Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Going back to R3
- - By Wayne Lowrance (***) Date 2010-07-21 19:34
Until mpv bug is fixed. Outside of this I really like the lil girls play overall
Wayne
Parent - - By Banned for Life (Gold) Date 2010-07-21 19:56
R4 is certainly a double edged sword, indispensable in certain scenarios, but not to be relied on. Zappa is more reliable, but will get out-searched on a regular basis by R4 or SF. SF brings an interesting and independent view of positions.

I'm not so sure about using R3 at this point in time because the clones offer similar evaluation in a shorter period of time.

IMHO Rybka has always been less useful for MPV than other top engines, even without the current bug. For some reason, Rybka slows down more than other engines. Vas has provided a formula for this slowdown which is almost linear in terms of the number of MPV lines, while other engines slow down less. I'm not sure why this is, but it may be do to Rybka spending less time on non-MPV moves, or getting less advantage from hash entries of other lines.
Parent - - By Wayne Lowrance (***) Date 2010-07-21 21:12
Your are most likely right. I guess right now I am really down on R3 mpv.
I have always relied on mpv in r3 and others
Wayne
Parent - By Loboestepario (****) [us] Date 2010-07-21 23:18
simultaneous single-pv + IDeA is an option when mpv fails
Parent - - By Uly (Gold) [mx] Date 2010-07-22 08:48
There's no rule that says you can't use both engines. There's very few overlap.

I agree with Banned for Life that relying on Rybka 4 is too risky, sometimes it just has no idea on the positions, and I just keep using it everywhere because I'm biased (on a double blind test of move quality, I'd have stopped using Rybka 4 on some of my games), and I'd play significantly weaker if I couldn't use Rybka 3 (and Human and Dynamic) on my games.

But relying just on Rybka 3 would be a big risk too! You just never know when Rybka 4 could have been useful, I've also seen positions where it's the only engine that finds the win or the draw in critical variations (and I keep saying the same thing about all my main engines, otherwise they'd not be main). I could imagine you easily entering into a position that Rybka 3 thinks is 0.40 while Rybka 4 thinks it's 1.20, and while both engines may be wrong, Rybka 4 would be closer to the truth, you'd just be better following Rybka 4 here, it's all about move choice! But if it happens that you are black, you'd be better choosing a move both engines rate as 0.50 than the one in the example, and there's just no way to know unless you analyze with both engines.

I tend to analyze my games with 6 engines, that seems the sweet spot, 7 or 8 seem overkill as I can't analyze as deep as I should, 5 make it very likely that I'm not analyzing with the right engine.
Parent - - By Wayne Lowrance (***) Date 2010-07-22 14:38
Wow, 6 engines. How deep do you average in each engine ?
Parent - - By Uly (Gold) [mx] Date 2010-07-23 04:07

> How deep do you average in each engine ?


For the Rybkas (4, 3, Human and Dynamic) I average 17 and rarely go over 18, Stockfish reaches about 27 in the same time on some positions. The trick is to move out and analyze positions that matter.
Parent - - By NATIONAL12 (Gold) [gb] Date 2010-07-23 21:55
you play completely different from me,when i talked about my play with Zappa and the need for big hardware to find true AN of position.that does not mean i do not use Jump AN(well remembered) however i have the luxury of having 2 comps and hopefully 3 shortly.i have found in many games a change in the root move which is often very deep.this means that superficial AN and jumping forward from this is worthless.also of course i have the luxury of using my brain and testing my ideas deeply.:smile:
Parent - - By NATIONAL12 (Gold) [gb] Date 2010-07-23 22:05
you play far to many games at once.you acknowledge this yourself.:sad:
Parent - By Uly (Gold) [mx] Date 2010-07-24 02:12
Yes, big mistake from my side.

Actually, one of my opponents, x64 at OC Forum, offered me to postpone our game as he noticed I was overloaded of games, and I gladly accepted, it has been very useful since my position on that game was also very complex!

