Milos wrote:That's simply because you and all ppl who basically are fans supporting a particular engine (usually SF or Komodo) stubbornly refuse to test Houdinis with proper settings for these kind of matches i.e. with contempt 0.
Milos,
I truly do not have a bias for a particular engine but rather am simply searching for the strongest. It seems like you are the one defending H4's LTC performance inspite of the numerous results at TCEC and here showing it's slightly weaker than SF DD and KTCECr.
Give me the settings you recommend for H4, SF and KTCEC and I'll run my own tests at 75 min and we'll see . . .
P.S. - I suppose I could have used General Custer at Little Big Horn but I would have been seen as showing bias toward the prevailing party, although I'm neither Cheyenne, Lakota, nor Arapaho.
Milos wrote:That's simply because you and all ppl who basically are fans supporting a particular engine (usually SF or Komodo) stubbornly refuse to test Houdinis with proper settings for these kind of matches i.e. with contempt 0.
Milos,
I truly do not have a bias for a particular engine but rather am simply searching for the strongest. It seems like you are the one defending H4's LTC performance inspite of the numerous results at TCEC and here showing it's slightly weaker than SF DD and KTCECr.
Give me the settings you recommend for H4, SF and KTCEC and I'll run my own tests at 75 min and we'll see . . .
P.S. - I suppose I could have used General Custer at Little Big Horn but I would have been seen as showing bias toward the prevailing party, although I'm neither Cheyenne, Lakota, nor Arapaho.
I haven't tested H4 yet, but what I see is huge discrepancy between tests I currently run with H3 C0 and what AH is running. And somehow I believe my test more (despite contempt issue, in his tests H is particularly bad, much worse than in any other test, I'll maybe even repeat some of them using his positions because something there looks really fishy).
Regarding setting you've been given them numerous times (H4 contemp=0, K drawscore=0, SF all default) but you haven't started a tournament. Somehow I have a feeling you don't really intend to run it, just talking about it.
P.S. Indians and cowboys was just a joke, I didn't think you'd had any real bias in that matter, it just sounded funny to me that you used Zulu-British example .
How about you point me out to source code, or even just executable of that Ordo version you used for your famous "predictions"??
Somehow, I feel both of you and Miguel will again run away.
Milos wrote:... (despite contempt issue, in his tests H is particularly bad, much worse than in any other test, I'll maybe even repeat some of them using his positions because something there looks really fishy).
Is funny to see that when you don't have any reasoning to support your ideas, you accuse me of "something fishy" . This tells all about you. Fortunetly, people in the forum only needs to read some of my posts in the last years to know who I am, and read some of your posts to know who you are ...
Komodo 6TCEC is about +16 Elo above Komodo 6 at CCRL 40/40 (and at CEGT 40/20)
So when Komodo 6 still losses with 60% margin against H4 at CCRL 40/40 than it would be very, very doubtful that Komodo 6TCEC can get 50% at same time control conditions http://www.computerchess.org.uk/ccrl/40 ... 4-bit_4CPU
My best guess is that at TC> 240 for a game with 1 core (for instance 40/120 CEGT conditions) Komodo 6 TCEC will be equal to Houdini 4
They are testing it right now so we have to be patient but I think this will be the best guess.
(Komodo 6 got 51% against Houdini 3 in 50 games at CEGT 40/120)
Adam Hair wrote:You made a blanket statement about Ordo, implying that the program in general is no good.
Adam, perhaps this debate should moved off this thread?
Actually, there is no place for it to logically go. But you are right. It does not belong here, and so I will remove my response and send it via pm to Milos.