WW HW + SW week ending 31st Jan

[binary solo]

Jabbamk1 said:

binary solo said:

Actually it does. If the release date isn't updated in the database - which is the actual error here - and the data sample does not have that game among confirmed sales the algorithm will still estimate sales that the data base tells the algorithm it is on the market. This does nothing to prove the sales estimate system is broken or unreliable. But it does seriously call into question the reliability of the game database management.

If you know how statistical modelling works you'd have understood these things, so whatever you do with the market research business it seems statistical modelling isn't it. Is it bad that vgc has recorded sales for a game that isn't out yet? Yes it's bad. But not for the reasons you are suggesting.

The point I'm making is that it's based off no real world data whatsoever (referring to actual DW8:Empires sales). 

What were to happen if the game database did update? Would we have the same number but just it comes out next month? It throws VGChartz entire premise out the window. You can't "track" a product that hasn' been released but it seems VGChartz already has algorithms in place to create numbers out of thin air. 

Anyone claiming VGChartz is a reliable tracker is wrong. It's a bunch of childs guesses. 

You're really showing a lack of understanding here. If vgc gets actual game sales data from 20 retail locations in the country and none of those locations record a single sale for, let's say, sunset Overdrive does that mean vgc should record SO sales for that week as zero? Of course not, because there are thousands of retail stores in the country and some of them will have sold SO. The algorithm uses various statistical techniques to estimate game sales for thousands of games for which no direct sales data is available. That's not child's guesses, it's actually sophisticated statistical processing. The basis of your complaint is groundless because you're aiming your critique at the wrong flaw in the system. The flaw in the system is that the game database scheduled the release date for this game wrongly and thus the algorithm picked up that game as requiring a sales estimate. The algorithm relies on the database having a correct launch date to trigger when sales estimates should start. If you get the input information wrong then the information coming out is bad. The QA on base game metadata is at fault here, not the algorithm used to create sales estimates. 

The fact is no one is under any illusion that game sales numbers are at all accurate. There are too many games and not a big enough sampling plan to yield highly reliable figures. But getting the launch date of a game wrong and thus creating an obvoliously incorrect sales number for it proves nothing about the accuracy of the system to estimate sales.