scary article

[Parsad]

"must buy stocks today" yields 113,000,000 hits

 

Did you try "must buy stocks today 2010"?  Yields only 510,000 hits.  You do know that the World Health Organization now uses Google to track the spread of the flu virus based on the number of posts coming up in various regions and searches on the web?  Cheers!

[Munger]

now we debate semantics????

 

i'll take the higher road (it is much less traveled) and agree to disagree

 

best

[opihiman2]

"must buy stocks today" yields 113,000,000 hits

 

Did you try "must buy stocks today 2010"?  Yields only 510,000 hits.  You do know that the World Health Organization now uses Google to track the spread of the flu virus based on the number of posts coming up in various regions and searches on the web?  Cheers!

 

I have a hard time believing that.  I require citation here. 

[opihiman2]

To see how pointless this game is, I did a search for "economic recovery" 2010 vs. "economic collapse" 2010.  Wow, 10x the result set in favor of recovery!  Sweet!  I guess that settles it, then.   ::)

 

We could go on and on using various semantics for recovery and collapse, but I hope this exercise shows how futile it is to use Google result sets to prove an argument.

[Munger]

I have a hard time believing that.  I require citation here.

 

Opihiman2 -- if you type "buy stocks 2010," the result is 52,200,000.   As the debate devolves into a matter of semantics and how cute one can be with a google search, we have entered the land of silliness.

[Partner24]

Amen.

 

Munger, this community works best with positive contributions and that includes valid criticism.  Criticism in socially unskilled ways is best left to other forums.

 

-O

I've got an easy way to resolve this.  Do an advanced Google search for a few of the words you've entered in the sentences below and add "2010".  How many hits do you get?

 

C'mon -- you're smarter than that my man.

[rick_v]

I tend to see this argument in a completely different light.

 

If you think about it statistically... there would be no bubble unless a majority of the capital deployed believed the hype. So its a matter of fact that the amount of people who did not believe the hype were a minority. But that to me is not as interesting as this.

 

To write a few sentences about something is very easy to do, Nostradamus was a pro at it. You can structure visions in a way that will formulate one way or another. As investors we need to focus on the guys that actually deployed their capital and gained from the downturn.

 

Now I recall in the 2009 Berkshire Annual Meeting, Munger made a comment that kind of pissed me off. He said that a Manager which was all cash in 2008 was not the Manager he would want to have. I strongly disagreed with that statement as I believe that a manager in tune with the markets who had the understanding of the fundamentals relating to the events could have easily made a decision to scale back and even short the obvious.

 

And those are the guys I admire:

 

#1 - Michael Burry

#2 - Albert Friedberg

#3- Tom Kaplan

 

There are others as well but these are guys you won't really hear about have made billions of dollars combined, and this was the foundation of their investment strategy... They knew what was going on, deployed capital accordingly, and killed it!

 

Now back to Parsad's point, what impressed me about Prem is not that he talked about it but that he actually performed amazing in 2008.

 

Deployment of Capital in the real world is so different from practicing in theory. I cannot stress this enough.

[Dorsia1]

http://mohel.dk/grafik/andet/Someone_Is_Wrong_On_The_Internet.jpg

[Parsad]

There is a problem with using the Google methodology for argument rebuttal.  That result set you're looking at?  It's estimated.  If you continue to iterate through the result set, you might be surprised to find only a few thousand URLs.  Also, here's another big problem (and this, IMO, is much bigger than the cardinality of the data set problem) with using Google to prove something: key words can skew the result set.  For example, does a search on banana bread return a true estimation of the popularity of banana-bread?  Or does it just show how popular banana or bread is?  Or, maybe there are links that mention banana or bread in passing.  One way to resolve this is to use quotes to use the AND operator on the key words.  However, with more keywords, there is more subjectivity to the result set.  Also, how do you resolve that "Cheap large cap stocks" may be equivalent to "cheap large capitalization stocks", or that a keyword search on cheap large cap stocks may not even be about cheap large cap stocks but something entirely different?  The result set is not meaningful in these cases.  I would probably only use this kind of analysis if the keyword was a single word or a compound word (for example, "banana bread").  Anything beyond that is suspect.

