|
< Earlier Kibitzing · PAGE 1684 OF 1784 ·
Later Kibitzing> |
| Jul-12-08 | | zsoydd: @<kb2ct>:
Hi Ken,
our old team was then strong enough to win a few games with black, against strong otb and also corr players. And our new team?
best regards,
zsoyd |
|
| Jul-12-08 | | kwid: Jul-12-08 < zsoydd > Have a look at the <Preparation for the GMAN game> We are trying to find a hole in his book. |
|
| Jul-12-08 | | RookFile: This must be frustrating for Timmerman. He's about as close as you can get to a draw without actually getting it. |
|
| Jul-13-08 | | kb2ct: <RookFile:> GMT did not understand that rook pawns are an advantage not a liability in rook versus knight endings. :0) |
|
| Jul-13-08 | | RookFile: Well, you're speaking music to my ears, of course. |
|
| Jul-13-08 | | falso contacto: i voted Re4 cause its already decided.
hope it helps anyway |
|
| Jul-13-08 | | vikinx: If we play well, we will have a theoretical win, so try best! |
|
| Jul-13-08 | | MostlyAverageJoe: <f54280: So, at the end, I would pedantically side with <Dan>: conceptually, the eval the engines print are the eval of the possible *moves*.> Actually, these are evals of the positions after the first move of each line, obtained under earlier described assumptions and constraints. What Dan fails to comprehend is that attributing a value to a position or to a best move that can be made in that position is completely, unquestionably, 100% equivalent. < f54280: Generally, those are the evals of the *positions* reached by those moves, which are often the eval of the terminal *position* reached by what the engine considers best play from those positions.> You've just rephrased the gist of what I posted. The fact remains that all these numbers come from computations done on *positions* reachable from the current one, and Dan can play his semantic games and display condescending attitude till he turns blue, and it won't change the reality. |
|
| Jul-13-08 | | DanLanglois: <ontocaustic: a very philosophical way of assuaging dan's confused stubbornness, hms> I'm not confused & I'm not stubborn. |
|
| Jul-13-08 | | Xenon Oxide: Question: Can we reach page 2008?
Hint: note the circular/highly semantic "engine eval" arguments on the main page ;) |
|
| Jul-13-08 | | DanLanglois: <Domdaniel: A point that possibly hasn't been stressed sufficiently here is the sheer variety in the way people use their engines. Some play out engine-vs-engine games. Some set the comp to infinite analysis of a single line. Some slide forward frequently, others believe that the deep-ply eval produced by an untouched engine has extra significance. And so on. I'm just glad that all of these methods are being used simultaneously, as each has its merits. Myself, I like to set an engine on infinite analysis, but configured to display its top 3-4 lines. I'll vary this in positions where 5 or 6 lines might be more interesting, or where the top lines clearly transpose. I also like to 'interrogate' the machine, by inputting moves that do not form part of its analysis.
>
This discussion is brilliant. |
|
| Jul-13-08 | | DanLanglois: <Domdaniel: I'm just glad that all of these methods are being used simultaneously, as each has its merits.> I'd like something more concrete in terms of knowledge of which of these efforts really is improving the 'trustworthiness' of the computer eval numbers..? tall order |
|
| Jul-13-08 | | sentriclecub: <MAJ> and <Langlois> I think you two are doing a great thing. Both of you have valid points, and I'm sure several people have learned a lot. I think you guys need to take a step back, and agree on terminology. Also, this will help others (including myself) understand better what you both are talking about. I sort of understand what DanL is trying to say, but I think he is unfairly putting the burden on MAJ to try to figure out what DanL is trying to say (which instead gives MAJ more rope to hang DanL the person, and not the argument). Thanks
|
|
| Jul-13-08 | | sentriclecub: Specifically, MAJ has been entirely consistent with MAJ's words. DanL, either define the words you use once and for all and post them and "copy link" to refer back to it, to help you remain consistent. OR adopt MAJ's lingo. Of course, you two might need to hammer out a couple definitions, but that is a lot easier to do than trying to proceed (either side) with the lack of clearity that has been carried forward so many pages and grew into such a worthwhile topic. |
|
| Jul-13-08 | | DanLanglois: sentriclecub,
I have one particularly controversial point about computer eval numbers. When I try to 'win' this point, the discussion tends to devolve into a pendandic debate about search algorithms & such. Here is an eval number: 6.26
and here is a line from RandomVisitor:
(6.26): 59...Rc7 60.Kd4 Rf7 61.Kxc4 Nf2 62.Rd4+ Ke7 63.Rxa5 Ke6 64.h5 Rc7+ 65.Kb4 Rc1 66.h6 Now, we have an eval number, a line, or, several moves, and of course several positions to consider here. But, only one eval number. It's worth reflecting upon what specifically the eval number is telling us. I don't really want to 'lecture' on this point, so much as 'focus' on this point. |
|
| Jul-13-08 | | DanLanglois: secntriclecub, the eval number (6.26) is not telling us that the position after 66. h6 is 6.26. It is telling us that the position after 59...Rc7 is ALREADY 6.26. I have been condescendingly corrected by many here, but i'm the one who is right. I'm laughing cuz of the highhanded way I've been corrected on this. Hey, we can drop it for a while, bring it up again in a month, whatever. I think it's an important point, potentially. |
|
| Jul-13-08 | | ajile: < hms123: My list of hard workers: <ajile>.
