in the last weekend I checked the new Fritz 11 engine. I must say that I was really impressed.
The improvement over Fritz 10 were great!
So Vas, you have now a much more challenge to improve and put out Rybka 3 as soon as possible...
What are the impressions of the community about the new Fritz11.
New, dramatically lobotomized engine. Weaker than the previous version on the same hardware, provided your hardware is good enough. A quad should do.
And they're still force-feeding the GUI and Playchess with it.
And the database is most likely a subset of what I already have.
And there's no download version.
We don't know anything about Deep Fritz 11.
Fritz 11 : 2914 19 19 820 58.0 % 2858 36.7 %
Fritz 10 : 2839 11 11 2948 50.0 % 2839 29.8 %
So there is a differnce of 75 points.
> You should compare single Fritz 11 vs. Single Fritz 10.
Why? I have a dual, I have no use for single Fritz 10. Being limited to one processor obviously hurts F11, but there's nothing I can do about it. CB could have just as easily released Deep Fritz 11 now: how else could they already know how strong it is?
>We don't know anything about Deep Fritz 11.
We know it's going to be released next March.
You can analyze chess games with more than one engine and give Fritz11 one processor and give Rybka another processor.
Uri
I'd like to define 'use for Fritz 11' as something that could possibly be construed as being worth 50€.
Fritz 11 : 2914 19 19 820 58.0 % 2858 36.7 %
Fritz 10 : 2839 11 11 2948 50.0 % 2839 29.8 %
So there is a differnce of 75 points.
Anycase ,it's not stronger than Kramnik who is playing elo2920 on Tal memorial.
:)
I think that it's definitely stronger. First, the CEGT list is deflated compared with the SSDF list, which is scaled to the FIDE elos. Second, the SSDF list is for Athlon 1.2 GHz. If you're dealing with an Athlon 2.4 GHz or equivalent or stronger, you're talking about another 60-point elo increase from that. On the other hand, rating increases between Chessbase versions tend to be smaller on the SSDF list, and this is probably closer to reality: I doubt that Fritz 11 is more than 50 elo stronger than Fritz 10 on the new SSDF list (compare Shredder 10 with Shredder 9, which is supposed to have an 80-point elo difference), and the difference is probably less. Even so, taking everything into account, the actual elo rating of Fritz 11 on a fast single-core system is on the order of 100 elo stronger than the value given for Fritz 10 on the SSDF list. This brings it to 2956, which I would guess is probably pretty reasonable for what would be its results against human competition.
He would say Fritzy is drawish .
Therefore of the same strenght as him.
Fritz 11 2919 16 16 1120 60.5 % 2837 37.6 %
Fritz 10 2839 11 11 2948 50.0 % 2839 29.8 %
So we have now the predicted +80 Elo from CB :-).
People who use more than one processor are minority.
They always used the name deep Fritz for more than one processor so it was easy to guess that the 80 elo is relative to Fritz10 and not relative to Deep Fritz10.1
Uri
I think, today it might be fact, that most people have a processor with two cores.
If this is factually not true, in a vew months it will become true.
(has someone information about the percentage of PC-users using a multi-core-processor today?)
Quap
http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?pid=32109;hl=#pid32109
Of course, better for the customer would be being able to buy mp engines for the price of today's single engines, soon... :-D But I guess that will not happen as long as enough people are willing to pay a higher (or double) price for the mp versions, each. That is the crucial point. As long as the existing concept makes sense (in the economical, or revenue sense), why should companies change it?
> Using them on single core remains useful under many circumstances; see for examples:
I've also tried to bring those point in front, but have to face the fact that most members in this forum seems to be keen computerchess freaks, foremost ;)
That is, less interest in analysis with (human)chess in mind, preparation etc. Maybe the majority of those who cares for those parts is just being quiet?
