Yeah. We'll see the chessbase interview in about 43 days.
And chessbase will also include a refutation of the King's Indian in the same article.
My nephew beat HOUDINI?!?!
Sort:
So, how difficult is Houdini compared to the Hard level on the computer at Chess.com?
naturalproduct wrote:
So, how difficult is Houdini compared to the Hard level on the computer at Chess.com?
Houdini would win 1000 out of 1000 games.
Same with computer impossible, Houdini would win 1000 out of 1000
Cnacnel wrote:
naturalproduct wrote:
So, how difficult is Houdini compared to the Hard level on the computer at Chess.com?
Houdini would win 1000 out of 1000 games.
Same with computer impossible, Houdini would win 1000 out of 1000
wow.
naturalproduct wrote:
Cnacnel wrote:
naturalproduct wrote:
So, how difficult is Houdini compared to the Hard level on the computer at Chess.com?
Houdini would win 1000 out of 1000 games.
Same with computer impossible, Houdini would win 1000 out of 1000
wow.
That is a joke match. That is like me vs. Superman in a boxing fight. No contest.
Feb 17, 2013
0
#127
Here's another way of looking at the situation.
www.chessmetrics.com/cm/CM2/AgeLists.asp
From the above link, you can see roughly the top 12 twelve-year-olds in recent chess history. Even if the kid was on par with these elite dozen, and even accounting for the extra time that the kid took during the game, he would only have about a 10 percent chance (at best) of winning a game against Houdini. Possible, but the odds are against it.
I've beaten the hard level here on Chess.com....ONCE 
....Rybka, Houdini, etc... NO WAY!!!
EscherehcsE wrote:
Here's another way of looking at the situation.
www.chessmetrics.com/cm/CM2/AgeLists.asp
From the above link, you can see roughly the top 12 twelve-year-olds in recent chess history. Even if the kid was on par with these elite dozen, and even accounting for the extra time that the kid took during the game, he would only have about a 10 percent chance (at best) of winning a game against Houdini. Possible, but the odds are against it.
Chessmetrics is its own rating system, it's not actual ratings.
Computer chess ratings are also not FIDE ratings.
But even pretending they're all the same thing, you somehow got the percent wrong. Although the rating formula shouldn't be accurate this many standard deviations out, the math shows it would be less than 1 in 100.
Feb 17, 2013
0
#130
waffllemaster wrote:
EscherehcsE wrote:
Here's another way of looking at the situation.
www.chessmetrics.com/cm/CM2/AgeLists.asp
From the above link, you can see roughly the top 12 twelve-year-olds in recent chess history. Even if the kid was on par with these elite dozen, and even accounting for the extra time that the kid took during the game, he would only have about a 10 percent chance (at best) of winning a game against Houdini. Possible, but the odds are against it.
Chessmetrics is its own rating system, it's not actual ratings.
Computer chess ratings are also not FIDE ratings.
But even pretending they're all the same thing, you somehow got the percent wrong. Although the rating formula shouldn't be accurate this many standard deviations out, the math shows it would be less than 1 in 100.
Yeah, you could well be right about the inaccuracy of the standard formula at large deltas. I was giving big benefits of the doubt to the numbers. 3100 for the engine to account for 1 core and reduced rating against humans, 2900 after factoring the extra thinking time for the kid (probably too generous, and of course, ponder would have to be off), and 2500 for the elite dozen people. That's a delta of 400 (approx 9 percent.)
I did ignore the difference between human standard ratings and human blitz ratings, so maybe the 9 percent should be a little lower.
EscherehcsE wrote:
waffllemaster wrote:
EscherehcsE wrote:
Here's another way of looking at the situation.
www.chessmetrics.com/cm/CM2/AgeLists.asp
From the above link, you can see roughly the top 12 twelve-year-olds in recent chess history. Even if the kid was on par with these elite dozen, and even accounting for the extra time that the kid took during the game, he would only have about a 10 percent chance (at best) of winning a game against Houdini. Possible, but the odds are against it.
Chessmetrics is its own rating system, it's not actual ratings.
Computer chess ratings are also not FIDE ratings.
But even pretending they're all the same thing, you somehow got the percent wrong. Although the rating formula shouldn't be accurate this many standard deviations out, the math shows it would be less than 1 in 100.
Yeah, you could well be right about the inaccuracy of the standard formula at large deltas. I was giving big benefits of the doubt to the numbers. 3100 for the engine to account for 1 core and reduced rating against humans, 2900 after factoring the extra thinking time for the kid (probably too generous, and of course, ponder would have to be off), and 2500 for the elite dozen people. That's a delta of 400 (approx 9 percent.)
I did ignore the difference between human standard ratings and human blitz ratings, so maybe the 9 percent should be a little lower.
Yeah, giving the kid somebig breaks there :)
Feb 17, 2013
0
#132
I take issue with the title. The OP states that his nephew beat Houdini but the title has two question marks in it. What does "?!?!" mean anyway? my guess is "doubly dubious".
With what a ball bat after he lost ?
This thread reminds me how my nephew once beat Houdini. Aaah, good times ...
Haha, I was thinking of replying to this last night and it would have been the second reply. I guess I'm not totally surprised this one got so much attention!
The dog may be real.
hoynck wrote:
(No better name for such a 'MG' than Adam, I suppose).
And what a fitting name for a Messiah that is, since Adam means nothing more than 'Man'! That's deep.
you know what? I think the OP has cowered from this thread. No more fuel to backlash against the accusations.
Here are results of a quick analysis of that game, using Houdini 1.5a, 512MB hash, 10s/move, at least 14 plies, analysis starting from move 26 (per pfren's diagnosis of 25 book moves). CPU: 2.66 GHz i7, single processor used.
For those unfamiliar with top-3 methodology for detecting engine use, top-1 is the percentage of positions where the 1st egine choice was played, top-2 is the percentage of positions where either the 1st or 2nd engine choice was played, et.c.
Adam Top1 match: 24/36 = 66.67 %
Adam Top2 match: 33/36 = 91.67 %
Adam Top3 match: 35/36 = 97.22 %
The 10s/move analysis time was based on the OP's statement of having given Houdini 5 minutes for the entire game. This setting is not really correct, but not much can be done without knowing the exact time that the opponents took for each move. I have only the free version of Houdini, so that's another source of inaccuracy.
Given the inaccuracies and small sample, there's no way to draw firm conclusions, however, the numbers are consistent with what might be expected from analyzing a Houdini-Houdini match.
Repeated analysis at minimum 40s/move, at least 16 plies deep:
Adam Top1 match: 24/36 = 66.67 %
Adam Top2 match: 31/36 = 86.11 %
Adam Top3 match: 36/36 = 100.00 %
Adam Top4 match: 36/36 = 100.00 %
Houdini 3 x64 Top1 match: 20/35 = 57.14 %
Houdini 3 x64 Top2 match: 27/35 = 77.14 %
Houdini 3 x64 Top3 match: 29/35 = 82.86 %
Houdini 3 x64 Top4 match: 34/35 = 97.14 %
If the OP comes back and says that this was all a big joke and the game shows Houdini playing against itself, with one engine instance given substantially more time than the other (note how the increased analysis time does not affect Adam's rates a lot, but substantially decreases those of Houdini), I will not be surprised.
OTOH, maybe we'll see that interview with ChessBase...