 

You'd think that was correct Ophi, but you may be partially incorrect.  Using the same initial words in the search, but add in "2007" (pre-recession) instead of "2010":

 

Entered "Coming Depression 2010".  Result:  96,000,000 hits

Entered "Coming Depression 2007".  Result:  6,210,000 hits

 

Entered "Double-dip recession 2010".  Result:  1,330,000 hits

Entered "Double-dip recession 2007".  Result:  437,000 hits

Entered "Recession 2007".  Result:  17,400,000

 

Entered "Cheap large cap stocks 2010".  Result: 61,000 results

Entered "Cheap large cap stocks 2007".  Result: 518,000 results

 

Entered "must buy stocks today 2010".  Result:  510,000 hits

Entered "must buy stocks today 2007".  Result:  430,000 hits

 

As you can see, the number of hits for the word "coming depression" using 2010 is multiples higher than if you entered 2007.  The same result for "cheap large cap stocks" for 2010, was multiples lower than 2007.

 

There was no possibility of a double-dip recession in 2007, as we had not entered the first one yet.  Thus the hits were less.  I would think most of the hits for 2007 had to do with the initial words "double-dip recession", as Google's search alogrithm accumulates all hits on the words, past and present.  Why did this not happen to the other two?

 

Your reasoning becomes apparent in the fourth test "must buy stocks", where the date has no real effect on the number of hits.  It seems as though Google's search alogrithm is hitting the first four words and there is little difference between 2010 and 2007 in the number of hits.  Cheers!

 

 

[opihiman2]

I have a hard time believing that.  I require citation here.

 

Opihiman2 -- if you type "buy stocks 2010," the result is 52,200,000.   As the debate devolves into a matter of semantics and how cute one can be with a google search, we have entered the land of silliness.

 

Well, in that post I was referring to the WHO's usage of Google for epidemiology.  I just find that an organization of WHO's caliber would not use a faulty idea.  

[opihiman2]

There is a problem with using the Google methodology for argument rebuttal.  That result set you're looking at?  It's estimated.  If you continue to iterate through the result set, you might be surprised to find only a few thousand URLs.  Also, here's another big problem (and this, IMO, is much bigger than the cardinality of the data set problem) with using Google to prove something: key words can skew the result set.  For example, does a search on banana bread return a true estimation of the popularity of banana-bread?  Or does it just show how popular banana or bread is?  Or, maybe there are links that mention banana or bread in passing.  One way to resolve this is to use quotes to use the AND operator on the key words.  However, with more keywords, there is more subjectivity to the result set.  Also, how do you resolve that "Cheap large cap stocks" may be equivalent to "cheap large capitalization stocks", or that a keyword search on cheap large cap stocks may not even be about cheap large cap stocks but something entirely different?  The result set is not meaningful in these cases.  I would probably only use this kind of analysis if the keyword was a single word or a compound word (for example, "banana bread").  Anything beyond that is suspect.

 

You'd think that was correct Ophi, but you may be partially incorrect.  Using the same initial words in the search, but add in "2007" (pre-recession) instead of "2010":

 

Entered "Coming Depression 2010".  Result:  96,000,000 hits

Entered "Coming Depression 2007".  Result:  6,210,000 hits

 

Entered "Double-dip recession 2010".  Result:  1,330,000 hits

Entered "Double-dip recession 2007".  Result:  437,000 hits

Entered "Recession 2007".  Result:  17,400,000

 

Entered "Cheap large cap stocks 2010".  Result: 61,000 results

Entered "Cheap large cap stocks 2007".  Result: 518,000 results

 

Entered "must buy stocks today 2010".  Result:  510,000 hits

Entered "must buy stocks today 2007".  Result:  430,000 hits

 

As you can see, the number of hits for the word "coming depression" using 2010 is multiples higher than if you entered 2007.  The same result for "cheap large cap stocks" for 2010, was multiples lower than 2007.

 

There was no possibility of a double-dip recession in 2007, as we had not entered the first one yet.  Thus the hits were less.  I would think most of the hits for 2007 had to do with the initial words "double-dip recession", as Google's search alogrithm accumulates all hits on the words, past and present.  Why did this not happen to the other two?

 

Your reasoning becomes apparent in the fourth test "must buy stocks", where the date has no real effect on the number of hits.  It seems as though Google's search alogrithm is hitting the first four words and there is little difference between 2010 and 2007 in the number of hits.  Cheers!

 

 

 

Hi Parsad,

 

There is some mis-communication going on here.  Entering "Cheap large cap stocks 2007" into Google.com returned zero results for me.  Are you entering the quotes?  That makes a big difference.  Without the quotes, I actually received 200k plus more result set size for Cheap large cap stocks 2010 than 2007.  I think this is a domain issue.  You must be using Google.ca.

