Again about browser statistics...

An old article that is still up to date

[starway]

I found an excellent article today: Debunking Browser Stats [pantos.org] and wanted to share this between forum members (hope its URL won't be deleted).

Some members probably read it in past, but I'm sure lots of other didn't know about it. It is rather old (mid 1997) but I was surprised that ideas represented in it didn't change now. Same questions, same problems - and excellent author's comments...

Highly recommended.

[Brett_Tabke]

That's a great synopsis of the current situation.

The biggest thing to remember about browser counting is that most of the newer 3rd party browsers must "lie" about their agent to gain compatibility with some sites. That skews all stats by atleast (guess) 10%.

[c3oc3o]

I don't think I know any browser that lies by default to an extent that it can't be identified.
Opera, K-Meleon and surely some others can be configured to do so, that's true. Also, Opera by default claims it's MSIE (which in turn claims to be a Mozilla), but does identify itself as Opera later in the UserAgent string: "Mozilla/4.0 (compatible; MSIE 5.0; Windows 98) Opera 5.01 [en]"
With careful analysis, I'd say you can identify at least 95% of browsers.
Of course bots claiming to be browsers may skew the stats too, but you can circumvent this by detecting the user agent through JS or images, which many robots don't request.