Forum Moderators: open
Result 3 (8788546316820000685) stays 2.716327
Result 4 (17634787461871553436) was 2.715453 now: 6.230899
I wonder if we can extract any sort of meaning from those numbers. The one in parentheses looks like an ID, but seems awfully long even for a 4 billion page index. The other numbers might have something to do with rankings...hmmm.
[edited by: ciml at 5:19 pm (utc) on July 19, 2003]
[edit reason] No private messages in the threads, please. [/edit]
Cabbie, would you mind checking the pr of the examples using the toolbar and report back if there's any similarity between the second numbers and their pr?
seems like a re-ranking to me, it's not april, is it?
/claus
<edit>given that 4 is max, that number 0 site has a mighty high PR... yes, debugging is the word for it - we probably can't recreate</edit>
[edited by: claus at 4:57 pm (utc) on July 19, 2003]
I actually think this is some sort of total rank score, as the sites are all listed in decreasing order by this number. Could reflect on-the-fly pagerank though.
I'm leaning towards #3.
I got this:
Result 0 (12945892823490907288) stays 3.353761 "
1+2+9+4+5+8+9+2+8+2+3+4+9+0+9+0+7+2+8+8 = 100
(a) (b) (c)
100 / 3.353761 = 29.81727
(a) (c) (b)
100 / 29.81727 = 3.353761!
A / B = C
A / C = B
B=?
C=?
ok my math teacher would laugh at me... for not beeing able to solve those last 2 equations.
1+2+9+4+5+8+9+2+8+2+3+4+9+0+9+0+7+2+8+8 = 100
Interesting, but seems conicidental. It doesn't match a pattern with the sums of the other numbers, AFAICT.
(a) (b) (c)
100 / 3.353761 = 29.81727(a) (c) (b)
100 / 29.81727 = 3.353761!
This works for all numbers A, B, C.
<edit> i see your point, change the 100's with 50 or an other number and get the same result.
This is always true for constant A,B,C. Any A,B,C you pick will satisfy this rule.
Its the same as B * C = C * B = A.
Multiply both sides by the divisor and you'll see.
Your high school algebra teacher is crying right now. ;)
We have three numbers. The first one seems an ID, and I'm not interested about it.
The second number is descending. It could simply be the relevance
value of the web page, that is the value created on the fly by
Google, taking in account all the parameters (PR, page contents,
backlinks anchor text, penalties, etc.).
The third number is very interesting. For some pages it doesn't appear
at all.
Since it's very unprobable that the result of the calculation
produces exactly the same value of the second number, the absence
of the "now:" value for some pages leads me to think that the third
number is the result of a calculation that is not performed on all
the pages, but just on pages that meet some conditions.
*If* it's the re-ranking algorithm I read about, it would be
interesting to check if pages with "now:" value receive links by
other pages/sites in the SERP and if pages with "stays" flag don't
receive links by other pages/sites in the SERP.
I repeat: just speculating.
- i tend to agree on that. If we pretend that these debugging messages are actuallly NewScore being calculated, the two other numbers could make sense in this way:
The "was" figures = NewScore
The "now" figures = OldScore (~PR (ie. perhaps not only that, see part 2 of this post))
The "stays" figures = NewScore being equal to OldScore (relatively)
I still do not state that this is the case, only that it could be. We have these SERPS (that have not been confirmed since cabbies first posts, which is not surprising, since they were probably intended for internal use only):
Result 0 (12945892823490907288) stays 3.353761
Result 1 -missing-
Result 2 (4248721765241641425) was 2.730382 now: 7.290199
Result 3 (8788546316820000685) stays 2.716327
Result 4 (17634787461871553436) was 2.715453 now: 6.230899
...
From this list, it seems that result 2 has an OldScore of 7 and result 4 has an OldScore of 6 (rounded numbers). Further, result 0 and 3 "stays". This can be interpreted so that their OldScore and NewScore are relatively the same.
If we further pretend that my calculating example from the very long thread "Google's 2 rankings & you" [webmasterworld.com] holds vith respect to the values of the constants in the equations, then the max NewScore is 4. Assuming again that max OldScore is indeed 10, we can compare relative values.
Result 0: NewScore: 3.353761
Result 1: -
Result 2: NewScore: 2.730382 ¦ OldScore: 7.290199
Result 3: NewScore: 2.716327
Result 4: NewScore: 2.715453 ¦ OldScore: 6.230899Using 4 and 10 for max values, we see this re-ranking:
Result 0: NewScore(relative): 83.84% ¦ OldScore(relative): 83.84%
Result 1: -
Result 2: NewScore(relative): 68.26% ¦ OldScore(relative): 72.90%
Result 3: NewScore(relative): 67.91% ¦ OldScore(relative): 67.91%
Result 4: NewScore(relative): 67.89% ¦ OldScore(relative): 62.31%
Meaning that Result 4 has climbed and result 2 has dropped, and the rest stays.
Dolemite
>> The first number is in descending order for the data we have, which PR almost never is.
It could be, that the PR you see on your Gbar might not be the exact PR, rather a derived measure or approximation (perhaps just a rounded number) in stead. I mean: with three-something billion pages, and 1,000 results for large serps, a one-digit number seems like a somewhat poor measure to sort on. A figure like, say: 7.290199 seems more appropriate in this case than a figure of, say, 7. Anyway, even if this was the case, a PR7 would still be higher than a PR6.
So, what you are saying is in fact, that some other measure than PR seems to get higher weight already. This could be keyword density or something like that, i don't know, but i don't think i disagree with you either.
Whatever it is, it must be part of the "OldScore" concept for all this to work out. This, in turn, means that OldScore does not need to be exactly the same as Page Rank, rather PR + "other mechanisms" (where "other mechanisms" can be equal to zero). We do know, however, that PR is the key concept G uses to rank pages in the index (the index being the total amount of pages, and the SERPS being just a sample).
I added the comment "given that 4 is max, that number 0 site has a mighty high PR" to my first post in this thread. What i ment by that was that a score of 8 based on PR alone seems very high. Take the first digit of the "OldScore(relative)" above, and you see what i mean. A PR8 is not very usual AFAIK, but for an adult query, i don't really know, some of those sites are really big on backlinks, keyword density and most everything else.
Well, enough guesswork for now, i only wanted to illustrate my first post in this thread, and this is done. Still nobody knows and all options are open.
/claus
The highest third value in that list is 14.346827. Further, real PR values are not between 0 and 10, that's just a logaritmic representation for the humans. There's no need to use an human representation in algorithms.
In the mentioned list, pages are ordered by the second value, except for pages with a "now:" (newer?) value. Pages with a "now:" value sometimes loose positions, sometimes gain positions.
It's a bit difficult to explain it without showing the mentioned list of datas. Beside, my english is too poor. :-)
Oh cabbie Where are thou?
if posting keywords is not welcomed in this case them someone sticky them to me please...
this forum's charter [webmasterworld.com] says it all!
<added>
>someone sticky them to me please
jon, cabbie is the only one who has specifics. So instead of asking here at this thread, i'd better send him a sticky, if i were you. ;)
</added>
I certainly cannot.
Really curious because this is a very thrilling discovery.
If anyone else could confirm this and provide perhaps all 10 of them maybe others could help as well.
