Hope you'll be there for our next game against Arno Nickel starting sometime later this year.

Tabanus, I think, agrees w/me. I don't think the rest of you are agreeing w/me, actually, but Tabanus, I am ready to trust on this subject. Note, that he is impatient with the extreme technical turn this debate has taken, as am I: <DanLanglois: the discussion tends to devolve into a pendandic debate about search algorithms & such>.

Mueller & hms123 have been very technical, in their posts, now you're joining them--on to your query: <sentriclecub: <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 >

I say that the eval number 6.26 represents the score of the move 59...Rc7. Of course, I haven't changed my position: <DanLanglois the eval number (6.26) is .. telling us that .. the position after 59...Rc7 is 6.26.>

I think it's clear that you're disagreeing w/me. Why is my rather carefree general statement that the eval number 6.26 represents the score of the move 59...Rc7 so inadequate to you? Because you flat disagree.

<sentriclecub: Here is my proof..> Well, your proof is irrelevent. I haven't offered any proof for my own view. If you understand why proof is irrelevent here, you will get my point. It doesn't matter how the number is generated, see? We have a number. The point now, is to disregard the details--like how it was generated, and just use the number. Question (my question) is: How is the number properly used.

Look. I'll make up a number. 4.00. I pulled it out of the sky. You know that you can't trust it. I'm telling you--you can't trust it I made it up. But if I type this:

(4.00): 59...Rc7 60.Kd4 Rf7 61.Kxc4 Nf2 62.Rd4+ Ke7 63.Rxa5 Ke6 64.h5 Rc7+ 65.Kb4 Rc1 66.h6

then, what is my message to you here? My message is 59...Rc7 is a 4.00 move. That is the proper interpretation of the number. Do you agree?

yes, you are leaning in my direction, here. The terminology is rather offputting. I don't panic at the use of the term 'heuristic', for example, but I'm skeptical of the need.

Hallelujah, I see you accepted that evaluation can be assigned to a position. But you're still incorrect. EACH POSITION along the line you quoted form RV has valuation 6.26 (and not only in that line, but way past it, likely past the 25 plies from the start -- due to extensions in search of quiescence).

Since you announced your unwillingness to engage in algorithms, I'll be brief. Simply put, if there was any position with black on-the-move along this line with a lower valuation, that number would propagate down the line to the start. Similarly, if any position with white on-the-move had higher valuation, that number would be used.

Okay, now that it has been established that the 6.26 applies to each node of that line, note that the conditions under which 6.26 has been computed differ for each node. The most important difference is that each position was examined at different depth. Also, the branching strategies (telling the engine what next positions to explore in each stage in the analysis) may differ depending on the depth form the starting position, the current position, and the valuation adjustments may depend on a number of other factors (see <g.mueller>'s post -- *he* knows what I am talking about, and so do many others).

Time for another experiment: from the current position (one ply past the previous example I gave), I run 20 plies analysis by Hiarcs which produced this line:

(+3.44) 57. Re4 Rxe4 58. Kxe4 Nf2 59. Ke3 Ng4 60. Kd2 Kd4 61. Rb5 c3 62. Kc1 Ne5 63. Rxa5 Nd3 64. Kd1

I advanced the line by 10 plies and run 10-ply analysys. Bingo, (+3.44). One more ply forward, 9-ply analysis. Bingo. One more ply, 8-ply analysis. Bingo.

QED (although I think I got lucky here by not running into possible consequences of the aforementioned heuristics that determine the processing of each position).

Now consider this: towards the end of the line, the valuation is really unreliable. Who'd believe 10, 9, 8 -ply evaluation? (for example, when I advanced 10 plies from the starting position and run 20-ply analysis, now I got +5.32).

Yet the same number propagated all the way back to the start of the line is now considered a reliable indicator of the valuation of the position (and, yes, of what move to select to maintain that valuation). Think about why this belief in reliability is (usually) justified, and perhaps you'll figure out the rest.

<I have been condescendingly corrected by many here, but i'm the one who is right.>

<Well, your proof is irrelevent.>

Usually that's a bad sign...

this is the proof...

<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...

All that technobabble was not relevant at all, but it did trip you up and expose an issue again.

Per <whatthefat> pointed out, your enthusiasm and exuberance hinders communication. Try to slow it down a little. You are sometimes too fast to respond, giving the impression that you aren't a good listener and maybe are inattentive to details.

I don't have the willingness to conform to discussing this <eval number discussion> in a non-technical manner, however I do like observing it, as long as I'm not in it.

