ioi speaks out about ergh "VGC analysts"

[DonFerrari]

Zkuq said:

DonFerrari said:

Zkuq said:

Cumulative numbers will be quite a bit off if you round the numbers at every turn.

Not if you make a balance in the adjustments... some up, some down, and you can always adjust when you have more data, like they do now... but in math, its wrong to put numbers of significance above your margin of error...

Lets say you have a scale (analog) that measure up to 1 pound differences... So you would measure someone 150Lbs or 151... would be acceptable to use 150.5 +- 0.5 as well... but would be totally wrong to use 150.56738 Kg because you have no way to accertain that imprecision... but I don't know much about statistics methods and if it is acceptable to use this many significance numbers.

First, I must admit that I don't know a whole lot about statistics. That said, rounding the numbers to an arbitrary direction every time doesn't sound like a sound method, nor does it sound like a convenient one. Who could ever tell which direction to round the numbers each time? A good rule of thumb, and the only rule I know I can trust, is to round the final results only and never round any intermediate results. I suppose you could round intermediate results in some cases but it requires special care so that no accuracy is lost for the final result.

Not arbitrary... what i'm saying is... instead of using 238,738 consoles sold W1 use 240k (added 1,262) so next week lets say it sell 132,000 so use 130k (took 2k) (inbalance of 738), and so on... that is because they are not pulling that exact number on statistic they are guessing a range lets say from 200k to 250k and by their method 238,738 is the most representative number for them so they use it...

But if they are 5 or 10% uncertain of it they shouldn't use precision to the last number because the imprecision is of 11,936 to 23,873 so using 240 or even 250k would be better description of the numbers... hope that helped. But in the end that doesn't make much difference.