Re: Browser Stats

Benjamin Franz (snowhare@netimages.com)
Mon, 15 Sep 1997 07:48:04 -0700 (PDT)


Date: Mon, 15 Sep 1997 07:48:04 -0700 (PDT)
From: Benjamin Franz <snowhare@netimages.com>
To: www-html@w3.org
In-Reply-To: <199709150625.IAA21590@se2000.sebank.se>
Message-ID: <Pine.LNX.3.96.970915065929.18432A-100000@ns.viet.net>
Subject: Re: Browser Stats

On Mon, 15 Sep 1997, Chapman, Hass wrote:

> >From: Mike Meyer
> >However, I seriously doubt your numbers. While I'm pretty sure that
> >NSN/MSIE between them have 90% of the market, I'd be surprised if the
> >number of downrev users is as low as you claim. This is based on the
> >last time I went and checked the numbers for NSN, which has been a
> >while. What are you basing your claim on?
> 
> 
> Browser stats can be found at:
> 
> http://www.tamos.com/bin/browser.cgi
> http://www.ericsson.se/stats/BrowserStats.html
> http://browserwatch.internet.com/stats/stats.html
> http://www.cen.uiuc.edu/%7Eejk/bryl.html
> http://205.238.136.65/cgi-bin/wholog.cgi

You can also find them at
http://www.xmission.com/~snowhare/statistics/browsers.html. The apparent
low N of only 50K or so is the result of throwing out all non-text type
references from a much larger dataset to remove the
graphic/multimedia/active content biases. The data represents ONLY
references to text-like things on the server during the previous 24 hours.
I believe this to provide a fairly representive snapshot of current
browser usage.

Note 1: www.tamos.com's numbers represent a LONG TERM average over the
        last 8 months for a fairly low N of only 30K hits total. Since
        browser percentages can and *do* change significantly on much
        shorter time scales than that, their numbers really do not
        represent current conditions very well.

Note 2: BrowserWatch's stats are skewed by 'observer effects'. Basically,
        people primarly go there *because* they want to know browser
        percentages - and so they throw the numbers off with
        self-selection effects for the low percentage browsers. Hence
        their completely unrealistic percentages for CyberDog. The
        operators there do not seem to understand the problem this
        presents even after explicit explanation.

Note 3: The stats at www.cen.uiuc.edu are over a year and half old (as
        you might guess from the fact that it assigns Mosaic a 18%
        marketshare....) 

Note 4: The stats at 205.238.136.65 suffer from *extremely small N*
        Like only 68 hits in the last WEEK (!!!!).

If you want to generate your own stats, I have a script available at
http://www.nihongo.org/snowhare/utilities/ that is quite fast
and allows the elimination of the most common sources of browser 
stats skew. This is the same script as is used by www.ericsson.se
and www.xmission.com.

My breakdown for things with more than 1% share for the last 24 hours is:

   8124 15.4% Mozilla 3.01
   6259 11.8% Mozilla 3.0
   6236 11.8% MSIE 3.02
   5653 10.7% MSIE 3.0
   4431 8.41% MSIE 3.01
   3345 6.35% Mozilla 4.01
   2118 4.02% Unknown
   1987 3.77% Mozilla 4.02
   1442 2.73% MSIE 4.0
   1418 2.69% Mozilla 2.02
   1185 2.25% AOL-IWENG 3.0
   1169 2.22% Mozilla 4.03
    960 1.82% Mozilla 2.01
    945 1.79% Vnet
    695 1.32% Mozilla 3.03
    636 1.20% InfoSeek Sidewinder 0.9
    590 1.12% Mozilla 4.0
    536 1.01% Mozilla 2.0
    530 1.00% AltaVista Top

-- 
Benjamin Franz