DonFerrari on 29 December 2013
Zod95 said:
DonFerrari said:
Zod95 said:
ioi said:
If we put out a figure saying that a game sold 600k units in a given period then we are not saying that the game sold exactly 600,000 units in that period. It is an estimate, based on much smaller data and extrapolated up. Think of it as a probability - when we publish a figure of 600k we are saying that we are 95% certain it has sold somewhere between 300k and 900k, 85% certain of it being between 400k and 800k and 70% between 500k and 700k and so on. |
So, if I want to work with a 95% certainty, everytime I look at recent figures I should rather see them as half, same or 50% more of what they are. For example, in the last week (ending 14th December 2013), the best selling game was COD Ghosts for X360 with 551k. In 5th there is Just Dance 2014 for Wii with 434k. But, in reality, COD Ghosts may very well be at 276k while Just Dance at 651k (over 2 times more than COD). COD Ghosts would be 10th rather than 1st and Just Dance would be 1st rather than 5th.
And if we apply this reasoning to every game in that chart (either adding 50% more or cutting by half the number we see there) we realize there's no point of seeing it as a sales ranking. I know, I know, "those are just estimates and they aim to be at the most accurate expected values". But my point is: the range for error is just too large to present the numbers as that chart does. I'm sure NPD and other tracking services don't work with "X - 3X" ranges, so it's ok for them to do rankings. Not the case of VG Chartz, so when someone points out that this site may be fooling people (which could be implicitly rather than explicitly), try to understand why they are saying such thing instead of calling them stupid.
In regards to the way you get the numbers, you seem to work ok and you definitely deliver a service no other site does (as far as I know). I think the problem is the way your site communicates, the way the numbers are presented are not coherent with the assumptions and methodology used. The example I just showed is proof of that.
|
What you didn't understood is that the numbers posted is like he is 95% confident that the number is within an interval with 5% of margin of error... so there would be a possibility of it being way off like you posted but that number or section would be less than 0.5% of probability and the combination of those 2 errors by that magnitude could be 1 in 1M. So you have to reread to understand what he said...
And in no place the admin staff says any different than ioi have posted here, and he have been doing it yearly, so maybe you just didn't see enough of the site to think that those were hard figures and that they have accounted all sales done individually.
|
Basically you're telling me that ioi is wrong when saying "when we publish a figure of 600k we are saying that we are 95% certain it has sold somewhere between 300k and 900k", since a 5% margin (as you claim they use) of 600k doesn't go to 900k or 300k. Is that right?
|
Partialy... they have 95% of certainty of this interval, not that 300K or 900K are significant in the analysis... like its stated in the same post it would decrease certainty the narrower you get the band... so 400k-800k would be less and 550k-650k even less... but even so to have both numbers as you said one lets say at 300k and other in 450k the probability of they being inverted is quite low.
