< Earlier Kibitzing · PAGE 3 OF 3 ·
|Dec-05-06|| ||WannaBe: Yeah, I can see it now... 1. d4 (20 seconds later) 1... e5 Black mates in 83 moves.|
|Dec-05-06|| ||TefthePersian: We don't really know how many atoms there are in the universe. But when they say, "particles in the observable universe" what do they mean? Are they including quarks? Matter and energy convert from one to the other regularly, so...the number always changes. Even estimating it is very, very complicated.|
|Dec-05-06|| ||TefthePersian: And dark matter makes the situation more complicated. It (probably) does not react electromagnetically so we can't see it. How do we then count the amount of particles in it? Exactly.|
|Dec-05-06|| ||rover: <TefthePersian> Dark energy particles must be heavy, otherwise we would be able to observe their production in particle accelerators. So the number of dark matter particles is much lower than the number of baryons in the universe.|
|Dec-05-06|| ||TefthePersian: <rover> "Dark energy particles must be heavy, otherwise we would be able to observe their production in particle accelerators. So the number of dark matter particles is much lower than the number of baryons in the universe."|
The amount of dark matter is far, far more common than vanilla matter, and the composition of dark matter is unknown. So...
I don't know what you're saying about particle accelerators and dark matter creation.
|Dec-05-06|| ||rover: What I'm saying that various particles get created in high energy collisions in particle accelerators. Part of the energy gets converted into matter. The reason why the top quark was the last one to be discovered becouse it is the most massive one so it needed the most massive collisions to be created.|
Now if dark matter particles (no idea why I wrote dark energy in my last post) were light they would be created in these collisions. Even if it doesn't get detected after the collision (quite likely), there would be some energy missing, so the creation of dark matter would be detected.
Now the fact is that this phenomenon is not detected, so the conclusion is that dark matter particles are heavy. So despite the fact that dark matter is more massive in the universe than baryonic matter there is probably less number of them.
|Dec-05-06|| ||TefthePersian: <rover> "What I'm saying that various particles get created in high energy collisions in particle accelerators."|
Yes, but these are all normal matter/radiation. (Not dark matter/dark energy *Whatever the hell dark energy is*)
"Part of the energy gets converted into matter."
I don't know many particle acceleration experiments, physics is just an interest of mine, but I have no idea how this means that dark matter would be created from collision. The elecromagnetic force is the primary thing in particle collision, but also the strong force. We know that dark matter probably doesn't interact via the electromagnetic force, because uh...we can't see it.
"The reason why the top quark was the last one to be discovered becouse it is the most massive one so it needed the most massive collisions to be created."
I again don't understand the link between normal matter and dark matter in particle collision of normal matter and normal matter.
"Now if dark matter particles (no idea why I wrote dark energy in my last post) were light they would be created in these collisions."
If you could explain this, I would be happy.
"Even if it doesn't get detected after the collision (quite likely), there would be some energy missing, so the creation of dark matter would be detected."
We know that dark matter must interact via gravity, and it seems to fill in the the massive gap required of the universe's mass, so I don't mind saying that it's heavy. But I don't know the situation we're talking about here. What is being collided into what that would probably create dark matter if dark matter was "light"?
"Now the fact is that this phenomenon is not detected, so the conclusion is that dark matter particles are heavy. So despite the fact that dark matter is more massive in the universe than baryonic matter there is probably less number of them."
The universe is composed of roughly 3/4ths dark matter. It would have to be really, really massive to account for that difference in percentage in ...uh...dark-atom size. I think more likely is that Tal is inside each one of the dark atoms, and he makes the force of gravity act like a real man.
|Jan-27-07|| ||Dionyseus: Rybka has just dismanted Deep Fritz 10 in a 6 game classical time control match by a score of 5.5-0.5, the machines that were used were identical to the one used in the Kramnik - Deep Fritz 10 match:|
You can download the pgn and the commentaries here:
|Feb-01-07|| ||micartouse: A brief timeline of anti-computer strategy for world class players:|
20 years ago - Play some crazy gambits and demolish the computer every game. Shock all the nerdy computer scientists in the room.
15 years ago - Take it safely into the endgame where its calculating can't match human knowledge and intuition. Laugh at its pointless moves. Win most the games.
10 years ago - Play some hypermodern opening to confuse it strategically and avoid direct confrontation. Be careful and win with a 1 game lead.
