I do indeed: browser statistics are fundamentally flawed. That said, they're reasonable for spotting trends.
To compare the two:
- NetMarketShare analyzes 40,000 websites. It makes statistical adjustments to weight for users and other factors. They have corporate sponsors - although I'm sure that wouldn't alter their findings...
- StatCounter analyzes 3 million websites. It makes no adjustments and they are independent.
In simple terms, NMS treats every user equally. If your Granny uses IE6 to visit one site, she's treated the same as you using Chrome all day to visit two thousand sites. It's one vote for IE6, one vote for Chrome.
StatCounter looks at visits. Using the same example, Chrome would have 2,000 votes, IE6 would have 1.
So they're different. Personally, I think StatCounter is the simplest and most realistic example. It doesn't attempt to weight users but that doesn't matter - the large sample size means averages and anonomolies are reasonable.