep enough.

This is not a place to advertise. 

Fritz 17 with the said hardware advantage would beat Stockfish 100-0. Pentium 3 is too slow. Old engines like Fritz 5 cannot cope up because it has no ability to formulate a plan, just pure calculations.

So you started this whole thread because someone is annoying you? You're one of the most knowledgeable players on this site why do you even care.

chess.com's engine is weak because it only calculates moves up till a depth of 14 moves.

An average engine would calculate till 20-30 moves (even though it may take more time).

That's not a problem with the engine.  It's a problem with how much time you're prepared to give it.  Also, the online engine is probably single-threaded, whereas you could download the same engine and run it locally with multiple threads.

Is it 14 moves or 14 ply.  Depth of 14 ply can see things that a world champion cannot see.

14 ply is only seven moves.