 

EDIT:

 

I've also tried variations, because "Cheap large cap stocks 2007" just seems too constrictive.  So, I tried "Cheap large cap stocks" and entered 2007 and 2010 to have the search engine parse those separately, but look for "cheap large cap stocks" in either period.  Still, I get about 8x more results in 2010.

[Munger]

moving from silly to insane

 

 

 

 

[Munger]

opihiman shows a valid distinction in my humble opinion

[Parsad]

Sorry Ophi, it wasn't WHO, but the CDC.  Cheers!

 

http://www.scientificamerican.com/article.cfm?id=google-flu-trends-on-par-with-cdc-data

[rick_v]

You guys do realize that google has an option to filter search results based on the first crawl date right? lol

 

http://www.google.ca/search?q=%22subprime+problem%22&hl=en&sa=X&ei=4YNtTJK-L4-WvAPkk-GWDQ&ved=0CAwQpwU&source=lnt&tbs=cdr:1,cd_min:2003-1-1,cd_max:2005-1-1

 

http://www.google.ca/search?q=%22subprime+problem%22&hl=en&sa=X&ei=8INtTK7SFce2ngfeobyBCA&ved=0CBAQpwU&source=lnt&tbs=cdr:1,cd_min:2005-1-1,cd_max:2008-1-1

 

 

Cheers?

[opihiman2]

You guys do realize that google has an option to filter search results based on the first crawl date right? lol

 

http://www.google.ca/search?q=%22subprime+problem%22&hl=en&sa=X&ei=4YNtTJK-L4-WvAPkk-GWDQ&ved=0CAwQpwU&source=lnt&tbs=cdr:1,cd_min:2003-1-1,cd_max:2005-1-1

 

http://www.google.ca/search?q=%22subprime+problem%22&hl=en&sa=X&ei=8INtTK7SFce2ngfeobyBCA&ved=0CBAQpwU&source=lnt&tbs=cdr:1,cd_min:2005-1-1,cd_max:2008-1-1

 

 

Cheers?

 

Really, on the crawl date?  That would be very helpful.  I did not know that.

 

EDIT:

 

Actually, hah, yeah, I did know that.  Sorry, I was thinking about the indexed date vs crawl date.  Those should be negligible.

[rick_v]

Argument Settled?

 

:)

 

Its either Crawl date or actually returns the absolute dates the content was written as opposed to just using a search with keywords which as Munger pointed out is extremely unreliable and hits really mean nothing...

 

Now with regards to the Flu, what these government bodies do is actually measure the amount of people typing the word "flu" into google and that forms a matrix which allows them to gauge spikes which can then be correlated to IP addresses in geographic locations.

 

When certain IP's from certain locations type the word flu that means the flu is spreading and they can get a head start on dealing with it because its basically instantaneous.

 

Get it?

 

 

[Parsad]

You guys do realize that google has an option to filter search results based on the first crawl date right? lol

 

Ah...no!   ;D  Thanks!  Based on that using the crawl date January 1, 2010 to August 1, 2010:

 

Entered "Coming Depression".  Result:  40,500 hits

Entered "Double-Dip Recession".  Result: 1,080,000 hits

Entered "Quality Large Cap Stocks".  Result:  47 hits

Entered "Cheap Large Cap Stocks".  Result:  0 hits

Entered "Cheap Big Cap Stocks".  Result:  40 hits

Entered "Quality Big Cap Stocks".  Result:  4 hits

Entered "Economic Recovery".  Result:  2,460,000 hits

Entered "Economic Collapse".  Result:  820,000 hits

 

Using crawl date January 1, 2007 to August 1, 2007:

 

Entered "Coming Depression".  Result:  668 hits

Entered "Double-Dip Recession".  Result: 96,800 hits

Entered "Quality Large Cap Stocks".  Result:  7 hits

Entered "Cheap Large Cap Stocks".  Result:  0 hits

Entered "Cheap Big Cap Stocks".  Result:  0 hits

Entered "Quality Big Cap Stocks".  Result:  0 hits

Entered "Economic Recovery".  Result:  238,000 hits

Entered "Economic Collapse".  Result:  224,000 hits

 

Cheers!

