orignal computation is trying to take into consideration the increaseing gap between the ps3 and x360 sales.  "if" trending would continue and sales gap would continue to increase, chances are that it wont take 3 years for the ps3 to catchup to the x360. (emphasis on if)

if we take a look at sales, ps3 and xbox 360 sold almost the same during the first year of their life.

if sales, continued this way, then there is no way in hell that the ps3 is catching up to the x360, but, this isnt the case. now, lets take a look at the 2nd year sales. and we start seeing a different picture

the x360 has sold for a full year and only managed 7.8m while the ps3 on its second year has already sold 6.2m in just 5 months.  now, if we compare same period for both the x360 and the ps3 we see the ps3 outselling the x360 by 0.9m in 24 weeks.

so what does this all mean? nothing really, to lazy to actually do my own computations. just saying that "if" trending continues and the ps3 continues to outsell the x360 and is increaseing the sales gap then it is incorrect to say that it will take the ps3 123 weeks to catch up to the 360s install base. (again, emphasis on if)

if you want to do an accurate prediction, we should consider seasonality, game realses, price drops, new features, new modes, etc etc etc. 

Trending like that doesn't scale linearly and fails for this kind of computation. Using the basic method of sales difference divided by difference in sales rate gives you the most accurate method of computation without accounting for seasonality, price drops and software releases.

it never does. using the basic method is the safest though not necessarily the most accurate method. if it were the most accurate, vgchartz would be using it and not bothering with complex computations to forecast its sales figures.  the same also holds tro for npd and every other firms and companies specializing in data collection and predictions.

I said most accurate when not accounting for variations such as seasons, price cuts and software, etc... When you take those into account, a much more sophisticated formula certainly comes into play.

The problem with Masterb8tr's method is that it continually narrows the sales gap disproportionally to the actual sales.

For example: If PS3 outsold X360 by 20k units this week. Chances are that it will outsell X360 by 20k units next week. Masterb8tr's method would have the PS3 outsell the X360 by 21k the next week, then 22k the following week, etc....


A better method, yet still simplified, would be to take an average weekly sales difference for the past several weeks or even months so that shipment fluctuations and software releases trends can be factored into the equation.

In fact, I'll work on that tomorrow. I bet you we still come up closer to 123-125 weeks than 66.

so you did, so you did, must have missed that, was in a hurry to eat luch. then i appologize, when not considering any other constraints, it is the most accurate.

now here's where it gets tricky, it is probable that if the ps3 outsells the 360 by 20k this week that it will again outsell it by 20k next week, but it is also probable that it could outsell it by 21k or 23k or higher and again, it is also probable that it wont outsell the 360 at all.

i'd also work on one too but i work in the merchandising department of a hypermart chain and its my job to make sure that the categories/items that i handle are always on stock at all our stores (7 hypermarts and 22 supermarkets).  to be honest, forecasting demand was fun especially when your right at first, but after a while (5 years) it gets old...really old.