http://blog.chess.com/pumpupthevolume247/me-vs-houdini-20---queen-vs-2-knights-endgame
http://blog.chess.com/pumpupthevolume247/me-vs-houdini---queen-vs-rook
Check these for computing power 
Humans v Houdini chess engine (Elo 3300)
Sort:
Jul 11, 2012
0
#203
-kenpo- wrote:
Haiku575 wrote:
While I am not currently aware of how big of a difference cores make on ELO ratings, I find it highly unlikely that running Houdini on one core would cause it to play at a 2300 level. What I can say is that at 4 cores on a 64-bit processor, Houdini reaches a speed of around 3-4 million nodes per second. On a single core it usually reaches 700 thousand.
mind you, the hardware they run the engines on during the computer chess world champions, from which these official elo lists are derived, is normally way above 10 cores, no?
That is correct. Usually the engine authors build their own supercomputer for tournament play.
Jul 11, 2012
0
#204
-kenpo- wrote:
Haiku575 wrote:
Oh boy, I sense another conspiracy theory beginning. Now we'll have a bunch of people running around claiming Kasparov learned chess from aliens.
I don't have any idea what you are talking about.
I don't have any idea what I am talking about either.
an_arbitrary_name wrote:
Hmm, but does strategy matter much when you can see 64 moves ahead in every single line? It seems to me that strategy is essentially a bunch of thinking shortcuts that humans use because we're so intellectually limited.
Interesting comment. Instead of saying "intellectually limited', I would have said "slower calculation speed" though. Humans are and always will be more intellectual than computers. We can reason, feel emotions and do other many things that computers can't do. Calculations are what computers do best though, which gives them a leg up in tactical chess. I believe that a Fischer, Kasparov or Carlsen, given enough time, could beat (or at least draw) your Houdini though. Houdini wouldn't come up with a better move after looking at a position for an hour. It would find it's move within 5 seconds and stick with it. A Magnus Carlsen would find better moves after 5 seconds though. Maybe we should have a correspondance match, Houdini vs Carlsen and see what's what. Since the computer has access to a complete opening book in it's memory and taps into it, Carlsen could have access to an ECO while he's playing. No computers or Grandmaster helpers though.
-kenpo- wrote:
sometimes engines do change their evaluations after several hours.
Really? Times have changed then. I don't play against computers much these days, so I'm not up on the latest technology.
Jul 11, 2012
0
#207
JeffGreen333 wrote:
-kenpo- wrote:
sometimes engines do change their evaluations after several hours.
Really? Times have changed then. I don't play against computers much these days, so I'm not up on the latest technology.
Houdini often changes its evaluation in a couple of seconds.
Interesting and insightful post!
Indeed. Post #273 might bring this sleepy 6 month old thread back to life. 
Thanks for penning a compelling argument, @FirebrandX.
Usually these posts are just blather. You have broken the mold.
So FirebrandX is one of the best CC players in the world?
And there is still no answer to the question of how much a difference hardware makes for your engine to play well (specific number)? My Houdini 2 even with an ordinary PC still seems extremely strong, although I probably can't tell the difference between, say, 2800 and 3100 level 
Yeah, but 2300 gets to the point where it, although strong, can become visually fallible -- if you played against it enough, you'd probably see an engine of such strength blunder to you. It's too "mistake-free" for me to estimate it that low.
Jul 11, 2012
0
#212
FirebrandX wrote:
Elubas wrote:
So FirebrandX is one of the best CC players in the world?
Not the best (yet) since I'm still climbing the ladder, though what got me started in ICCF was I personally headed a vote-chess team that defeated ICCF-GM and 15th World Champion Gert Jan Timmerman (rated #1 at the time with over 2700 ICCF). Through my analysis and charge, we eventually outplayed Gert in a technically drawable endgame. I was given the MVP status from the game since it was my analysis that took Gert down, and subesequently a free year's subscription to the site at the time. It was then that I knew I had a natural OCD-knack for correspondence 'centaur' chess.