But I couldn't miss the opportunity of playing turbojuice or Alkelele, or on a team game! Something I've always wanted to do, and it's unknown if the chance would appear again (or at least soon, what if Alkelele disappeared for another 3 years?!, what if turbo quits corr computer chess for real?!, will we ever get a team as strong as you, Bobby, Tony and me?! if those games were going to happen, they must have happened now).

Now I'm trying to rescue my games :yell:, but I don't blame too many games at once, I blame my analysis methods, but I keep having new ideas of a way to do analysis, sometimes very good, sometimes too crazy, some look decent at the start and I end losing the game, but it's unknown how they'll work unless I try them in actual games.

You'd probably laugh about some of them like "Ping Pong Analysis" or "Election Analysis", that in retrospective are as ridiculous as their names :lol:
Parent - By NATIONAL12 (Gold) [gb] Date 2010-07-23 22:50
i remember my arguements with DarkHorse on this on Hiarcs forum.sure i am right.time will tell.
Parent - - By maceisthemax (*) [us] Date 2010-07-23 23:48
It sounds like this bug in the MPV doesn't effect the evaluation of the position, it just causes the engine to crash and do crazy stuff that is detectable ("red flags").   I'm using it to analyze my games after the fact...

1.) if it crashes, or effects the evaluation, I will be aware and able to work around it, right?

2.) Or will I need to go back over all of the analysis I have done with DR3?

Thanks for all your sharing of your expertise...I feel like I should be paying for this.
Parent - - By Uly (Gold) [mx] Date 2010-07-24 02:16
The MultiPV Bug just makes the engine stall for hours (say, 29 hours that most people will find unreasonable).

That is, Rybka would reach depth 15 at MultiPV and then just stay there, as if you stopped her (expected behavior would be for her to reach depth 16).
Parent - - By maceisthemax (*) [us] Date 2010-07-24 05:43
yes, every once in a while, it will do this, like if I try to leave it on overnight, but it seldom happens during short-term analysis (12-14 ply).  I just close the program and re-boot.  It's a nuisance, but if it get's me a quality move, it's worth it.  Other than that and the occasional end-game mess-up, the opening analysis is still the best in the business, right? I won't be getting any lemons handed to me on move 10, will I?
Parent - By Uly (Gold) [mx] Date 2010-07-24 09:30
You'll get lemons if you analyze like that, never leave Rybka at MultiPV analyzing overnight, it'll be a waste of time.
Parent - - By MWYoung (**) [us] Date 2010-07-22 16:45
I have also done this. I like Rybka 4, but the bug makes it were I can not use the program, the way I need it. Rybka 3 is not a option,  since the other programs out are faster and better then Rybka 3, and maybe a bit stronger and for sure faster then Rybka 4.

What makes my blood boil about this whole bug thing not being fixed. I paid 100 bucks for Rybka 4, but I get better customer service from the free programs. We let them about a MPV bug in their program, and BAM it is fixed with a new release in a few days. With a Sorry from the programmer for putting out a version with a bug!! :)

I am starting to think that Rybka 4 is going down the same road as Rybka 3, and no bug fix will be provided. I hope I am wrong.
Parent - - By maceisthemax (*) [us] Date 2010-07-22 21:22
Maybe I fell down a rabbit hole or I'm in a dream where I can't remember how I got here.:confused:

I'm have a a Deep Rybka 4 DVD but I have been waiting to find out if I really want to install it.  Am I to understand:
A.) There is a bug on Deep R4 which makes it LESS trustworthy analysis machine than Deep R3?:eek: 
B.) This bug is NOT going to be fixed? :cry: 
C.) I wasted my money?:mad:
Parent - - By MWYoung (**) [us] Date 2010-07-22 22:04
A. No, in MPV mode, Rybka 4 will stop working. So this is worse then Less Trustworthy, it is just broken.

B. No bug fix yet. Vas said he may or may not do a bug fix. Here is his history, Vas said he would fix Rybka 3, it never happened. So I would not hold my breath for a bug fix with Rybka 4. That is his history.

