ouachita wrote:Jeffrey,
There are many daily tests with innumerable HW/SW/settings performed by many testers at numerous sites to review and consider. Nevertheless, on balance, neither the STC or LTC top tier hierarchy has changed over the past month or so. Basically, H4 is King of Blitz, K is King of LTC, and SF continues to get better, perhaps in both.
Not surprisingly, there are many different views on these topics.
SF vs Houdini, wins SF.
When there are many engines under test, houdini is best because its overall performance.
All we need is to change the default contempt of SF to solve the problem. But Marco doesn't want to. In head to head SF is at least on par with Houdini, yet Houdini is significantly above in rating lists. The difference is that Houdini destroys the weaker engines more than SF destroys them, because it doesn't accept 3-fold repetitions where SF would.
I _guarantee_ you that the current dev version of SF+syzygy+contempt=20cp would beat Houdini (a fortiori Komodo) in any rating list condition (except perhaps super fast ones like Stefan Pohl's one).
I do agree, still there is another way to see the same: Instead asking Marco on SF default settings, we can ask Houdart to modify H4 default settings.
Or, ask testers to set houidini contempt=0 in their tests OR do not change default houdini settings when playing a match against a strong opponent.
If not, we are favoring an engine by changing their settings according what is going to play.
OR we can do nothing, leaving houdini with #1 rank (we know why) and using SF and Komodo to find the best move for analysis.
ouachita wrote:Jeffrey,
There are many daily tests with innumerable HW/SW/settings performed by many testers at numerous sites to review and consider. Nevertheless, on balance, neither the STC or LTC top tier hierarchy has changed over the past month or so. Basically, H4 is King of Blitz, K is King of LTC, and SF continues to get better, perhaps in both.
Not surprisingly, there are many different views on these topics.
SF vs Houdini, wins SF.
When there are many engines under test, houdini is best because its overall performance.
I am not sure that SF wins vs Houdini
I saw the following result in one of the rating lists (CCRL FRC 40/4 rating list)
Houdini4 179.5 − Stockfish DD 120.5
(+121−62=117)
The latest developement version of stockfish is better but my guess is that it is not good enough to beat houdini4 in that list(and I do not talk about rating but about winning direct match of 300 games).
Uri Blass wrote:do not talk about rating but about winning direct match of 300 games.
I agree; I really have very little interest in the many STC matches (except for mine) or matches involving many dozens of engines, but am mainly interested in direct matches between "the Big 3", defined as H, K and SF, a/o 12/27/13.
ouachita wrote:Jeffrey,
There are many daily tests with innumerable HW/SW/settings performed by many testers at numerous sites to review and consider. Nevertheless, on balance, neither the STC or LTC top tier hierarchy has changed over the past month or so. Basically, H4 is King of Blitz, K is King of LTC, and SF continues to get better, perhaps in both.
Not surprisingly, there are many different views on these topics.
SF vs Houdini, wins SF.
When there are many engines under test, houdini is best because its overall performance.
I am not sure that SF wins vs Houdini
I saw the following result in one of the rating lists (CCRL FRC 40/4 rating list)
Houdini4 179.5 − Stockfish DD 120.5
(+121−62=117)
The latest developement version of stockfish is better but my guess is that it is not good enough to beat houdini4 in that list(and I do not talk about rating but about winning direct match of 300 games).
But that is FRC.
Probably the evaluation and search parameters of Stockfish are tuned mostly specifically to normal chess, while Houdini's to more general chess rules, or maybe Houdini has some specific FRC-related parameters, etc.
ouachita wrote:Jeffrey,
There are many daily tests with innumerable HW/SW/settings performed by many testers at numerous sites to review and consider. Nevertheless, on balance, neither the STC or LTC top tier hierarchy has changed over the past month or so. Basically, H4 is King of Blitz, K is King of LTC, and SF continues to get better, perhaps in both.
Not surprisingly, there are many different views on these topics.
SF vs Houdini, wins SF.
When there are many engines under test, houdini is best because its overall performance.
I am not sure that SF wins vs Houdini
I saw the following result in one of the rating lists (CCRL FRC 40/4 rating list)
Houdini4 179.5 − Stockfish DD 120.5
(+121−62=117)
The latest developement version of stockfish is better but my guess is that it is not good enough to beat houdini4 in that list(and I do not talk about rating but about winning direct match of 300 games).
I am sure.
Here is the Head to Head match results of Stockfish DD at 40/40. It wins every match. This is my findings also. And there are stronger version of stockfish out now. When the list catch up to reality, I guess then you will be sure.
"The worst thing that can happen to a forum is a running wild attacking moderator(HGM) who is not corrected by the community." - Ed Schröder
But my words like silent raindrops fell. And echoed in the wells of silence.
Well, if a 34-game match (+9−5=20) is a proof to you...
We are talking about 189 games against the best programs just on this list. It wins every match and is 4 elo behind Houdini 4. There are stronger versions of stockfish out now. Houdini 4 is the same houdini 4. It is not hard to reason that stockfish if not already the best engine. It soon will be.
"The worst thing that can happen to a forum is a running wild attacking moderator(HGM) who is not corrected by the community." - Ed Schröder
But my words like silent raindrops fell. And echoed in the wells of silence.
If you're looking at contests between the top three only, I'd put Stockfish slightly ahead of Houdini, with Komodo in third spot.
However, Houdini has a better record against engines outside the top three, so when all is said and done, Houdini might have a slightly better overall record than Stockfish. I'd still put Komodo behind both.
Graham Banks wrote:If you're looking at contests between the top three only, I'd put Stockfish slightly ahead of Houdini, with Komodo in third spot.
However, Houdini has a better record against engines outside the top three, so when all is said and done, Houdini might have a slightly better overall record than Stockfish. I'd still put Komodo behind both.
The next TCEC should be interesting.
Yes, that is what I have been hearing: Houdini is higher rated not because it is a stronger engine but because of contempt settings it beats up the weaker engines more effectively than Stockfish and Komodo.