I know that Vas has persuad the idea of Rybka being an super engine, and with the perspetive of a later GUI, for long. Then Covekta came in, and we, that the target is there :)
As very sweet "bi-effect" Rybka has proven extremely strong against the exsisting engines, which has caught the attention of those who's main interest is computerchess and those likes both cc and human chess where analysis is a concern. And just lately it seems, there begins to be a breakthrough, a connection, to the chessworld. And that's where ChesBase has dominated for year.
Still, back to cores and "ordinary" people (you've touched the subject yourself more than briefly, if I remember correctly, and so has Uri and others): These people is often playing in clubs or on the net, and seeks ways to improve their chess, they almost give a shit about CC-ratings, they foremost want what the other guy's having. And what that is can change. Regarding the computers, many I know, is cathegorieds in two rough groups: Those who knows they have a dual all by them self, something they aimed at, and those who've discovered they have a dual and need the term explained. In both cathegories, there's some knowledge, and the understanding varies a lot. Naturally.
But those who've aimed for a dual, usually wants to use one processor for chess analysis/game, and the other for "something else". They understand the significance of one vs two processors. In some cases, the thought of using both processes for an overnight analysis appeals, but often you need a special occasion, or being a pro chess player to legitimate that. The rest is most likely not interested, except when they've just bought a new program that yeilds it's "deep" analysis in some way.
Then there's the super-normals: They have dual, and dont care, or they have a typical household comp, maybe a new one (but still one cpu/tread): They don't care if there's a benefit im using 2 treads, even less when they're told how "much" this benefit is in time.
In both groups there's quite some who may buy just just the most expensive of everything, but still they would wonder why a program uses "both" cpu's when 1 is enough... I know a guy is buying Deep Fritz.... has done for years...He is an unormal "ordinary", changing harware often. Now he has dual-laptop, and the main problem is that when he uses Deep Fritz, everything is so slow, just like before... He is for many years a the IT-manager for a big company, with a substance of technicallity in his education.
My point is, that many buys computers and computers and software without more prior understanding than they have when buy TV or phones. It's superficial knowledge, and it will probably alwas be that way. No blaim here.
So the aim at any time, when the majority is clear on benefits and caveats in one area, like hardware, is of course to present the most desent product, with all the quallities to the typical user. But where? And who?
My personal guess, is that it will last long before the inital target audience of Rybka has more more than sp's - 3-5 years??
And still we'll like to use sp's because "diffence of oppinion" matters - some of us at least.
I hope the "market" wont milk the cow too much... adjusting to the circumstances woud seem too me, as bieng able to stop benefitting from mp in time. That's about two years from now, when looking at the Europeans, the USA and Asia.
If the minority of all home computer users is single processor then it is barely minority.
For many have core duo (1st one), athlon x2, pentium D, and core 2 duo machines.
So the users of more than one processor will definitely be majority very soon (if not now).
> Disappointing. They promised about 80 Elo, but CEGT results show it's only about 40 points stronger than Deep Fritz 10 on a dual.
You are disapointed, because the new singel core engine ist only 40 ELO stronger than the old multi core version?
Why?
This ist not the engine you waited for.
So wait until Depp fritz ist out.
And don't be disapointed because the weaker SP-version couldn't outdistance the old MP-Version.
But it seems the sp-Version has increased her strength significally.
Perhaps the MP will do in a similar way.
You wrote ' They promised about 80 Elo...'
Are you shure that the promised that the new SP-Version has 80 ELO points more than the old MP-Version?
Maybe the SP-Version isn't interesting for you, OK.
I think, you shouldn't be disapointed then.
You should wait.
Quap
>You wrote ' They promised about 80 Elo...'
>Are you shure that the promised that the new SP-Version has 80 ELO points more than the old MP-Version?
What they promised is undefined. I defined it in the only way that makes any sense to those who aren't particularly interested in buying SP engines. Fritz 11 is a giant backward leap: they had a relatively good parallel implementation with DF10.1, but now it's gone. Deliberately weakening your engine is a marketing strategy I heavily disagree with...
>This ist not the engine you waited for.