C. Yes... If you use Rybka 4 for analysis because of the MPV bug, unless the bug is fixed.  I also would not use Rybka 4 for many endgame positions, not sure this is a bug, just bad evals in many endgames. Rybka 3 also had this problem. If you just play games with Rybka 4, and don't care about analysis,  Rybka 4 is very strong, but there a free programs out there that are just as strong or some say even stronger then Rybka 4.
Parent - - By maceisthemax (*) [us] Date 2010-07-23 04:30
If there were some kind of bug with DR4 which made it inferior to DR3, I would think that would be reflected in the latest ratings list.  Yet, all the ratings lists still show R4 and Deep R4 at the very top.
Am I to understand this bug only manifests itself in "analysis", and NOT when playing?
Parent - - By Uly (Gold) [mx] Date 2010-07-23 09:06

> Am I to understand this bug only manifests itself in "analysis", and NOT when playing?


Yes.
Parent - - By Wayne Lowrance (***) Date 2010-07-23 16:42
Howdy Vytron. I have  enough engines to do 6 eng analysis to depth 17. I
do not go that way though. I have 2 or 3 that I favor besides the lil girl. I
Like Zappa Mexico, Hiarcs 12 and the Rybka's. I have Stockfish & that group
of Free down load version but do not trust them.
MPV (deeper than 17) + multiple engine analysis has worked for me. But I
understand your depth 17 mode limitation. So for me mpv is a must. Some folks
here are skeptical that Vas will fix this, I am not one of them. I believe he will
as not too would be very very damaging to his reputation. It is a trust thing.
Parent - - By Banned for Life (Gold) Date 2010-07-23 17:18
Some folks
here are skeptical that Vas will fix this, I am not one of them. I believe he will
as not too would be very very damaging to his reputation. It is a trust thing.


I'm sure he will fix the problem where the kn/s goes way up and the engine hangs. This is much worse than any of the R3 problems. I don't see it as being as much about trust as about having an engine that is usable for analysis. I can't believe it got through Beta testing this way. I have to assume that most people were just using it for engine-engine games and very few were using it for analysis.
Parent - - By Wayne Lowrance (***) Date 2010-07-23 17:55
I dunno, Either the testers were not thorough, or Vas just did not heed their input, either way it is a shame.
You know I have bought  Hiarcs, shreddar, fritz, Junior etc in the past and I don't recall as many bug reports
from these Engines. So Vas needs to consider where he has been or where he is going with his products.
Wayne
Parent - By tomgdrums (****) Date 2010-07-23 18:27

> I dunno, Either the testers were not thorough, or Vas just did not heed their input, either way it is a shame.
> You know I have bought  Hiarcs, shreddar, fritz, Junior etc in the past and I don't recall as many bug reports
> from these Engines. So Vas needs to consider where he has been or where he is going with his products.


I agree.  Hiarcs and Shredder are usually very clean and if there is something they clean it up quickly.  Even the latest version of Junior from the Hiarcs site came out kind of buggy BUT they fixed it quickly!!

That is the difference.  Bugs are bugs and they will show up from time to time.  It is how the author reacts to the bugs that makes the difference.  Does the author take care of his or her customers? 

On a recent trip, Hertz screwed up my car rental reservation BIG time!  BUT they made arrangements and then they even gave me $50 off the rental.  That made a difference to me.  They cared enough to rectify the mistake AND try to hold out an olive branch with the $50 off.

I don't expect Vas to give any of us money back, BUT to not fix what are now KNOWN problems is just disrespectful to his customers, and a good way to lose customers as well.
Parent - - By Uly (Gold) [mx] Date 2010-07-24 02:22

> I have to assume that most people were just using it for engine-engine games and very few were using it for analysis.


I was using it for analysis but don't use MultiPV at all.

Kapaun reported the bug but Vas thought that the engine stalling for hours was normal (even after several forum members were complaining about it after R4 was released, it took a long while to Vas to confirm it was a bug).
Parent - - By Banned for Life (Gold) Date 2010-07-24 03:18
R4 is flat out broken when trying to analyze with MPV. But if you never use MPV it still feels like you are stuck in molasses while you are analyzing.
Parent - By Uly (Gold) [mx] Date 2010-07-24 03:30

> But if you never use MPV it still feels like you are stuck in molasses while you are analyzing.


