Improve Computer Analysis

There are problems with Chess.com's computer analysis, even as a premium member:

1:It incorrectly states "so and so has a decisive advantage" when the move it's commenting on is checkmate.

2:It incorrectly marks many moves as mistakes-even if I'm winning-if there was an alternative. I hate that b/c those moves are NOT mistakes.

3:It's not accurate. Rybka, rated 3000+, provides much better analysis than Chess.com analysis.

Chess.com should also allow us to analyze ANY game. To stop cheating this way, the program should filter out any positions that are in current games.

And a big improvement also would be Chess960 game analysis.

Chess.com could also allow us to view the computer going through our games live and giving exclamation marks like Rybka does.

This could change if Chess.com invests in lots of multi-socket servers (host it's own datacenter essentially) for this feature.

You should be happy that there actually is a computer analysis on chess.com. It's a privilege so stop whining about it. 

This comment is no help. I like the feature but it needs to improve. Just b/c there's no competition doesn't mean Chess.com shouldn't improve this feature.

True, I just tried the computer analysis and it seems quite weak (for a computer). In this position:

The chess.com computer analysis recommended 12.Qa4+ and a long line following (which was winning), but surely a computer should excel in this tactical position and find 12.Bd5+ with forced mate in 7!

Don't know what engine is powering chess.com, but the site admins might want to take a look at Stockfish. It's open-source, free, and very strong (up there with Rybka, see computer chess ratings for 40 moves/40 minutes games). Also plays Chess960.

given enough computational resources, i'm sure the engine they use would find it, but due to the amount of games that the chess.com servers have to analyse, the basic analysis for non members is limited. stockfish and rybka (and quite a few others) are damn good engines, but they are resource hogs in that to analyse a game they require a fair amount of CPU time.

Yeah could just be a matter of resources. If that's the issue, I think they'd be better off making people wait a little longer to receive their analysis. Not much point getting it out the door quick if it's going to be (relatively) inaccurate! That said, the site does offer a lot for free, and it's not very difficult to run a PGN through stockfish/firebird/rybka at home.

Yeah. I did it a couple days ago and it did the same thing your talking about. I hope they fix it soon because I learn a lot from it. They shouldn't send out an inaccurate analysis. I agree with drleper.

i use it sometimes when i'm too lazy to do it on my own.

1:It incorrectly states "so and so has a decisive advantage" when the move it's commenting on is checkmate.

If you choose to read about engine evaluations and how programmers prefer to convert that to a natural language, you'll note that a +3.5 evaluation is no different than a +10.00 evaluation. Both are "decisively" winning.   I don't see why Chess.com's programmers need to offer  a "I has Checkmate in one, n00bz" addendum. However, it can/may offer a "mate in X" countdown.

2:It incorrectly marks many moves as mistakes-even if I'm winning-if there was an alternative. I hate that b/c those moves are NOT mistakes.

I agree to some extent.  When one is clearly winning, whether he picks the +5.0 evaluation scored move or the +2.5 one, he is still theoretically "winning". Though when you ask a computer to rate your move quality, you're going to get what you paid for.

Perhaps you'd rather see the less hurtful "Your move is sub-par, but is still winning" as opposed to "mistake".

3:It's not accurate. Rybka, rated 3000+, provides much better analysis than Chess.com analysis.

Said the apple to the orange. :)

It is only fair to compare an online server's evaluation engines with "OTHER" online server's evaluation engines.  

Chess.com is currently #1 in a field of #1.  Sure, they can strive to add enough hardware to bridge the gap you indicated, but there's no validation of the cost involved to do that yet. 

Let's see:

- No competition (even  in the slightest) to warrant an investment like this.
- Most people on chess.com do not even have an engine on their home PC and they are extremely thankful for this feature.
- Even Computer analysis from a freeware engine like Crafty is 2500 strength when it comes to tactics => which is what most people use them for.

Finally, if you're somebody whose level of play actually "needs"  Rybka (over all the other 2500+ strength engines out there) to give you the winning edge, then ... can I hire you as a coach? :)

 Chess.com is a startup first of all. They don't even have their own datacenter (yet). Maybe they can get some multi-chip servers to do this. those are known to run 128+ CPU threads at once... meaning up to 127 computer analysis simultaneous-just for one server.

 That's if you have Stockfish (never heard of), Firebird (never heard of either), or Rybka at home (or any other programs); not all of us do...I know I don't but wish I did.

I do appreciate the one they have to offer here; even though it may be inaccurate, it still finds moves that I miss. I would not complain if they did upgrade it though...

The good news is you can stop wishing because Stockfish and Firebird are both free and rated 3000+!

I would like to see more natural language in the analysis I think Fritz does an excellent job at this.

 but they need hefty systems.

As pointed out, they are both freely available :) Here are the links for Stockfish and Firebird (Firebird may be of dubious legal status in certain countries, I read somewhere they reverse engineered some Rybka stuff). Stockfish is proper open source, and comes with it's own interface. Both are pretty much up there with Rybka in terms of strength. Rybka would have more bells and whistles though, so probably still worth buying if you're really into comp analysis.

The programs don't need supercomputers to run, they both go fine on my oldish (~4 years) Dell laptop. Anyway, was just putting the info out there, not demanding anything from chess.com, I think their free service is very good.

Rybka's reccomended CPU is an Intel Core 2 Quad 2.4 GHz-exactly what I have. Now try running Rybka for millions of game analysis. You probably would need a whole datacenter or supercomputer to run Rybka for all those. It's a good investment.

I also expected the premium analysis to be better than basic :( and it isn't.

So you wouldn't consider this a mistake?

White is clearly still wining, yet white made a huge mistake.

Well white is still a rook up there.. 

I was a bit disappointed myself when I first upgraded to premium membership status and had a game analyzed only to find it was still with a 2500 rated computer.  Chess.com should get a stronger engine and servers for the Diamond members at the very least it seems.  Just paying for hardware alone and going with Stockfish would be a good option it seems.  I wouldn't even mind waiting 12-24 hours for games analyzed at top level perspectives.

The whole point of a good engine is that for the same system, it can play much stronger. That is, to reach the same playing strength, the better engine only needs a less hefty system.

So a rated 3000 engine would require a much less hefty system or much less time to process than an engine rated at, say, 2500.

Also, note that for every DOUBLING of time spent on processing, the strength of the engine improves by approximately 50 elo (varies by position and engine and hardware, of course). But anyway a quick calculation yields that, if chess.com used the same engine, then it would take about 1024 times the amount of time (or processing power) to raise the computer analysis quality from 2000-strength to 2500 strength. Or, without increasing the amount of time, chess.com would need a thousand times more servers (prohibitively expensive!).

Refer back to my comment about 3000 rated engine vs 2500 rated engine. To let the engine rated at 2500 play equally as strong as the 3000-rated engine, it needs to have a thousand times the amount of time to process!!

Therefore it is crucial for a site like chess.com to have an efficient engine. Indeed, I have read that Erik has posted on the chess engine forum talkchess.com in search for a good engine, so it's good to know that Erik is aware of what's important. :)