5 years ago - Block up the position to avoid all tactics. You'll probably lose a game, but maybe you can win one by taking advantage of the horizon effect. Draw the match.
Now - Play reputable solid openings and make the best possible moves. Prepare everything deeply, and never make a tactical mistake. If you're lucky, you'll get some 70 move draws. Fool some gullible sponsor into thinking you have a chance.
|Apr-21-07|| ||Voxation: Out of Fritz 9, Deep Fritz and X3D Fritz, which is the strongest and which is the weakest?|
|Feb-24-08|| ||The Rocket: Well the strongest version is deep fritz, then comes x3dfritz(but i dont know if it really exists anymore,?), and lastly fritz 9 heres a site which provides computer rankings:|
|Feb-07-09|| ||WhiteRook48: however the strongest computer is Rybka|
|Sep-21-12|| ||Conrad93: We can never be certain about how many particles exist in the universe, but we can "estimate" the number.|
|Nov-28-13|| ||whiteshark: testing <Deep Fritz 14 x64>:|
|Dec-04-13|| ||numbersguy70: Deep Fritz 14 definitely calculates quicker than Fritz 12 on my cheapo PC, but comparing exact same positions against Fritz 12, 14 reaches more drawish numerical evaluations every time. Kind of annoying, as I'm used to 1.0 to 1.5 being clearly winning for white, and now have to adjust to something less (still figuring it out).|
|Jan-23-14|| ||SChesshevsky: I noticed this game
B Alterman vs Deep Fritz, 2000
where Fritz didn't seem to play very well. Yet in a couple of years it looks like he could beat and split a match with Kramnik and by 2006 appeared easily Kramnik's equal and possibly better.
Any computer chess fans know the story on this seemingly sharp development by Fritz?
|Jan-23-14|| ||mrbasso: Deep Fritz 14 is a marketing stunt.
It is not related to Fritz 12 in any way. It is Pandix with a new name.
Computerchess programming made a big leap forward between 2000 and 2006 and the hardware got much better as well. The improvement of Fritz (Quest) was in reality not big compared to the improvement of other engines in this period.
|Jan-03-15|| ||The Rocket: Any computer chess fans know the story on this seemingly sharp development by Fritz?|
Version 9 to 10 they changed the algorithms for the Kramnik match, while significantly improving the tactical search.
Version 11-13 used Old rybka code intermixed with other stuff and is a completely different engine from version 9-10. Which is why the ponderhits, eval and material value is completely revised. DF10 and DF11 belong to different "familys" of engines in ponderhits. DF11-13 is similiar to strelka/old rybka.
Versions 14 is the Pandix engine, after Fritz author Frans Morsch retired and so did the program.
|Jan-03-15|| ||AylerKupp: <Fritz versions, ratings, rankings (part 1 of 3)>|
For those interested in Fritz's version and release history, see http://en.wikipedia.org/wiki/Fritz_(chess).
To get an idea of the relative strength of various Fritz versions over time, here are the ratings and rankings of selected versions according to the results of the CCRL 40/20 tournaments. CCRL is not consistent in indicating whether the engine was running on a 64-bit vs. a 32-bit computer so assume a 64-bit computer:
Feb-04-11: Version, Cores, Rating, Rank
11 4 3096 6
12 4 3087 6
10.1 4 2993 16
10 2 2953 19
10 4 2933 21
9 1 2843 29
8 2 2833 30
Jan-29-12: Version, Cores, Rating, Rank
11 4 3095 11
12 4 3087 11
13 1 3067 15
10.1 4 2992 21
10 2 2951 23
10 4 2931 26
9 1 2841 36
8 2 2831 37
Jan-14-13: Version, Cores, Rating, Rank
13 4 3052 12
11 4 3003 15
12 4 2998 16
10.1 4 2898 26
10 2 2855 29
10 4 2836 30
9 1 2746 43
8 2 2736 45
Jan-05-14: Version, Cores, Rating, Rank
14 4 3064 11
13 4 3049 15
11 4 2999 18
12 4 2994 18
10.1 4 2894 28
10 2 2851 34
10 4 2832 36
9 1 2743 50
8 2 2734 53
Jan-03-15: Version, Cores, Rating, Rank
14 4 3087 15
13 4 3048 18
11 4 2998 22
12 4 2992 22
10.1 4 2893 41
10 2 2851 47
10 4 2831 50
9 1 2743 66
8 2 2733 66
For information on other engines see http://computerchess.org.uk/ccrl/40....
|Jan-03-15|| ||AylerKupp: <Fritz versions, ratings, rankings (part 2 of 3)>|
And here is the evolution of each Fritz version over time:
Version 8: Cores, Date, Rating, Rank
2 Feb-04-11 2833 30
2 Jan-29-12 2831 37
2 Jan-14-13 2736 45
2 Jan-05-14 2734 53
2 Jan-03-15 2733 66
Version 9: Cores, Date, Rating, Rank
1 Feb-04-11 2843 29
1 Jan-29-12 2841 36
1 Jan-14-13 2746 43
1 Jan-05-14 2743 50
1 Jan-03-15 2743 66
Version 10: Cores, Date, Rating, Rank
2 Feb-04-11 2953 19
4 Feb-04-11 2933 21
2 Jan-29-12 2951 23
4 Jan-29-12 2931 26
2 Jan-14-13 2855 29
4 Jan-14-13 2836 30
2 Jan-05-14 2851 34
4 Jan-05-14 2832 36
2 Jan-03-15 2851 47
4 Jan-03-15 2831 50
Version 10.1: Cores, Date, Rating, Rank
4 Feb-04-11 2993 16
4 Jan-29-12 2992 21
4 Jan-14-13 2898 26
4 Jan-05-14 2894 28
4 Jan-03-15 2893 41
Version 11: Cores, Date, Rating, Rank
4 Feb-04-11 3096 6
4 Jan-29-12 3095 11
4 Jan-14-13 3003 15
4 Jan-05-14 2999 18
4 Jan-03-15 2998 22
Version 12: Cores, Date, Rating, Rank
4 Feb-04-11 3087 6
4 Jan-29-12 3087 11
4 Jan-14-13 2998 16
4 Jan-05-14 2994 18
4 Jan-03-15 2992 22
Version 13: Cores, Date, Rating, Rank
1 Jan-29-12 3067 15
4 Jan-14-13 3052 12
4 Jan-05-14 3049 15
4 Jan-03-15 3048 18
Version 14: Cores, Date, Rating, Rank
4 Jan-05-14 3064 11
4 Jan-03-15 3087 15
|Jan-03-15|| ||AylerKupp: <Fritz versions, ratings, rankings (part 3 of 3)>|
And here are some observations:
1. Version 14 (based on the Pandix engine) is the strongest Fritz version ever. So, while it might be a "marketing stunt" as indicated by <mrbasso>, at least its customers are getting a slightly stronger engine.
2. Version 11 has always been slightly better than version 12. So, in contrast to Version 14, I would suspect that version 12 customers would not be too pleased to have paid for an update and received a slightly weaker version.
3. Version 10 with 2 cores consistently outperformed version 10 with 4 cores. Again, I would imagine that any version 10 customers who upgraded their computer from 2 to 4 cores would not be too pleased to find out that their engine's strength went down as a result.
4. Version 10.1 finally got its multi-core performance right, as with 4 cores it outperformed version 10 with either 2 or 4 cores. I don't know if version 10.1 with 2 cores outperformed version 10.1 with 4 cores (no data for the 2-core version was provided), but I suspect that the Fritz developers fixed their scalability problem with version 10.1 so that the 4-core version would outperform the 2-core version.
5. The latest version of Fritz has been going slightly backward relative to other engines. The highest ranked versions (both version 11 and 12, 4-core) on Feb-04-11 were ranked #6 but the highest ranked version on succeeding years were ranked #11 (both version 11 and 12, 4-core), #12 (version 13, 4 core), #14 (version 14, 4-core), and #14 (also version 14, 4-core).
6. As should be expected as a result of lowered rankings, the rating of each version has been decreasing over time, as a result of facing increasingly stronger competition. The exception is version 14 whose rating increased by almost 20 Elo rating points between 2014 and 2015.
Happy New Year everyone!
|Jan-03-15|| ||The Rocket: <Version 11 has always been slightly better than version 12. >|
The margin of error is around 10-15 elo in testing engines for rating lists. Versions 11 and 12 are thus pretty much tied. They are none the less unrelated to the previous Fritz versions. Different engines.