Is this the game you have in mind: http://www.chessgames.com/perl/chessgame?gid=1464744 ?
FirebrandX wrote:
LegoPirateSenior wrote:
Is this the game you have in mind: http://www.chessgames.com/perl/chessgame?gid=1464744 ?
That's the one. That's where I got the free MVP year subscription for my contributions in that game. I remember closing that game out to a win was a monumental effort of analysis mapping. I recall as a I said before, I'd determined Gert could have drawn the endgame with 'perfect' play, but we got lucky in that he simply was unable to keep up with the branching sequences until he eventually lost the position. It was after that game that I joined the ICCF.
I've learned quite a lot since that game. Each year since on ICCF has been a learning process for me.
Edit: Going over my old notes from that game, I have to apologize for a bit of egocentric false memory syndrome. There were in fact many others that contributed quality work in the game. I just remember having to do a lot of "campaigning" to get everyone to vote for the best lines we analyzed. Often times, people would spam the discussion with weak computer moves, and I'd have to waste a bunch of posts and energy busting them. We learned fairly quickly that if we ignored the bad analysis, casual voters would take it for gospel truth and vote in line with it. So it became a crowd-control issue for much of the posting of analysis.
At any rate, that's why I like the vote chess system even better here on chess.com. Less casual voters and a closer-knit group to bounce ideas around with.
I think casual players are actually more difficult to play against. They are less predictable.
With a good player you already have a notion of what move he will play. It is usually the best for that position. Good moves, at least in the opening, are easy to predict.
pfren wrote:
fburton wrote:
I'd be kind of surprised if humans do better in correspondence chess against a computer than in a 'normal' timed (but not rapid/blitz) game, because it can spend every second calculating (unless restricted in some way) whereas a human has the matter of life to attend to.
1. In longer time controls, humans can avoid most major tactical oversights.
2. Computers cannot acquire positional grasp, no matter how long the think. If they can pick between a few moves of about equal value according to their algorithms, there is a serious chance picking up an inferior one.
Currently in ICCF word championships the usage of computers is allowed. While there have been a few weak players who have made good results, the vast majority of champions and top rated players are/were also very strong OTB.
Point 2 is wrong.
Modern engines have lots of positional understanding built in that guides the calculation. This is how they jumped so quickly from 2500 to 3000+
Haiku575 wrote:
That's true, especially with the chess engine Komodo.
I don't think anyone underates it.
It's considered the best way to play the Scotch Game.
Jul 13, 2012
0
#217
Engines don't have any real positional understanding, or any understanding at all. They fake it with evaluation functions.
browni3141 wrote:
Engines don't have any real positional understanding, or any understanding at all. They fake it with evaluation functions.
Humans positional understanding is also based on evaluation, just they don't assign numerical values to their evaluations.
In order to claim that engines don't have positional understanding you have to define what does it mean first.
browni3141 wrote:
Engines don't have any real positional understanding, or any understanding at all. They fake it with evaluation functions.
You call this fake positional understanding, but I think "substitute" might be a better fit. Obviously they don't think like a human does, but if the evaluation function allows them to usually value the right stuff, then it can often have a similar effect. It obviously doesn't have as much (potential) depth as a human's positional understanding, but I still think it's fair to qualify it as some sort of positional understanding nonetheless.
browni3141 wrote:
Engines don't have any real positional understanding, or any understanding at all. They fake it with evaluation functions.
I wonder if human playing with unknown opponent will be able to say with certainty if it's another human or an engine (kind of Turing test for chess). If not then judging by observable behavior engines do have positional understanding.
While I am not currently aware of how big of a difference cores make on ELO ratings, I find it highly unlikely that running Houdini on one core would cause it to play at a 2300 level. What I can say is that at 4 cores on a 64-bit processor, Houdini reaches a speed of around 3-4 million nodes per second. On a single core it usually reaches 700 thousand.