The rest of you are total slackers :-)
Especially me.
(It is impossible to come up with a list and not overlook many others. I managed to overlook everyone else all by myself!)> lol
:o)
|
|
| Jul-13-08 | | sentriclecub: <DanL>
Several people including Mueller, F150, and myself agree with the substance of your debate. However, we're not arguing against you. MAJ on the other hand, has to keep his arguments and ideas sequestered from yours, and he has to be able to explain the merits of his argument and weaken yours. This is a positive thing for the team, who observes and learns from this exchange. However, I see a sad roadblock because of language barriers. <On a side note> Let me throw my opinion in and see what you think of it. (6.26): 59...Rc7 60.Kd4 Rf7 61.Kxc4 Nf2 62.Rd4+ Ke7 63.Rxa5 Ke6 64.h5 Rc7+ 65.Kb4 Rc1 66.h6 The eval number 6.26 represents the score of the FEN position well past move 66. h6 Here is my proof
Rybka evaluates "kilonodes per second"
The number of nodes is a function of time. The longer you wait, the more nodes will be analyzed and the more accurate. (10 ply is more accurate than 5 ply) The number of nodes is a function of nth-root-mean-^n and of ply depth and pruning selectivity. root-mean-squared is what that confusing term means when n = 2 for n = 3 it is cuberoot-mean-cubed
etc...
The total number of nodes is proportional to the nth power of the nth-root-mean^n ply depth and proportional to the pruning/selectivity. What does this mean? That means if you haven't bothered to learn why Rybka slows down reaching successive plys, then you might not understand what an eval number is. Not because of being incapable, but because of looking at the issue from the "chess" angle. That F150 guy, I was pretty sure he had a very good technical knowledge of computer science, and when I saw his massively technical post, it was no surprise. The answer is that the eval 59...6.59 is for a unique FEN position way way way way far out from the root position. Well beyond move 66. Afterall, if Rybka can evaluate 100 kn/s, then what if you leave it running for 10 hours, and come back it has only evaluated the starting move 59...Rc7 Then the kn/s second is bogus, which isn't the case. http://en.wikipedia.org/wiki/Reduct... Whose side am I closer to in my independent opinion? <MAJ> or <DanL> |
|
| Jul-13-08 | | sentriclecub: Short answer though, (ignoring the long post) is that the eval number is a DEFINITE measurement of a FEN well beyond move 66. The eval number is a HEURISTIC for move 66 and it is a HEURISTIC for move 59... |
|
Jul-13-08
 | | kamalakanta: This is my first time in a team. I love chess, and I am seeing a discussion about computer terminology. I think White is winning after 57.Re4, Re4+ 58. Ke4, threatening 59. Rb5+. Can we do some analysis WITHOUT computers, please? |
|
| Jul-13-08 | | Artar1: <kamalakanta>: <I think White is winning after 57.Re4, Re4+ 58. Ke4, threatening 59. Rb5+. Can we do some analysis WITHOUT computers, please?> Doesn't Black have a more stubborn defense with 57...Rc6, avoiding the exchange of rooks? With every simplification of the game, Black draws that much closer to defeat, wouldn't you agree? |
|
| Jul-13-08 | | sentriclecub: If you click the "sticky" at the top of the page, it will list the categories. We have divided up the likely outcomes into manageable categories. If you have any further questions, this main page is a great place to ask, as we're always looking for opportunities for team growth (analysis especially). |
|
Jul-13-08
| | kamalakanta: Artar1:
Doesn't Black have a more stubborn defense with 57...Rc6, avoiding the exchange of rooks? With every simplification of the game, Black draws that much closer to defeat, wouldn't you agree? Artar1,
I agree with you, but can Black avoid the exchange of rooks? If 57...Rc6, can't White play 58.Rb5+, Kd6 59.Rxh5, threatening 60. Rh6+, exchanging rooks anyhow? This is why I feel Black is lost: He has trouble defending the c4 pawn, and the h5 pawn will fall as well. The White rook on b8 is covering Black's only threat of counterplay: invading with the King through the b4 square. What can he actually do to prevent White from increasing his advantage? |
|
Jul-13-08
| | kamalakanta: Thanks to Sentriclecub for directions to the forums in the stickie, and to Artar1 for his analysis in said forums. It seems to me Black is definitely lost. |
|
| Jul-13-08 | | sentriclecub: <DanL> I dont have much interest since I'm too technical-leaning to debate the nontechnical side. What about though...
<Short answer though, (ignoring the long post) is that the eval number is a DEFINITE measurement of a FEN well beyond move 66.The eval number is a HEURISTIC for move 66 and it is a HEURISTIC for move 59...Rc7> and one last thing to add is that the eval number as a heuristic is less volatile for 59...Rc7 than it is for 66.h6 (this could definitely be interpreted in your favor). Addionally, the smaller uncertainty in the eval number as a heuristic for move 59...Rc7 also helps your argument. |
|
 |
|
|
< Earlier Kibitzing · PAGE 1684 OF 1784 ·
Later Kibitzing> |
Kibitzing
