Human Vs Computers

Of course, 2200-2300 is pretty decent play without consuming all the atoms and, thus, all the resources, of the universe.

What would it take for a computer to play at 3000, by computation alone....rather than database storage...with these gross positional errors?

Seems to me, even the best players of all time were not error less. That's why none of them had a 3000 record...not even Kasparov.

Such a computer would be ample sport...without consuming the universe's entire capital.

Without opening books (databases) computers are many hundreds of points below their present playing strength.

Hmmm...

Since the entire universe's atoms...and perhaps more...would be needed to store the entire database of chess play...and, thereby, creating a "perfect" god-like chess playing machine...is it feasible to create, instead, a computer that calculates near the 3000 level (of course Garry K.'s level would be optimum, too, IMHO)?

I think there's no question that a computer could play near a 3000 level.  They are probably close to that now.  But they will still not play perfect chess, and therefore theoretically beatable by a human.

The only way that I'm aware of to play perfect chess is to have a computer compute every possible game (unless somebody comes up with a mathematical proof for determining the best play each move, which is unlikely).  Then it would be solved, with no error in play.  Of course having a perfect playing computer isn't necessary.  Even the top players have a tough time of it against the current crop of computer engines.  But solving chess would certainly be a huge milestone (mostly for the technology that was discovered to do it).

Hey, redbird...nice thoughts.

Also, I see that...like me...you've been on board awhile yet haven't played anyone. What are your thoughts on that, if I may inquire?

Are there any chess engines that you are aware of that don't use a database?

Mainly because I suck!  :-)  I am a total beginner with chess.  I've been on chess.com for about a year now.  I've always been intrigued by chess.  It's a fascinating game.  I've actually only played one game to completion in my entire life.  And that was over 25 years ago.  I somehow won, so I'm probably one of the few undefeated players on the planet.  :-)  Maybe that's why I haven't played yet.  I don't want to lose my 25+ year winning streak.  :-)

I'm just estimating by watching games that when I started here, I was probably around an 800 rated player.  If I were to guess my rating now, I would guess maybe 1000.  Still pretty sucky.  I have no idea how close that is to reality, since I've never played.  I mostly have been doing tactics trainer and chess mentor and watching some of the videos.

You can turn that feature off with the ChessBase engines. It is also possible to turn opening books on or off for any engines running in Arena. I'm not aware of that option with Chessmaster software, but it might be hidden in their labyrinthine, poorly designed menus.

You can do it in Chessmaster by creating a new opponent.  You can set that opponent to have no book openings, as well as no pondering (thinking while it's not the computer's turn), as well as turning off endgame tablebases.

I see. There are some parallels, here. I have gotten back to chess a little over a year and a half...after a 40 year leave of absence. I was very, very rusty...lost almost all sense of pattern recognition...made blunders. Yet, I was still able to beat everyone in our local chess club, three times each...except one guy. That guy was a good, solid chess player. I had him on the ropes...but he just didn't make mistakes. Asked him to play again, since I've had a little practice but he won't do it.

Anyway, like you, I've been practicing on Tactic Trainer here, Chessmaster, Shredder, Rybka, and several other databases. I suck, too...trying to break through 1700 (I guess that would be around 2000 on a database enhanced, cyber turn-base basis). Getting close. Don't want to play until I hit a good, solid 1800. 

Plateaus...one progresses...hits a plateau...keep on keepin' on...bust through the plateau...then there is always a higher one. I know this from many life experiences.

Anyway, hang in there!

P.S.: When I do hit 1800 in the real world, I'm only gonna play a fast chess game...such as blitz...on line. That way...well...I've already publicly stated my concerns, more than once...and have been called an "elitist pig" for doing so. Life.  

1700?  You don't suck!  I suck!  :-)  I wish I could play at a 1700 level.  I guess it's all relative.

Interesting! I have Chessmaster 11. Didn't know these things. I'm going to "tinker around" and see. Thanks, my friend.

Hmmm...I'm still very rusty. I'm making less and less blunders...but not nearly to my satisfaction. Maybe the eyesight...the brain cells...slowing down with the senior years.

But...there is hope. And I see the improvements. Just played Shredder 4. They say I have an Elo of 1792. Not sure I believe them, though.

I imagine...that if that were true, I'd be over 2000 playing turn-based and using a database. (Which...I will NEVER do; perhaps I am getting old, but I don't need a crutch to play chess.)

I like the story, but let's face it: at the very top level, true blunders (I'm talking missing mate in 1 or dropping a piece for nothing, not missing an obvious fork, which would also count as a serious blunder for a top level player) like the one in your story happen extraordinarily rarely. As for how to beat computers, I'm afraid that there is literally no way any more: the gap has become far too big recently. A software like Houdini can beat fairly high-level computer programs 10-0; those programs could beat the low-level ones 10-0; and even these low-level computer programs would comprehensively beat any chess player in history, with perhaps 8.5-9.5/10. Computers have become far too strong for humans to challenge them: all of these so called "anti-computer strategies" crumble when faced with the real thing.

I found your comment about looking four or five moves ahead very interesting. I did a little calculation; taking the average number of possible moves in a given position as 30, a computer would have to calculate through 656.1 billion variations to look 4 moves ahead (8 half-moves) or 590.5 trillion variations to look 5 full moves ahead (10 half-moves in total) - in reality, this is child's play, and even a home program could do it in a couple of seconds. The strongest programs run on dozens of CPUs at competitions with huge processing power can look up to 6 full moves ahead (analyzing 531.4 quadrillion variations) in a second or so. However, with this sort of speed, in a tournament match these computers would be more likely to look 6.5 full moves ahead (6 moves from both sides, plus their own final response), as they could do this within half a minute; if they were in a tough position and needed to look 7 full moves ahead, it would cost them 15 minutes.

However, this doesn't mean that the computers would produce perfect play, as their positional evaluation at the end of however many moves they calculate may be imperfect and will not show all of the correct options - if it were able to do that, there would be no need at all to calculate.

Good comments. Thanks. Someday...I'm gonna finish that story. Maybe, it turns out, Otto is an android with a Rybka 10 built in. Any blunders that he makes are to throw Arthur off. Maybe they are playing on a comet. Maybe Arthur is a...

Oh, sorry, I didn't mean to refer to it incorrectly as a story; I was talking about your recount about the player who blundered his Queen in the post above, not any other story you may be talking about. ;)

Oh, I see. Well the fellow was quite rusty. More so than I.

I set an engine (Rybka 4) thinking about a position from a GM game still in the opening phase and went shopping. When I returned, it was looking 22 plies ahead (eleven moves). Earlier in the day, from a different position, after a shorter period of time Rybka 4 was looking 18 ply ahead, but Hiarcs 12 only 14. This depth is partly a consequence of pruning algoritms. The engines do not examine every conceivable variation to maximum depth, but prune unproductive lines to concentrate on the ones that matter. Rybka's pruning is more severe than Hiarcs's, thus it probes deeper.

I was answering ChessStrategist's question about what kind of advantage it would bring to analyze every possible move until six or seven moves ahead. I am well aware that computers focus on the important lines and go ahead up to 20 moves, rather than looking at every possibility for seven moves.