kowenicki said:
Kasz216 said:
BigBoobieHead said:
Kasz216 said:
fxa5209 said:
Cant believe that this web site would question its integrity with such a huge adjustment. This is not even within any kind of error margin. Serioulsy, i am starting to doubt any number here. |
1.1 million divided by 31.21 gives me an error of 3.5%.
1.1 million divided by 13.89 gives an error of 8%.
Both are within VGC's margin of error of 10%.
It's an adjustment going back... pretty much forever as they refine their process. So yeah... it is within the margin of error.
|
That's spin and you know it. If your going to be fussy like that you should know that probably the majority of software on this site is out of the margin or error. Also, if a console sells 100M, do you think a 10M error is acceptable just because it's in the 'Margin of Error?'
|
How is that "spin". It's percentages... using the number based over the timespan given.
Also.. if a console sells 100M and their is a 10M error... is that acceptable? Only if you have a hard count of every system sold. Since that would be the entire swing of your margin of error.
Bit if you knew in fact that 110 or 90 million consoles were sold then... yeah. That's acceptable.
That's what a margin of error is... a margin... a percentage. Just because a number gets higher doesn't mean that the acceptability of the margin of error gets lower.
Being 1 off on a sale of 10 is no different then being 10 Million off a sale of 100 Million.
Hell... i'd say being 1 off a sale of 10 is WORSE... I mean your only tracking 10 things... vs you know... 100 Million.
|
You are being wayyyyyyy too simplistic.
A margin of error of 10% should NEVER mean something is overtracked or undertracked by 10% over a very long period of time, that would be a very weird fluke if that happened. A margin of error will sometimes mean overtracking and sometimes undertracking.
So if we accept as a fact that we have a weekly or monthly error of 10% then that should lead to lifetime figures much closeer than that 10% error.
|
It wouldn't be a weird fluke.
It would be a systematic error in the formula.
Unlike the sampling bias type of error your talking about.
Which is exactly what Ioi admited to in this thread.
Their others formula had a systematic error in it causing them to overstimate others hardware sales on a number of platforms.
Systematic errors are always going to cause huge changes like this....
and every company has them. Even NPD I'm sure has had their fair share of systematic errors. It's just you don't hear about those because you don't pay for the private data.
Every tracking company further refines their formulas... and when they do so it leads to large changes in numbers.
