Well, my guess would be that some people actually do assume it is correct :)

I agree that it is a good website in that regard, and there could be some good discussions on the forums. I would personaly change some things around to make it more clear what it is about and change some other stuff too :p

But I'm not gonna try to tell anyone how they should run their web-page.

I'd definitely agree with this. Although I'd change one minor thing: you say that we shouldn't assume that VGCharts' numbers are "correct" -- I'd change that to "exact," or "almost exact," as NPD's numbers aren't correct either, they're just more exact.

Well, I just question the value of providing weekly numbers when those numbers are difficult or impossible to get even for professional marketing firms like NPD. And even NPD's monthly numbers are called into question at times, and they have, what, 65% coverage? With additional polling to further establish purchase patterns? I would much rather see VGChartz provide officially released numbers from recognized sales agencies rather than speculating independently. A 10% margin of error is pretty huge when you're talking about hundreds of thousands of units, and I know for a fact that on some months VGChartz' average margin of error is even higher.

Bump? 

What?

Listen buddy, bump a thread if you've got something to add...I figured your question has been well answered by now anyway.

Not really. No one's really said anything about how VGChartz assembles its weekly North American sales figures, except to say that, "They have sources," which doesn't really help anything.

And then there's also my last post, where I question the wisdom of providing weekly sales figures at all. I wanted to bring attention back to that point, hence the bump.

I'll speak to this.

I asked ioi directly the other day how much sales coverage he has in the Americas - it is increasing constantly - but it is no where near NPD coverage.  Still, the data is representative, and statistically sound because when dealing with data spread out from tens of thousands of stores selling videogames and videogame hardware across the continent, even 1% representation would equal over 1,000 stores - more than enough to have estimates for sales trends.  Now, whether or not ioi is getting partial samples or full data from the retail channels he does cover is something we don't know.  Statistically, the greater the sample size, the more nuance you see in trends and change. 

It is also possible to elimiiate gaps - for instance I would guess ioi can estimate PS2 numbers now because he has added stores that sell alot of that console still, while before the stores covered did not sell many/any PS2's, as it varies across the USA.

I don't think I'm allowed to reveal the numerical estimate of retail coverage ioi thinks he has...but it has roughly doubled since he started providing the American charts, which is why the margin of error for everything was quite accurate when compared to NPD last month.  He has passed what I consider the necessary threshold for accurate figures in the past few weeks, and in the coming weeks should approach a figure that allows us to see more of the aforementioned nuances that histrorically only professional agencies could provide.

Thank you!

That's more what I was looking for. Obviously I would've liked to know more specific figures, but I can understand why that might be kept confidential. In any case, I'm curious now as to how much VGChartz' deviation from NPD numbers has changed over time.

TheSource has posted a few news articles showing the difference in percentages between NPD and VGCharts. The first NPD after American charts went weekly was off by a large margin, but the most recent NPD saw the weekly charts much closer to what was reported. I imagine, as ioi collects more and more data from more sources, you will see more accurate representations. I'm certain TheSource will provide another breakdown to show the differences between VGC and NPD for May when the time comes.

You say you'd much rather VGC provide "official" numbers from "recognized" sales agences rather than speculating independently.   So you would rather VGC be "dependent" on other organization's data?  What, exactly, is the value in that?

Do you realize the irony of that statement?  NPD numbers are "official" in only one way: They are the official NPD numbers.  Beyond that, they are a statistical estimation of sales.  Period.  They are "recognized" in that they report their "official" numbers and are pretty much the only game in town.  NPDs numbers are "independent" in that they gather data, statisically analyze it, and report it but are completely dependent on the data they gather from retail and distribution channels.

VGC's sampling data, estimating, statisically projecting, etc. is no different than NPD and it doesn't make their numbers any less independent or dependent as well as any more or less official.  VGC most certainly is estimating sales (just like NPD is an estimate) on a smaller budget and a smaller data set but that doesn't make VGC's numbers any less "offical"...

You said you have a higher degree in statistics.  While my degree isn't in statistics, I have taken quite a number in my studies in support of my chosen field.  There is no doubt that a larger sample will reduce deviation, but it is entirely possible to create fairly high quality estimates with small sample sizes.  If the entire population was acting independently, sample size would have to be huge.  Since, in every entry in the sales charts for both hardware and software, each and every entry is a "single" sku, the need for sampling is to account for overall variances between different stores and regions.  Given past history of how each area in the U.S. along with how close each store tracks compared to others, it would be quite possible to get a pretty darn high i.e. 90% confidence with a very small sample size.   Only ioi knows his statistical methods and I'm sure he won't be sharing them with us.  (I certainly wouldn't)

Give it six more months and see how things are going in terms of his predictions.  As with anything statistical, over time we'll know more accurately how well his sales statistics compare to other data.  So far I'm fairly impressed with his results given the resources and time involved to produce such data.