(keitai-l) Re: Failure, schmailure, show me the money! (was Re: failure, succ

From: <hag_at_eatoni.com>
Date: 12/29/01
Message-Id: <20011229200908.77FDE264321@tapdance.eatoni.com>
| 
| > With reference to "walls" - I saw some stats. recently suggesting that
| > some huge percentage (80% ? something like that - feel free to post
| the
| > correct figure) of www page impressions are on just 60 sites.
| 

I'm not sure that this will help advance the discussion, but what the heck:

Most of the argument here can be subject to scientific, statistical
analysis. 

Quick Statement:

"open" or "closed" sites aside, there is always a distribution of
traffic to sites, usually a Zipf's law distribution. If you knew the
statistics of site access for official and non-official sites, you could
measure the height of the "wall" by looking at the difference (if any)
between the distribution of accesses to open and closed sites. The
exponent in Zipf's law might be different for the two, but probably not,
probably just the constant offset, which is a measure of the height of
the "wall".

Somewhat Longer Statement:

Even in a totally non-walled environment, a handful of sites get most of
the traffic. It says nothing about walls or whatever that 80% (say) of
traffic goes to the top 60 (say) sites. The same is true of any
language, natural language, or the language of accesses to hard disks,
"language" of access to web sites, whatever. For any natural language, for instance,
the top 50 most frequently used words will account for about 40% of the occurrence of
all words. In English, the top 5 words: "the, of, and, to, a" account
for nearly 17% of all word usage. The same is true of web sites, some
sites,   yahoo or google (say) get as much traffic as thousands or
millions of other sites combined. So, what is interesting is to look at
_deviations_ from the natural distribution of accesses, see if that
might be due to some systematic influence, like the presence of a
"wall".

To not risk further boring non-quants, I'll stop here.
Received on Sat Dec 29 22:15:49 2001