It's the same engine, just with an important strength-enhancing feature completely removed. Hence my use of the word 'lobotomized'.
I don't know what you mean by this. You have to buy the strength-enhancing feature, and this feature isn't available until next spring. The advertising was clear: with Fritz and Deep Fritz being separate engines, developed separately, the "predecessor" of Fritz 11 is Fritz 10. There is a 75-elo point difference between these two, which is statistically consistent with the advertised estimate of 80 elo increase, which was probably made by the programmers, not by Chessbase, who simply relayed the information. As much as I tend to deride Chessbase, this particular instance isn't a good opportunity :-).
> I don't know what you mean by this. You have to buy the strength-enhancing feature, and this feature isn't available until next spring
Alright, alright. They've completely removed the availability of the strength-enhancing feature for marketing reasons! Happy? Makes no difference to anything I care about, though. Especially since they don't even have an upgrade price.
>The advertising was clear: with Fritz and Deep Fritz being separate engines, developed separately
How many times did you roll your eyes while typing that? In the real world (i.e. everywhere but the advertising), they're the exact same engine, the difference being that Deep Fritz is compiled with
#define SMP
No, nothing was removed--unlike Junior, Shredder, Hiarcs, Rybka, etc., Fritz and Deep Fritz actually ARE two basically different engines. Even in the cases where they've been roughly the same overall strength, such as with Deep Fritz 10, which is the only time when the "Deep" version and the "Shallow" version have been somewhat similar overall, there is quite a noticeable difference in the evaluation of many types of positions. In the case of Deep Fritz 6, 7, and 8, there has been a noticeable strength increase on single processor from the ordinary to the deep versions--the deep versions always have more endgame knowledge and better play in many other types of situations. They're two different engines altogether. There are far, far more similarities between Strelka and Rybka beta, for example, than there are between Fritz engines and their "Deep" versions. It's much more than simply "#define SMP"--when the author works on a new Fritz version, he goes back to work from the previous version. When he works on a new Deep Fritz version, he goes back to work from the previous Deep Fritz version. Not all of the knowledge is compatible, either--it would be a mistake to simply copy the "extra" endgame knowledge from Deep Fritz 11 into Fritz 11 because Fritz 11 is different enough that this could induce some serious weaknesses.
I think the thing on which we agree here is that compared with Rybka, Fritz is poor buy in that you get a far weaker engine for a higher price. Use that point! That is worthy enough of mention without having to pull at weak straws. If the only thing in which one is interested is chess analysis in all different types of positions, then there is no reason why one should do anything more than buy the latest Rybka and download Fruit and Spike, which are free. If one wanted to make absolutely sure that everything is covered, then buy Zappa Mexico, too, since its style is so different from Rybka.
> No, nothing was removed--unlike Junior, Shredder, Hiarcs, Rybka, etc., Fritz and Deep Fritz actually ARE two basically different engines.
Even if that's true, it wouldn't be very difficult to port the parallel search, especially for the guy who developed both.
> There are far, far more similarities between Strelka and Rybka beta, for example, than there are between Fritz engines and their "Deep" versions.
There are more similarities between Loop and Toga than Strelka and Rybka beta according to the data you're referring to. It's just not reliable.
> I think the thing on which we agree here is that compared with Rybka, Fritz is poor buy in that you get a far weaker engine for a higher price. Use that point!
The problem is that all the extra clutter (GUI,Playchess,database,videos), while nearly useless to me, is generally pretty valuable.
Regards,
Alan
> Fritz 11 is a giant backward leap: they had a relatively good parallel implementation with DF10.1, but now it's gone...
Only in the same way as you can attack Microsoft, when they come
out with an new Works-Version and you proclaim heatedly, that their
Office now has done a giant backward leap!:-o
Not fair and not my opinion.
Quap
The fact is they can't ride this horse for too long. Sooner or later the trend will shift from legacy systems and they either had to prove fritz is the best engine or accept the fact and play equal in modern platform if they want to survive. We are already at the early stages of that shift in knowledge among chess users. It's just a matter time for others to get educated.