I'll make a couple responses, but wish to opt out immediately after this post.

<I say that the eval number 6.26 represents the score of the move 59...Rc7> Yes you are 100% true. See the super short post I wrote that begins with "Short answer though..."

As for the rest of what you wrote, I agree with the substance of what you're saying.

I think your creative mind and stubbornness to stick to your gut instinct is a critical quality to have in this team game. I give you a lot of credit with being such an outspoken advocate of 38. h4

With all this said, I feel ready to opt-out, but will continue reading and might make a comment or two when I see fit, but no more tonight.

yes as a heuristic, see the "short answer though" post

Bwah, hah, hah, so he used HEURISTIC as a short for "result of a heuristic method". Big deal, anyone familiar with the vernacular can understand what he meant.

This is what's just outrageously wrong :-)>

what do you think about my idea for a interface that shows graphically the first 10 levels of nodes (so about 4^10th total) and at each node (corresponding to a move or position, you chose) you can tweak the priority.

If you assign 100% priority to the root node (current position) then Rybka will handle everything past it and assign priorities per its MO

I think in Fritz GUI you can add a double questionmark "59...Kg1??" to a move and the machine wont analyze any nodes past it right? Even if I'm mistaken, follow along.

What if you could say assign it 1.9 question marks. If a regular move (a move without question marks) is normal priority. Maybe let the 1.9 question marks mean for Rybka to spend only 1/20th of the time on it as it would have normally.

i.e. (if it spends 2 * 10^(x-15) seconds) evaluating it before the GUI announces it reached a new ply level, now you can limit the amount of time it spends on that 1 subtree, thus speeding up (in this case only infinitesimally) the rate at which Rybka annouces it reached a new ply.

What if you could enable the "manage subtrees" checkbox to let you adjust up or down how many nodes in a particular subtree that Rybka evaluates before it gets the go-ahead to report that it reached a new ply.

That's my thought and supposedly the Aqua interface of Rybka 3 has something similar.

This arises from the fact that when you set the PV setting on rybka from 1 to 3 or 4, you make its ply-reaching rate CONSIDERABLY slower.

Not because you decreased the kn/s but because you in effect told the machine to spend more time (seconds) exploring nodes in the subtrees of 1st level (immediately below root level) subtrees.

So in effect, if you set the PV setting from 2 up to 3, you told Rybka to explore that 3rd move's subtree more fully before getting the thumbs up to report to the GUI a new ply level has been reached.

Please understand your ideas and the topics you raise reflect positively about you. You seem to be the kind of guy I imagine to be a Matrix movie fan.

I really hope that you understand you're not wrong in substance. I like both your receptiveness and skepticism of numbers! Everybody should always be both.

I think it's worth pausing here, to agree with you, or at least reassure you that I understand what you are trying to say (I wouldn't say it this way, because, well, you'll see).

<Okay, now that it has been established that the 6.26 applies to each node of that line>

Usually that's a bad sign...
>

Yes, that is usually a bad sign. We agree! :-)

sigh

<blah blah Bingo blah blah Bingo blah blah Bingo QED>

Look, I do appreciate the care you have put into this. But I can manage to follow your argument just fine. I understood the *previous* 'experiment' just fine. You generate a number, and then you flip some dials, toggle some switches, wait around, and the number changes. I could have predicted that for you. This is not the diffcult point at all. I keep offering to make up a number (4.00) and then change it (2.96)--look! BINGO! the number changed. got it.

<Now consider this: towards the end of the line, the valuation is really unreliable. >

the number is the number. See? A number is just a number. When you start talking about it being 'really unreliable', what you mean to say is that you are doing something really unreliable w/it. What are you doing with it?

<Yet the same number propagated all the way back to the start of the line is now considered a reliable indicator of the valuation of the position (and, yes, of what move to select to maintain that valuation). Think about why this belief in reliability is (usually) justified, and perhaps you'll figure out the rest.>

I wonder if you're trying to tell me my own point, here?!!!!? eh!?? hold on. What are you saying?

<Yet the same number propagated all the way back to the start of the line is now considered a reliable indicator of the valuation of the position (and, yes, of what move to select to maintain that valuation). >

touchy. Then I guess I haven't been condescendingly corrected?

All that technobabble was not relevant at all, but it did trip you up and expose an issue again. >

I'm flabbergasted by this--you're somehow lecturing me on how right I was to say 'your proof is irrelevant'. You respond: 'All that technobabble was not relevant at all..'