I haven't noticed a difference from Rybka 3 while relying on Exclude moves for my analysis, it's faster than MultiPV in most circumstances (though it can't be automated).
Parent - - By davidwhite (***) Date 2010-07-25 07:43
Alan,

I almost wish that you were right about R4 being flat out broken when trying to analyze with MPV because,in that event, Vas  would have no choice but to put out a quick fix.Just imagine the pressure that would be on Convekta,ChessBase, and any of the resellers
to immediately refund to all purchasers of the "Deep" version the premium they'd paid over the SP version.

My own experience has been that sometimes the Deep R4 analyzes properly and sometimes it doesn't. I've been unable to isolate precisely which kinds of positions I can trust Deep R4 and which I can't.I'm using a Paladin XLC V3 from IBUYPOWER  which features
a 6-Core CPU powered by a Core i7 980X Extreme Edition factory overclocked to 4.13GHz with 12 GB Memory,2x160GB SSD Primary Hard Drives and an Asetek Liquid CPU Cooling system so it's possible that my hardware alone is responsible for the R4 sometimes
operating correctly with MPV analysis.

Of one thing I'm certain,however, and that is that we had what we considered to be the elephant of chess programmers go into labor for 18 months only to deliver unto us...a relative mouse.
Parent - By Uly (Gold) [mx] Date 2010-07-25 12:51

> Just imagine the pressure that would be on Convekta,ChessBase, and any of the resellers
> to immediately refund to all purchasers of the "Deep" version the premium they'd paid over the SP version.


That's unrelated, MultiPV means Multi-Principal Variation mode and it's unrelated to the Multi-Processor version.
Parent - - By Banned for Life (Gold) Date 2010-07-25 14:36
David,

I'm glad to hear that you are able to use R4 most of the time. I have not been so lucky with my lessor hardware. Even in single PV mode when moving quickly back and forth through variations, I see frequent cases of R4 getting stuck at low depths for long periods of time (without the symptom of rising kn/s which is a trademark of the multi PV problem). This makes the engine seem sluggish and ineffective. I hope Vas fixes this problem, as it is much, much worse than the issues with R3.
Parent - - By Uly (Gold) [mx] Date 2010-07-26 10:17

> I see frequent cases of R4 getting stuck at low depths for long periods of time (without the symptom of rising kn/s which is a trademark of the multi PV problem). This makes the engine seem sluggish and ineffective. I hope Vas fixes this problem


This problem hasn't been reported and I don't think is acknowledged (seems like a new issue).
Parent - - By Banned for Life (Gold) Date 2010-07-26 13:33
I've had another engine analyzing a position for several hours now, and am reluctant to break off at this moment. I'll give you my settings and a position later. You would be a perfect person to test this because it seems to have some of the problems that were associated with persistent hash when doing backward analysis (i.e. needing to get to a high level to pick up previously discovered results).
Parent - - By keoki010 (Silver) [us] Date 2010-07-26 15:34
I've seen it also, but thought it was part of the mpv problem.
Parent - - By Uly (Gold) [mx] Date 2010-07-27 12:52
It's a bug, it could be related and self-fixed after the MultiPV bug is fixed, or it could be unrelated and unique.

The way it affects SinglePV makes me think it's unrelated.

Banned For Life, please, try to reproduce the bug with Large Pages OFF, I suspect it's related and knowing whether it's or not would save a lot of time.
Parent - - By Banned for Life (Gold) Date 2010-07-27 14:26
OK, as I've previously stated, all of my Rybka analysis to date has been with large pages on and with calculating exact scores. It's hard for me to imagine the problem being associated with large pages (this should be totally transparent to the application), but anything is possible.
Parent - - By Uly (Gold) [mx] Date 2010-07-27 19:06

