Atla Doubles Referrals in last year according to some stats.

[Brett_Tabke]

Statmarket is reporting their tracking of Altavista referrals has doubled in the last year. I chaulk most of that up to the switch Alta made from a primary text based interface to more graphics. People now run with more graphics on at Alta, thus triggering the Statmarket counters. It would be interesting to know how much of the increase is due to that fact - 25%?

[seth_wilde]

I've never really been impressed by statmarkets stats, they never seem to line up with my own (I've heard that because of statmarkets large porn client base, their stats are skewed)

Web Snap Shot seems to better agree with what I see in my logs so I use them now. (not to mention it's free) They actually had AV falling almost 3% since the begining of June
[websnapshot.mycomputer.com ]

[metaman]

Hello all!

I am new to the forum. I look forward to exchanging knowledge here with all of you.

I was wondering if any of you knew about what appears to be a points system at AV Belgium? Is this simply a numerical display of AV results?

[altavista.advalvas.be ]

I see #1 rankings coming in anywhere from 6,000 to 35,000 (points?)

[seth_wilde]

Hi Metaman, thanks for joining at SEW

This definitely looks to be a numerical value for site relevancy.
It definitely needs further testing but I did notice something pretty interesting. It seems as those when testing 1 word search phrases, the less common the word on the net the higher the possible number.

normal competitive 1 word search terms
#1 for "pokeman" - 15208
#1 for "mp3" - 15785

competitive search terms that are probably more common

#1 for "travel" - 6469
#1 for "sale" - 7874

competitve search terms that are very common

#1 for "computer" - 268
#1 for "software" - 268

[JamesR]

I believe it is a comparison of the single document in relation to the entire web collection. This is really interesting. I am even seeing exact results as the American version on 3 trial searches. This could be a huge find.

[metaman]

JamesR, I noticed the same thing. I find it gives the exact same results only this point system is showing up as well. I'm sure they both use the same database and algo.

Seth, good observation.

The top 8 sites under a search for "internet" all rank at 351. I wonder what determines a higher rank when they are tied?

As far as key phrases go:

3 word phrases seem to range from 15000-up
"florida real estate" - 15897
"stuffed koala bear" - 85553

2 word phrases seem to go about 12000-up
"real estate" - top 9 all at 12871
"computer game" - 15462 (98202 documents found)
"mole rats" - 52928 (366590 docs found)

Those last two are puzzling. There is more competition for "mole rats" than for "computer game"? Hmm, that kind of throws things off if we're assuming that the rankings correlate with number of documents found (low numbers for high competition).

Possibly they correlate to the number of times that phrase is searched on. That would make more sense but your guess is as good as mine...

[NFFC]

>Those last two are puzzling. There is more competition for "mole rats" than for "computer game"?

On competitive terms av searchs as if the term has been enclosed in quotes.

[seth_wilde]

>Those last two are puzzling. There is more competition for "mole rats" than for "computer game"?<

actually this fits perfectly into my observation -

"computer game" which is very popular tops out at 15462 (which is a relatively low number for a 2 word phrase)

While "mole rats" tops out at 52928 which is over 3 times higher than "computer game".

edited to add --> (was looking as phrase popularity not page numbers)

Edited by: seth_wilde

[JamesR]

Try doing a comparison between the search phrase and the number of pages found. My theory based on previous reading is that the more unique the search phrase in comparison to the entire indexed database, the higher the score you will get.

[seth_wilde]

James when comparing these 1 word search phrases it pretty closely follows that theory, there are some exceptions, but I would guess that's becuase of the level of optimization involved. (radiology could have a higher number than pokemon, but since it's less competitive the pages aren't as well optimized)

1 word phrases

software - 268 (36,542,495)
computer - 282 (33,252,320)
travel - 6,469 (18,106,953)
archive - 6,979 (20,199,533)
sale - 7,874 (10,079,736)
mp3 - 14,050 (6,464,960)
radiology - 21,010 (368,795)
pokemon - 27,851 (622,300)

(ps - misspelled pokemon in my first post/search so the numbers are different.)

[metaman]

NFFC, thanks for clearing that up.

I can't think of any huge benefit of this "points system" other having some idea of how close you are to having a higher ranking, or having a numerical value as to how well your site is optimized. It will be interesting to monitor a new site I'm submitting as the pages show up in the index. I will keep track of what the numbers do and report here if I notice anything significant. Any other thoughts?

[NFFC]

>I can't think of any huge benefit

I find the greatest benefit is that you can monitor the effect of changes made to a page even when your ranking did not alter.

Edited by: NFFC

[seth_wilde]

I would still like to do some more testing. I was thinking that since we know that "popular" words get a lower numerical value, then maybe when marketing a site under multiple word phrases, you might benefit from adding extra focus to the "less popular words" (increasing your total numerical value)

[NFFC]

From AV:

"Keep in mind that in any query, rare words count more than common words. If someone searches for fruit and pomegranates, pages with the word pomegranates will appear at the top of the list (a technique known as "inverse document frequency"

pomegranates 16780

fruit 16324

Edited by: NFFC

[seth_wilde]

NFFC - Thanks for the quote directly from the "horses" mouth.

How much have you noticed the position of the word in the search phrase effects this? It seems as though if the word is at the back of the phrase (even if it is rare) it still doesn't receive as much pull as AV implies it does.

[NFFC]

Try here [sims.berkeley.edu].

[seth_wilde]

Thanks NFFC, I think this might actually help me with some other research I'm doing, But unfortunately I couldn't find anything about my previous question in the slide show.

[NFFC]

Not sure what you mean by "at the back of the phrase" but as regards rarity...

Assuming the title is of primary importance and is unable to be practically cloaked:

car 9,766,120

loans 1,478,495

Making the phrase "car loans" [no quotes].

Repeat the "loans" do not repeat the "car".

[seth_wilde]

I understand IDF, what I was talking about is the level of IDF pull in consideration with the position of the word in the search phrase.

for example

pomegranates and fruit - In this phrase IDF is very strong

fruit and pomegranates - in this phrase IDF is still present but not nearly as much as in the previous search phrase

Nothing earth shattering here. I think what we've compiled here just shows that for accurate optimization you can't optimize for a search phrase (like most optimization software suggests), you need to optimize for each word in the search phrase seperately (but with the same page).