Just weight and watch the fun after that. :-)
I have a lot of respect for what they did to chess community for quiet a few years consistently. I got interested in chess computers only because of fritz and it's popularity like so many others. Thanks to their very good marketing team again. I used to have a fidelity designer 2100 when i was a kid and used to wonder how fritz 2 used to beat it left right and center in a modest 386/33Mhz machine :-)
Please don't think i am just here to thrash fritz. I would love to see fritz come back big like any other chess engine, which is always good for the chess world.
All that said and done, there is only one thing that bothers me, they don't have the stuff they are trying to sell and that makes us look like a fool in their hands. That sounds like a rip off. Moreover i always felt they had total domination over computer chess world at one point with top 5-6 engines as their product. This kind of monopoly is not good for chess industry.
Why?? Like Quapsel said this can be easily understood with MS analogy if you think a bit. (Again don't get me wrong, i respect both Bill and Steve equally). In todays world quality matters over riding a legacy goodwill horse.
I can keep writing more and more with examples to prove my point. But don't want to start a serious war over here between Rybka and Fritz fans. Peace :-)
BfF-Gauntlet (Aktivschach, C2D)
_
Shredder 11
Junior 9 : 20 (+ 12,= 5,- 3), 72.5 %
Fritz 10 : 20 (+ 7,= 6,- 7), 50.0 %
Rybka 2.3.2a 32-bit : 20 (+ 5,= 14,- 1), 60.0 % !*
HIARCS 11.2 SP : 20 (+ 9,= 3,- 8), 52.5 %
Fritz 11 : 20 (+ 6,= 4,- 10), 40.0 %
_
Fritz 11
Junior 9 : 20 (+ 15,= 3,- 2), 82.5 %
Fritz 10 : 20 (+ 10,= 5,- 5), 62.5 %
Rybka 2.3.2a 32-bit : 20 (+ 5,= 11,- 4), 52.5 % *
HIARCS 11.2 SP : 20 (+ 10,= 8,- 2), 70.0 %
Shredder 11 : 20 (+ 10,= 4,- 6), 60.0 %
Jouni
IMO, the Fritz 7 book is one of the best opening references at all (of course without the perfect "up-to-date" factor). It was bigger still, then the later Fritz 8 book. It containes a giant number of moves and variations from opening analysis including rare and unusal "off-theory", not just from imported games.
I think it is a masterpiece, not a crappy joke... Also, I do not assume that the recent Fritz book are worse. The book designer Alex Kure is considered to be in the same league with Jeroen Noomen and Sandro Necchi.
If anyone would like to send me the Fritz 11.ctg book I'd appreciate it; I don't think that would be an egregious case of piracy. I'd be happy to put it through its paces against my tournament book and offer an objective assessment. If it does comparably to Harry's at the same settings ("Normal") I'll be happy to eat my words and make a public statement to that effect.
Your comment on our hobby being a sideshow of a sideshow is well taken. You're right. But at the same time I think the computer chess community at large--everyone who buys a product like Fritz--would be better served if opening books were as muscular as possible within the constraints of what can be compressed onto a CD (or two).
You of all people are well aware that a good engine book is not independent of the engine that will be using it. On the contrary, the lines should be carefully chosen to attain positions the engine in question is relatively good at playing.
Since you clearly do not intend to purchase F11, its not obvious to me how you can evaluate the F11 book as it was intended to be used. I submit that it is only important for this book to work well for F11. How well it works for other engines is really not relevant.
Regards,
Alan
This is not good enough. No classically-trained scientist would tolerate such uncertainty and inexactitude. Systematic progress can only made with rigorous testing and very precise measurements. (Which, incidentally, is a hallmark of my style of analysis.)