> It's hard for me to imagine the problem being associated with large pages


It's easy for me because since I turned them off I've never seen Rybka 4 slowing down (assuming we're talking about the same slowdown).
Parent - - By Banned for Life (Gold) Date 2010-07-27 19:13
OK, That would explain why so few people have seen this problem...
Parent - By keoki010 (Silver) [us] Date 2010-07-27 20:02
Hmmm, I use large pages on my IA analysis, but no large pages on my idea analysis. Maybe that's why I haven't had any problems with my idea engine, of course it's also only using 1 core.
Parent - - By NATIONAL12 (Gold) [gb] Date 2010-07-27 23:24
ask BigBen he uses large pages also.
Parent - By Banned for Life (Gold) Date 2010-07-28 00:58
Has he noticed that R4 analysis resembles being stuck in molasses?
Parent - - By maceisthemax (*) [us] Date 2010-07-23 19:57
As long as Rybka's ELO is not effected, there is really not incentitive for fixing this bug, is there?

I don't understand much of this discussion, but I would like to ask a simple question, or two.

1.) Is the cause of this bug ALSO contributing to any gain in ELO?
2.) Is the MPV crashing and unreliability a designed-in "sacrifice"?   Making Rybka a stronger engine vs engine (higher ELO), yet throwing the analysis under the bus?
Parent - By Fulcrum2000 (****) [nl] Date 2010-07-23 21:32
No, No
Parent - - By titanium cranium (***) [us] Date 2010-07-26 14:36 Edited 2010-07-26 14:40
I here ya Buddy. I can understand why Vasik would want to issue an update to her, even if it wasn't any stronger - money. But I'm not sure R4 can beat R3 in a match.
Parent - - By Nimzovik (***) [us] Date 2010-07-28 07:00 Edited 2010-07-28 07:04
Ultimately----  Vas has defintely earned a place in chess history with Rybka 1 , 2 and 3, this simply can not be denied.  I am grateful as he obliterated the Chessbase hegemony.  I am grateful in that he gave us for the time the avant garde in chess.  He is an icon of  a point in chess history regardless of what ever the fruit controversy is.  Fruit did not achive what Rybka did.  Kudos to Vas.  If he can not keep up the pace, he is in very good company ----from rex chess, to chessmaster to junior, to fritz to shredder and of course Hiarcs.  Any whinning is juvenile and immmature.  He (vas) gave us quantum leaps in the chess world.  My hat tips to Vas.  There will be others to stand on his shoulders.  I did not buy rybka 4 .  I do not intend too .   The engines to date do quite well in concert. Until another Quantum and I do mean quantum leap creator appears I shalll stick with r3 as a main stay.  I WILL ANALYZE WITH OTHER ENGINES AS WELL.  Why?  Simply because unlike checkers...chess has not been solved.  Thank God.
Parent - - By Labyrinth (****) [us] Date 2010-07-28 10:29

>Any whinning is juvenile and immmature.


This reminded me of the "CoolHwip" bit on Family Guy.

Everyone loves to Hwin!

http://www.youtube.com/watch?v=lich59xsjik

>Fruit did not achive what Rybka did.


Like Vas said "It is a bit of a pity that Rybka won't make the same contribution to the computer chess community". Remember Fruit was open source unlike Rybka. Unless you count decompiled abortions.

Still, Rybka is no minor contribution.
Parent - By Uly (Gold) [mx] Date 2010-07-28 12:29
In a world without Rybka, I can easily imagine engines still stuck in the ~2950s ratings.
Parent - By vroger007 (**) [be] Date 2010-07-29 19:34
all,

I never went to R4 in the first place, a.o. because the promosed R3 bugfix never came,
and coz I'm using engines only for analyzing OTB games,
and also coz I'm still waiting for the version which includes BU (also "promised" by Vas).
Meanwhile I will switch from analyzing with rybka-only (blundercheck with low threshold)
into doing "compare analysis" with DR3, DF12, DH12 and DS12.
Hope to get DR5+BU soon and hope rybka never gets DR6-ed [ inside joke for the active professionals :-) ]
Up Topic Rybka Support & Discussion / Rybka Discussion / Going back to R3

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill