homepage Welcome to WebmasterWorld Guest from 54.205.247.203
register, free tools, login, search, pro membership, help, library, announcements, recent posts, open posts,
Pubcon Platinum Sponsor 2014
Home / Forums Index / Google / Google News Archive
Forum Library, Charter, Moderator: open

Google News Archive Forum

This 57 message thread spans 2 pages: < < 57 ( 1 [2]     
New Google code?
Anybody know what this is?
cabbie




msg:118001
 9:54 am on Jul 19, 2003 (gmt 0)

On one page of serps I have checked is anew tag from google bemeath each result.The one under the number! is this
"Result 0 (12945892823490907288) stays 3.353761 "
Anybody else seen this?
Ps.Its an adult query

 

g1smd




msg:118031
 5:34 pm on Jul 20, 2003 (gmt 0)

Maybe it was a hoax. Has anyone verified seeing the data live on Google? Is there just the one claim from Cabbie?

Now you have a cache and history full of stuff that you might have trouble explaining to the boss/wife/kids.

doc_z




msg:118032
 8:52 pm on Jul 20, 2003 (gmt 0)

Unfortunately, I cannot verify the data.

However, there was a post in another forum (thanks LowLevel) with the following data:

Result 0 (17858621969398070262) stays 7.673061
Result 1 (3777368099750945527) stays 4.878522
Result 2 (4941276811107807664) was 4.719801 now: 14.346827
Result 12 (3097663625355827139) was 3.651012 now: 11.077334
Result 4 (15388425574117200660) stays 4.358095
Result 5 (1269396484255989084) stays 4.233190
Result 6 (3957142746074129145) stays 4.092563
Result 38 (1003105703434443285) was 2.165533 now: 7.362286
Result 3 (14730126769846848047) was 4.634323 now: 9.625390
Result 8 (9216847400548364205) was 3.902606 now: 9.271439

If this is not a hoax, one can think of the following explanation (yet another theory):

- 'Result n' ist the old position based on the old score (first number)
- Results which 'stay' don't change the position
- Results with a different scoring can change the position. They are ordered by some (unknown) function(old score, new score) and they occupy the 'free positions'

It would be helpful to have more of these data (if this is not a hoax).

LowLevel




msg:118033
 9:10 pm on Jul 20, 2003 (gmt 0)

If it's a hoax it's a well constructed one, because I can see a "logic" in the numbers.

cabbie




msg:118034
 12:11 am on Jul 21, 2003 (gmt 0)

Look I am REALLY sorry about the delay in answering your posts.I am away in an exotic tropical island at the moment and I promised my wife (and myself) I would not touch a computer.
I have snuck away and am in a internet cafe where it takes 5 mins for a page to unload.
I answered all the stickys.The code doesn't show for the query now at least not where I am.I did take a snapshot of the 20 result page.I Thought the last bit was word density for the keyword.Page rank did not seem to figure in the cocde at all.
Sorry for seeming to ignore your posts .
cheers Alan

plasma




msg:118035
 1:02 am on Jul 21, 2003 (gmt 0)

The number seems to be in octal, thatīs whz it is so long.

IITian




msg:118036
 1:14 am on Jul 21, 2003 (gmt 0)

>The number seems to be in octal, thatīs whz it is so long.

No. Octals will contain only 0-7, I spotted a few 9's. Moreover, conversion from decimal to octal makes it only slightly longer. For example, 2^32 which is slightly above 4 billion is represented by 4 followed by 9 digits in decimal scheme. In octal, it will be 4 followed by 10 zeros!

claus




msg:118037
 2:33 am on Jul 21, 2003 (gmt 0)

g1smd:
>> you have a cache and history full of stuff that you might have trouble explaining to the boss/wife/kids

*lol* you're absolutely right, did a few queries for ...some very competitive KWs myself and found absolutely nothing new under the sun.

LowLevel:
>> PR values are not between 0 and 10, that's just a logaritmic representation for the humans

Although it does not have to have a finite endpoint, a logaritmic scale has the characteristic, that it gets increasingly more difficult to "climb a step", the higher you are placed (as distances between steps increase). The figure 10 i have used along the way is just an example number to make the calculations easier. I have never heard of or experienced pages above PR10, but you're absolutely right that it might as well be entirely different numbers.

And of course, the figures don't need to have anything to do with PR at all, i'm just guessing like the rest of you - only i'm biased because of the patent thing. It could be that it has nothing to do with this, there's a possibility that it has, but also a possibility that it has not, i hope i have been clear on that.

doc_z:
>> there was a post in another forum (thanks LowLevel) with the following data

These data - if they are an example of the same thing that cabbie posted about - makes my previous post (#18) invalid, or at least it makes the constant values that i used impossible. I have not tested with other constant values, but it could be that other values would fit. I just don't see the point in trying right now.

>> It would be helpful to have more of these data (if this is not a hoax).

I agree on both. Now we know that something like cabbies original post have been posted in two places. The syntax of these two posts are so close that this double posting could not likely have happended a) un-coordinated or b) without a real reason. Neither post seems to have a query string attached, descriptions or url's of sites; we suppose they have been there, as that is what at least cabbie has stated. All we have are silly ID's and some odd numbers that relate to each other in some odd way. So, we have cabbie on an exotic island, unable to recreate the thing, and some unknown poster on "another forum". Now, if Google is really trying out something new, why would they post debug code to the www in the first place? People tend to notice such things, although only two seems to have done so. Two observations out of 150 million queries is a very very little fraction. It could be the summer heat, and it could be an error, it could even be that G just like to see all the fuss they can stir up - a practical joke or whatever. Or, cabbie... do you have a friend with you on the same cafe posting to another forum?

I'm sceptical. The numbers make some sense, but it's not hard to design a string of odd numbers with some odd relationship. Only, Google does not normally have text like "Result 0 " in the serps - they're normally not numbered - then again, none of this is normal.

- 'Result n' ist the old position based on the old score (first number)
- Results which 'stay' don't change the position
- Results with a different scoring can change the position. They are ordered by some (unknown) function(old score, new score) and they occupy the 'free positions'

-this seems like a better explanation than the one i came up with as it explains the missing number one.

/claus

mil2k




msg:118038
 5:33 am on Jul 21, 2003 (gmt 0)

The thing whatever it was is gone now as cabbie said. Didn't get to see it first hand. :(

claus




msg:118039
 11:01 am on Jul 21, 2003 (gmt 0)

- just had a closer look at the new numbers. Like doc_z stated above, the "was" and "stays" seems to hold the same information as the "Result XX", and the "now" seems to be able to move results:

Result 6 (3957142746074129145) stays 4.092563
Result 38 (1003105703434443285) was 2.165533 now: 7.362286
Result 3 (14730126769846848047) was 4.634323 now: 9.625390

Result 38 seems like it has been placed too high. It is lower than the "Result 3" on both the "was" and the "now" number. For Result 38 (and 12), a number of 7 (and 11) moves the result up, while for Result 3 and 8, a number of 9 moves the result down. So 9-something has negative influence, while both larger and smaller numbers have positive influence.

/claus

swerve




msg:118040
 2:57 pm on Jul 21, 2003 (gmt 0)

Very intersting....

Here are some of my observations (some of which have already been stated in this thread):

There is a direct correlation between "Result n" and the "stays/was" number. From the limited data, if we list each "result" in descending order of "stays" or "was", the results appear in "order" according to n (Result 0 will appear before Result 1, etc.) As such, it seems that "stays/was" is a relevance (ranking) score, not Pagerank - we all know that results are not ranked according to PR, even though it weights heavily in the scoring algorithm.

Given the use of the terms "was", "stays", and "now", we can infer that something happened between "was" and "now". As mentioned, this could be the re-ranking algoritm. Perhaps it also could be the application of other scoring factors, spam filters, etc.

A "now" value can affect rankings.
However, the data suggests that a "now" value can result in no change in position, an increase in position, or even a decrease in position.

"Now" values are always higher than "was" values Of course, we don't have enough data to be sure of this, but it holds true for the data we have. In fact, in these examples, "now" is always more than 2X "was". It is interesting combined with the previous point, even a now value higher than the was value can result in a decrease in position.

Ranking in the SERPs in not based on the "now" value While this is obvious, it begs the question: if "was" is a ranking score, and "now" affects the new ranking, why doesn't the debugging info (as it seems to be) display the "new" ranking score (the score used to order the top ten results)? If this is debugging info, isn't something missing?

None of the original top ten (Results 0-9) were able to move above each other, compared to the orignal ranking For example, in no case is Result 5 ranked higher than Result 4, etc.

Speculation: The "value" affects ranking differently depending on the original rank (result n). The lower that a site is ranked, the larger the impact that "now" can have. For example, the site ranked 38th moved into position 8 with a "now" of 7.36. Another site ranked 12th moved into 4th with a now of 11.07. When a site is already ranked highly, the now value doesn't have as much impact: a site already ranked 3rd, with a "now" value of 14.35, does not move into 1st or 2nd. Perhaps it is near-impossible for a top ten site to move up based a "now" value. It is possible, however, to drop based on a "now" value. The sites ranked 9 and 10 were previously ranked 3 and 8, respectively. With now values of 9.62 and 9.27, thse sites dropped - perhaps there is a minimum threshold for sites to retain their positions (ie. sites in the top ten must have a "now" of 10 or greater...)

It is all very interesting, but much more data is necessary to draw any real conclusions...

-swerve

ciml




msg:118041
 7:00 pm on Jul 21, 2003 (gmt 0)

Very interesting indeed swerve. Just one thing I'm not clear on; "None of the original top ten (Results 0-9) were able to move above each other" and yet "the site ranked 38th moved into position 8"?

Now that is very odd. #38 can move up to #8, but judging from the data you saw #9 wouldn't move up to #8.

Thanks for bringing this up cabbie; you've certainly caused a lot of whirring brains.

swerve




msg:118042
 9:50 pm on Jul 21, 2003 (gmt 0)

Just one thing I'm not clear on; "None of the original top ten (Results 0-9) were able to move above each other" and yet "the site ranked 38th moved into position 8"?

We don't have enough data to know if this would always (or usually) be true. I highlighted this observation because it is potential support for my speculation that the "now" value carries a lesser weight for sites that already rank highly. A now value of 7 was able to propel result 38 to position 8, but a now value of 14 was not enough to move Result 2 up at all. Again, nowhere near enough data to be sure of anything....

ikbenhet1




msg:118043
 10:01 pm on Jul 21, 2003 (gmt 0)

Simple observation: The sites are ranked bij the "was" value.
("stays" is the same value as "was" and "now")
(was:) 0) 7.67 1) 4.87 2) 4.71 3)- 4) 4.35 5) 4.2 6) 4.0
Of course everbody knows that, i just figured it out.

So obviously these sites are not re-ranked yet, they still need to be "updated" and are still ranking by the old ("was") values.

Chndru




msg:118044
 10:31 pm on Jul 21, 2003 (gmt 0)

This just could be the comparision of the estimates of PR values between two monthly updates. (say, Dominic and Esmeralda).

LowLevel




msg:118045
 12:39 am on Jul 22, 2003 (gmt 0)

Swerve wrote:


Speculation: The "value" affects ranking differently depending on the original rank (result n). The lower that a site is ranked, the larger the impact that "now" can have.

I agree with this, it seems to me that the new position is calculed taking in account the original (previous) position.

Speculation time:

Result 2 (was 4.719801 now: 14.346827) keeps its position. Starting at 4.719801, it needs a "now-value" of 14.346827 to stay at position 2.

Result 3 (was 4.634323 now: 9.625390) drops because, starting at 4.634323, it needs a now-value greater than 9.625390 (probably very near to the "14.346827" of result 2), to stay at position 3.

Result 38 (was 2.165533 now: 7.362286) is now in the top ten not just because its now-value is 7.362286, but also because it started at 2.165533, a very low value compared to top-ten values.

So, some parameters of the ranking formula/algorithm of "now-valued" pages could be "where the page started from" (was-value) and "how much it gained" (now-value - was-value).

This stuff is funny (and maybe even a hoax), but absolutely useless with no more datas to analyze. :-)

doc_z




msg:118046
 11:58 am on Jul 22, 2003 (gmt 0)

Just another theory (while waiting for additional data from cabbie or someone else):

The 'old' and 'new' score mentioned in msg#32 are perhaps related to 'deep' and 'fresh' data. Thus pages which are unchanged (same 'old' score) doesn't change the position, while pages which were changed, are re-ranked (for the free positions, which were occupied by these pages) by some function of the 'old' and 'new' score. The 'old' score is the normal score (calculated from all 100 factors in the usual way). The 'new' score could be calculated in a way that on-page factors getting more weight, while off-page factors (anchor text, PR) are getting less weight (to compensate the disadvantage for new pages with few incoming links).

Of course, this is very speculative.

michael heraghty




msg:118047
 4:32 pm on Jul 22, 2003 (gmt 0)

This has been my favourite posting on the Google News forum so far!

It reminds me of a science fiction novel -- like you're trying to decipher a coded message from an alien intelligence. (Anyone ever read Stanislaw Lem's "His Master's Voice"?)

Hats off to all of you have actually managed to formulate theories from this small crumb that must have originated in an advanced civilisation -- in Mountain View!

cabbie




msg:118048
 5:32 am on Jul 27, 2003 (gmt 0)

If anybody is still interested this is the complete results I got from my original query.

Result 0 (12945892823490907288) stays 3.353761
Result 1 (12613868362284195648) stays 3.084599
Result 2 (4248721765241641425) was 2.730382 now: 7.290199
Result 3 (8788546316820000685) stays 2.716327
Result 4 (17634787461871553436) was 2.715453 now: 6.230899
Result 5 (4162163671003097492) stays 2.658952
Result 6 (16365760213835556644) stays 2.658946
Result 7 (16952869009410310664) stays 2.415777
Result 8 (9289898448292765357) stays 2.268252
Result 9 (12601404847159071662) was 2.256268 now: 5.260049
Result 10 (691312429157068477) stays 2.203274
Result 11 (17959278080108510766) stays 2.152111
Result 12 (7072029970207894277) stays 2.135290
Result 47 (787381731018850063) was 1.545682 now: 4.089366
Result 50 (1277532804845327735) was 1.528751 now: 3.764169
Result 15 (10958332758791235226) stays 2.063933
Result 16 (17396588693104660420) stays 2.027930
Result 17 (12197222240661626338) stays 1.963704
Result 18 (11610824377450667500) stays 1.787976
Result 20 (10753981164858177399) stays 1.786239

There was no result 19

jcoronella




msg:118049
 5:45 am on Jul 27, 2003 (gmt 0)

The chances of figuring out with any certainty what the numbers represent seems low. Whatever the metric is, the interesting aspect of all this to me is that a programmer put debug code into the search server.

To me this indicates that he/she was debugging on the fly calculations, or on the fly application of new data and wanted to see the old results vs the new results.

A rolling update is supposed to be coming, and this is just a further indication that it may have already arrived.

The Dance is dead. RIP

doc_z




msg:118050
 8:09 am on Jul 27, 2003 (gmt 0)

cabbie,

do you know which of the results had a fresh tag?

cabbie




msg:118051
 8:41 am on Jul 27, 2003 (gmt 0)

results 1
results 3
results 4
results 7
results 11
results 50
results 16
results 17
results 18 all had 17th July fresh tags

vincevincevince




msg:118052
 12:05 pm on Jul 27, 2003 (gmt 0)

FYI:

***** Regression Analysis *****

Response variate: code
Fitted terms: Constant + result + now_stays + result.now_stays


*** Summary of analysis ***

d.f. s.s. m.s. v.r. F pr.
Regression 3 1.693E+38 5.642E+37 2.00 0.154
Residual 16 4.505E+38 2.815E+37
Total 19 6.197E+38 3.262E+37


*** Estimates of parameters ***

estimate s.e. t(16) t pr.
Constant 1.19E+19 5.33E+18 2.24 0.040
result 1.20E+17 4.69E+17 0.26 0.801
now_stays 3.81E+16 1.27E+18 0.03 0.976
result.now_stays -8.50E+16 1.21E+17 -0.70 0.491

***** Regression Analysis *****

Response variate: result
Fitted terms: Constant + code + now_stays + code.now_stays

*** Estimates of parameters ***

estimate s.e. t(16) t pr.
Constant 20.7 14.6 1.41 0.177
code -3.63E-19 1.24E-18 -0.29 0.773
now_stays 1.04 3.79 0.27 0.788
code.now_stays -2.31E-19 3.30E-19 -0.70 0.494

So, we can see pretty much no correlation - the nearest we get is a factor at p 0.5 relating a combination of the code and now/stays value against the result.

doc_z




msg:118053
 12:31 pm on Jul 27, 2003 (gmt 0)

So, we can see pretty much no correlation - the nearest we get is a factor at p 0.5 relating a combination of the code and now/stays value against the result.

vincevincevince,

as far as I understood, you want to proof that the 'ID' is not related to the rest of the parameters. As most of the people, I think the first number is indded some kind of ID (even without your analysis). However, your results doesn't prove this. The reason is that your are just proving that a special ansatz failed to show a correlation.

At least as far as I can see without any explanation of the acronyms, i.e. d.f. = degree of freedom ...

vincevincevince




msg:118054
 12:54 pm on Jul 27, 2003 (gmt 0)

it doesn't show that the code is and ID, but it doesn't show that the code isn't an ID

alxdean




msg:118055
 1:12 pm on Jul 27, 2003 (gmt 0)

roflmao
vincevincevince, did you forget to take your medication again?

doc_z




msg:118056
 8:42 pm on Jul 27, 2003 (gmt 0)

I tried to analyze the data to see if there is a correlation between the status ('stays' or 'was now') and the time when the page was changed (msg#46).

If this theory would be true, pages with 'stays' should be unchanged (compared to some older version of the page) while 'was now' pages should have changed the content. To examine this, I looked when the pages were last modified (because I haven't older versions of these pages). Pages where the content have changed should by modified recently (compared to the date of the search). Pages which are unchanged could have either an older 'last modified' date or a new one (because they are changed after the date of the search or just the date was changed, but the page is still unchanged).