I submit it is only important for a book to work well with ALL engines, and that the measure of success is not in the final game outcome, but the position attained at book exit. After that the book has passed the baton to the engine; the engine has to perform in accord with the position it was handed and its relative strength against the opposing engine/opposing hardware.
All this said, I am not entirely discrediting other views. I have always maintained that there are other approaches to bookmaking that work extremely well precisely because they are extreme. Everyone has to follow an approach that their imagination and resources permit. The strongest bookmakers are all extremists.
Your thesis could only be valid if all engines play all position types equally well. This can, and has been, disproven from doing statistical tests on test group results, grouped by ECO. Results of these tests on large numbers of games show that percentage of wins or losses (to normalize for engine strength) of ECOs varies significantly from engine to engine in a manner that is well beyond statistical significance.
Please note that I am NOT claiming that engine developers typically tune the opening books they distribute with their own engines to work well with their own engines. Why not? Beats the hell out of me. Probably for the same reason that so many engine developers show up at tournaments with badly constructed books.
Alan
I will only say this: generally speaking, opening books in the public domain are inadequate. With a little effort, they can be better. With incredible effort, they can be better yet. One problem with public books, as I see it, is in the realm of philosophy: to what degree should books dominate the action vs. letting the engine strut its stuff. For instance, if the Fritz book was so strong that it was routinely playing 30 moves out of book, hardly giving the engine anything to do, how would that go down among the people who shelled out 100 euros for the latest iteration of a program? Pretty soon people would realize they were buying a book as much as an engine (perish the thought). So it seems to me engine authors tend to throttle back on their public books because they're standing at the precipice of a proverbial slippery slope. The canny authors put out milquetoast books and play in tournaments with private books on steroids.
To answer Alan, with any project, you have to have some long-term principles. For example, Rybka targets 64-bit, multi-processor systems. If something helps in 32 bits or on a single-processor machine, it better be really easy or I just won't do it. Maybe this isn't absolutely optimal today, but I'm ready to bet on it long-term.
Opening book authors also have their principles and their long-term strategies. You don't really want to scramble to overhaul your book and the way you work every time the engine you're booking for improves its evaluation.
Vas
> If anyone would like to send me the Fritz 11.ctg book I'd appreciate it; I don't think that would be an egregious case of piracy.
Why wouldn't it be piracy? I think the value of the book was charged with F11's price, so I don't see why people that didn't buy F11 could be able to own the book for free.
And if you mean that this wouldn't be bad piracy, are you implying that there are cases in where piracy is good?
The only unique thing related to Fritz single/mp releases is, that unlike with other engines, there is a delay of some weeks or month between the single release and the mp release. Often - if not always - the mp version contained improvements in addition to the mp capability.
(IIRC there was no Deep Fritz 9 at all, though?!)
I don´t believe it! Make a match over 100 games between Fritz 11 vs. Deep Fritz 10.1 4CPU with at least 30 minutes per game and you will see.
Disappointing. They promised about 80 Elo, but CEGT results show it's only about 40 points stronger than Deep Fritz 10 on a dual. What's worse, it's most likely not even an improvement compared to DF10.1 with enough processors.
New, dramatically lobotomized engine. Weaker than the previous version on the same hardware, provided your hardware is good enough. A quad should do.
And they're still force-feeding the GUI and Playchess with it.
And the database is most likely a subset of what I already have.
And there's no download version.
You should compare single Fritz 11 vs. Single Fritz 10.
We don't know anything about Deep Fritz 11.
Fritz 11 : 2914 19 19 820 58.0 % 2858 36.7 %
Fritz 10 : 2839 11 11 2948 50.0 % 2839 29.8 %
So there is a differnce of 75 points.
WITHOUT ADDRESSING THE CHALLENGE OF RYBKA, FRITZ WILL ALWAYS BE WEAK .
BECOMING OF: SPORTSMANSHIP, BE A MAN, STEP UP TO THE PLATE, SHOW US YOUR STUFF, STOP COWERING IN SHAMEFUL FEAR.
brian carson, toronto
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill
