Oh noes, someone on the internets is wrong :P
VGChartz is a very accurate site, and Brett has every reason to be proud.
As others have said, VGChartz receives sales data from retailers, and extrapolates on this to determine the final value. Brett has not told us how large his sample size is, but we have been told that it is not as large as the sample size for NPD data, but that it has increased recently (ever since the data started coming in a couple of days later in the week)
This is the exact same methodology as NPD by the way.
Brett will constantly monitor various other sources looking for games/consoles that have been overtracked or undertracked. Such sources include the numbers of other sales trackers - NPD, MC, Famitsu, the Aussie one, the British one etc as well as manufacturer's shipment reports. This means that even if a single weeks figures are out by 10% (or more in some extreme cases), the total will almost always be within 5% after the game/console has been out for a while.
The sales for obscure areas are harder to get, yes. I believe Norway is one of the countries that actually does get tracked, but obviously, Brett does not get retail info from all 260 or so countries in the world, in those case, I am under the impression that the methodology is to assume the games and consoles have the same attach rates as in a similar country that is tracked. As an example, you could probably not track New Zealand, and just assume that its sales of everything are 20% of Australia's sales (excepting LOTR games, which could be higher) Anyone who actually knows, want to confirm on deny that last bit??