I have the 10.1 version and it doesn't agree with DF early continuations in the QGA game which Kramnik blundered to mate. And in the final sicilian game - the Knight mirroring maneuver - move 19. Nb1 which, Chessbase praised, it does not rate as a candidate move and elects to play something else.
In reality Kramnik faced a single processing DF, which could account for why it played remarkably shallow in the QGA game. I still found it dubious from the Chessbase team to employ standard opening lines against Kramnik, who had a version of Fritz 10 at home for his personal use in preperation for the match, and could carefully set up his own variations.
|Jan-03-15|| ||AylerKupp: <Correction:> The CCRL games were played at a time control of 40/40, not 40/20 as I indicated. They are also repeating; an engine gets 40 minutes for its first 40 moves and another 40 minutes for its next 40 moves. This way a possible time trouble scramble is somewhat reduced.|
<The Rocket> I don't know why you say that the margin of error is around 10-15 Elo rating points in testing engines. The CCRL ratings are computed using the BayesElo freeware program and they provide the 95% confidence interval, so the top-ranked Komodo 8 64-bit 4CPU engine is ranked at 3303 with the 95% confidence interval being 17, meaning that there is a 5% probability that Komodo 8's true rating is below (3303 -17) = 3286 or above (3303 + 17) = 3320. But this is <not> the same as a margin of error, and both depend on the actual number of games played (which, to make matters more complicated, varies from engine to engine). So I take the easy way out and assume that the CCRL ratings are "accurate".
But I do have to admit that I am perplexed as to how CCRL calculates their rankings for their Complete List. For example, there are 6 engines ranked #3 in their latest list with ratings from 3273 (Houdini 4 64-bit 4CPU) to 3236 (Komodo TCEC 64bit 4CPU), a 37-point Elo rating differential. Fire 4 64-bit 4CPU is rated at 3226 and ranked #4, a 10-point Elo rating differential between it and Komodo TCEC 64bit 4CPU. So why is Fire 4 grouped with the 7 engines ranked #4 and not with the 6 engines ranked #3?
I think that all Fritz versions are "related" in the sense that they all go by the name Fritz, regardless of whether they contain any common code. As I said earlier, when one spends money on an engine upgrade, one expects (perhaps unreasonably) that the upgraded engine is stronger than the previous engine. If that is not the case, then I think that the engine vendors would be guilty of misrepresenting their product. Of course, the engine buyers would be remiss for not checking whether the upgraded engine is in fact stronger than the previous engine, but that's a different issue.
And I don't think that Kramnik faced a single processor Deep Fritz, at least not according to these stories by Chessbase: http://en.chessbase.com/post/deep-f... and http://en.chessbase.com/post/kramni.... In my CCRL records for Feb-04-2011 (the earliest I have) Deep Fritz 10 (2CPU) is rated at 2953, Deep Fritz 10 (4 CPU) is rated at 2933, and Fritz 10 (1 CPU) is rated at 2884, considerably lower than both Deep Fritz's. Since the Deep Fritz's were released by the time of the Kramnik match, for marketing reasons I would think that Chessbase would have insisted that the Deep Fritz multi-core version be used rather than the weaker single core Fritz version in order to maximize its chances of winning. Whether that would have made any difference in the match's outcome is anyone's guess.
|Jan-03-15|| ||The Rocket: Chessbase did not realise at the time of writing that their 10 version wasn't functioning correctly with more than 2 cores and bugged. Given that they attemped more than 2 cores, it's very likely that the bugg resulted in a performance decrease corresponding to 1 cpu, at most 2 CPU. I don't know the exact technical details, but judging from some of the games from the Kramnik match. Several DF continuations look quite dubious and weren't selected by my mere intel core duo laptop when I analysed the games with DF 10.1, in what supposedly correspond to 2 cpu.
Version 10.1 corrected all of this. It's a shame they dismantled it.|
As to version 12, there was no promise of elo gain in it's dvd. The interview with Fransch Morsch suggests improvements are being made, but not explicitly.
They more than made up for it with version 13 though, which is signficantly stronger. And then Frans Morsch retired.
|Jan-03-15|| ||The Rocket: Kramnik faced version 10, and the offical release of Deep fritz 10 was indeed the buggy one. 10.1 uppdate was provided online by chessbase.|
< Earlier Kibitzing · PAGE 3 OF 3 ·