Here are all results I got:

Result 0 (12945892823490907288) stays 3.353761:?
Result 1 (12613868362284195648) stays 3.084599: Last-Modified: Thu, 10 Jul 2003
Result 2 (4248721765241641425) was 2.730382 now: 7.290199:?
Result 3 (8788546316820000685) stays 2.716327: Last-Modified: Sun, 08 Jun 2003
Result 4 (17634787461871553436) was 2.715453 now: 6.230899:?
Result 5 (4162163671003097492) stays 2.658952:?
Result 6 (16365760213835556644) stays 2.658946:?
Result 7 (16952869009410310664) stays 2.415777: Last-Modified: Tue, 15 Jul 2003
Result 8 (9289898448292765357) stays 2.268252: Last-Modified: Thu, 24 Jul 2003
Result 9 (12601404847159071662) was 2.256268 now: 5.260049: Last-Modified: 25 Jul 2003
Result 10 (691312429157068477) stays 2.203274:?
Result 11 (17959278080108510766) stays 2.152111:?
Result 12 (7072029970207894277) stays 2.135290: Last-Modified: Wed, 25 Jun 2003
Result 47 (787381731018850063) was 1.545682 now: 4.089366:?
Result 50 (1277532804845327735) was 1.528751 now: 3.764169: Last-modified: Sat, 26 Jul 2003
Result 15 (10958332758791235226) stays 2.063933: Last-Modified: 02 Apr 2003
Result 16 (17396588693104660420) stays 2.027930:?
Result 17 (12197222240661626338) stays 1.963704:?
Result 18 (11610824377450667500) stays 1.787976: Last-Modified: Thu, 17 Jul 2003
Result 20 (10753981164858177399) stays 1.786239: Last-Modified: Wed, 11 Jun 2003

-----------------------------------------------------
Result 0 (17858621969398070262) stays 7.673061: Last-Modified: Tue, 14 Jan 2003
Result 1 (3777368099750945527) stays 4.878522: Last-Modified: Tue, 14 Jan 2003
Result 2 (4941276811107807664) was 4.719801 now: 14.346827: Last-Modified: Mon, 21 Jul 2003
Result 12 (3097663625355827139) was 3.651012 now: 11.077334:?
Result 4 (15388425574117200660) stays 4.358095:?
Result 5 (1269396484255989084) stays 4.233190: Last-Modified: Thu, 27 Mar 2003
Result 6 (3957142746074129145) stays 4.092563: Last-Modified: Thu, 27 Mar 2003
Result 38 (1003105703434443285) was 2.165533 now: 7.362286:?

Unfortunately, there are a number of results where I didn't got a date (marked by an '?'). There is no old/unchanged (i.e. mid of June) page with an 'was now' status. Therefore, no result violates the theory. However, there are too few data to draw conclusions. Thus the theory is still very speculative.

gmoney




msg:118057
 11:37 pm on Jul 27, 2003 (gmt 0)

There is a direct correlation between "Result n" and the "stays/was" number.

True. This leads me to believe that the was/stay values are used to determine the old rankings (Result n). The data also hints that the was/now values are used somehow to determine the current rankings.

I tried to arrive at a function of "was" and "now", f(was,now), that would predict the order of the SERPs shown a but I was unsuccessful. However, I did notice a trend which is that the ratio of (now/was^2) was quite high for the big gainers and quite low for the big droppers.

(big gainer) Result 38 was 2.165533 now: 7.362286 : (now/was^2)= 1.57
(big dropper) Result 3 was 4.634323 now: 9.625390 : (now/was^2)= 0.45

(big gainer) Result 47 was 1.545682 now: 4.089366 : (now/was^2)= 1.71
(big gainer) Result 50 was 1.528751 now: 3.764169 : (now/was^2)= 1.61

The other (now/was^2) ratios (for rankings that didn't change much) were all between 0.6 and 1.04.

Of course, this analysis is on meager data but I thought it was interesting.

This 57 message thread spans 2 pages: < < 57 ( 1 [2]
Global Options:
 top home search open messages active posts  
 

Home / Forums Index / Google / Google News Archive
rss feed

All trademarks and copyrights held by respective owners. Member comments are owned by the poster.
Home ¦ Free Tools ¦ Terms of Service ¦ Privacy Policy ¦ Report Problem ¦ About ¦ Library ¦ Newsletter
WebmasterWorld is a Developer Shed Community owned by Jim Boykin.
© Webmaster World 1996-2014 all rights reserved