[Cardboard]

When you guys are done playing with Google please read this from Fairfax 2007 annual report:

 

"Recently, we came across an interesting observation by the man who provided the intellectual underpinnings of “long term value investing” and to whom we are ever indebted. Ben Graham made the point that only 1 in 100 of the investors who were invested in the stockmarket in 1925 survived the crash of 1929 – 1932. If you didn’t see the risks in 1925 (very hard to do), it was very unlikely that you survived the crash! We think Ben’s observation may be relevant to what we have experienced in the past five years. We reminded you in our 2005 Annual Report that “Jeremy Grantham of Grantham Mayo said that of the 28 bubbles that they have studied in all asset categories (including

gold, silver, Japanese equities and 1929), this recent bubble in the U.S. stock market is the only one that has not completely reversed itself (just as it was about to in 2003, it turned and rebounded).”"

 

Then take a deep look at the actions that this same man has taken in Q2.

 

We have a bubble in Internet stocks, a bubble in treasuries and other yielding instruments (REIT's anyone?), a bubble in housing of most countries other than the U.S., countries deep in deficits and debt, a stock market that acts completely bizarre with HFT, an economy that has 500,000 new requests each week for unemployment.

 

I don't care how many people talk about it on the Internet or wherever, it remains that the reality is very fragile to any kind of shock. Very few are prepared or seem to accept the possibility.

 

This is not to say that JNJ is not a great bargain at this price. I think it is if you are looking for 10% returns. However, keep in mind that no brainers or very cheap stocks offering amazing rewards are now rare. Every time it is the case, something happens to change that.

 

Cardboard

[Munger]

the absolute absurdity continues...

 

"buy stocks" from 1/1/2010 - 8/1/10 yields 791,000 hits

[Munger]

thank you for bringing us back to reality cardboard

[Parsad]

Cardboard, you'll like this one:  

 

"Prem Watsa"

 

Crawl date - January 1, 2007 to August 1, 2007:  226 Hits

 

Crawl date - January 1, 2010 to August 1, 2010:  22,600 Hits

 

Cheers!

[opihiman2]

Argument Settled?

 

:)

 

Its either Crawl date or actually returns the absolute dates the content was written as opposed to just using a search with keywords which as Munger pointed out is extremely unreliable and hits really mean nothing...

 

Now with regards to the Flu, what these government bodies do is actually measure the amount of people typing the word "flu" into google and that forms a matrix which allows them to gauge spikes which can then be correlated to IP addresses in geographic locations.

 

When certain IP's from certain locations type the word flu that means the flu is spreading and they can get a head start on dealing with it because its basically instantaneous.

 

Get it?

 

 

Well, it depends on which side you're arguing.  :-)  Actually, no, because there is one thing Google, and actually web 2.0 in general, is notoriously bad at: metadata.  Web 3.0, the semantic web, is trying to address that.  Right now, there is a metadata tag for date in every HTML version spec that accounts for this, but from what I know, almost no author uses it.  For example, if you look at most news articles or opinion pieces, the authored date is clearly expressed along with the article; however, Google has no idea of knowing what that is.  Could it be a date referenced by the article that has no relation to the authored date?  How would it know?  It doesn't understand semantics.  Even with the author's best intention of letting people know what date the article was authored (without using the metadata date tag), the Google index of the article would lead to all sorts of inconsistencies in a search query using date filters.  For example, let's use this current debate over what period had most value for large cap stocks.  Suppose I wrote an article in 2010 on cheap large cap stocks; however, in that article I referenced 2007 as being a expensive year in the equities market.  Using that search query we talked about earlier, "cheap large cap stocks" 2007 or 2010, would return that article.  So, which query is right?  The metadata date tag would resolve this debate, but no one uses it, and from my understanding, Google's spiders don't even parse the tags and input those into the index.  I guess they are going to wait till web 3.0 becomes a reality, and a defined and sanctioned .RDF definition is being used.  Until then, it just bases the date on the first time the URL was indexed.  

 

EDIT: God, my writing is coming out like pure garbage.  I have to re-write for some clarity.

 

EDIT: By the way, the indexed date would resolve the 2010 and 2007 query I mentioned.  However, you could write an article in 2010 that discussed large cap values in 2007 and have the indexed date be in 2010.  This would lead to a similar inconsistency in the result set as we see earlier.

[Hoodlum]

Are you suggesting this is a sell signal for FFH?   ;)

 

Cardboard, you'll like this one:  

 

"Prem Watsa"

 

Crawl date - January 1, 2007 to August 1, 2007:  226 Hits

 

Crawl date - January 1, 2010 to August 1, 2010:  22,600 Hits

 

Cheers!

[Munger]

just to further point out the insanity

 

1/1/2007 - 8/1/2007 "buy real estate" (the supposed peak of the real estate bubble) yields only 210,000 hits

 

1/1/10 - 8/1/10 "buy real estate" yields 2,350,000 hits