Support NeoGAF

[ioi]

In response to a number of different threads about "Well Sony say they shipped 5m PSPs in Europe so they must have sold close to this amount" etc, I've decided to attempt to demonstrate just how wildly inflated Sony shipped numbers are.

Lets concentrate on North America.

This is what Sony have shipped to NA:

Date		Shipment		LTD
Mar-01		3.01		3.01
Sep-01		5.23		8.24
Mar-02		2.35		10.59
Sep-02		7.35		17.94
Mar-03		4.27		22.21
Sep-03		5.21		27.42
Mar-04		2.11		29.53
Sep-04		0.70		30.23
Mar-05		5.12		35.35
Sep-05		3.43		38.78
Dec-05		2.21		40.99

These are NPD numbers for USA only:

Date		NPD Sell-thu	NPD LTD
Mar-01		2.13		2.13
Sep-01		1.96		4.09
Mar-02		4.32		8.40
Sep-02		2.77		11.17
Mar-03		5.82		17.00
Sep-03		1.90		18.90
Mar-04		4.05		22.95
Sep-04		1.64		24.58
Mar-05		3.59		28.17
Sep-05		1.74		29.92
Dec-05		2.29		32.20

Now, I'm going to make a few assumptions:

1) Canada represents around 6% of the US market (supported by total revenue per year, snippets of sales we have etc) and that the pattern of the Canadian market is the same as the US with same games released, similar price cuts etc.

2) NPD isn't exact. We assume for simplicity that the error is consistent from month to month, which is fairly reasonable over the timescale of a few years as monthly fluctuations will average out over the large sample time.

The result of this is a "scale factor" that can be applied to NPD data to represent actual NA sell-through, taking into account both Canada (scale factor on it's own of around 1.06) and inaccuracies in NPD.


Luckily, we have a very handy 'pinch-point' at Sep 04, when very small shipments had been made by Sony as they wanted to use up PS2 stock to make way for PSTwo. Only 700k had been shipped between Apr 04 and Sep 04 for example, whereas NPD said that 1.64m had been sold in US alone.

Now, if we take the total NA shipments as of Sep 04, we have 30.23m. NPD sell-through in the US upto Sep 04 is 24.58m.

If we divide the 30.23 by 24.58 we get 1.23. Say we round down to 1.2 to take into account the small amount of stock that may have remained, we can assume that of the 30.23m shipped in NA, we had (1.2 x 24.58) 29.5m sold-through (using our 6% Canada split we have 27.83m in the USA and 1.67m in Canada).

This all seems reasonable so far, NPD is under-reporting PS2 numbers a bit (24.6m vs 27.8m). Of course we have made a number of assumptions. The scale factor could be lower than 1.2 if more stock was on shelves, but it couldn't be any higher than 1.23 as they can not have shipped more than they sold. To keep the sceptics happy, we will use the 1.23 factor from now on.

Now let's scale all NPD up by 1.23 to represent NA sell-through and compare back to Sony shipments:

Date		Sony		Sony LTD		NPD x 1.23	NPD x 1.23 LTD	Difference
Mar-01		3.01		3.01		2.62		2.62		0.39
Sep-01		5.23		8.24		2.41		5.02		3.22
Mar-02		2.35		10.59		5.31		10.33		0.26
Sep-02		7.35		17.94		3.41		13.74		4.20
Mar-03		4.27		22.21		7.16		20.90		1.31
Sep-03		5.21		27.42		2.34		23.24		4.18
Mar-04		2.11		29.53		4.97		28.21		1.32
Sep-04		0.70		30.23		2.02		30.23		0.00
Mar-05		5.12		35.35		4.41		34.64		0.71
Sep-05		3.43		38.78		2.14		36.79		1.99
Dec-05		2.21		40.99		2.81		39.60		1.39

Look at the end column. This is the difference between the shipped numbers and our sell-through, which would represent whatever "stock" was on shelves. Using even this most extreme case, there are apparently 1.39m PS2s on shelves at the moment, 1.99m at the end of Sept, and 4.2m at the end of Sep 02.

Would 2.5m PSPs sitting in Europe as of the end of 2005 be that unrealistic?


My explanation is not that it's a conspiracy or that Sony are lying. Likely that they shipped a lot of PS2s early on and are doing the same with PSPs as they know they can sell them. You'll find that PSP shipments this quarter will likely be very low, as with some periods above, when Sony wait for stocks to go down before shipping more.


My final point is that I think the 1.23 factor we used was too high. Dividing by the 1.06 for Canada this suggests that NPD are tracking about 10% to low which seems high to me. It would also suggest that there was no stock at the end of Sep 04 (which the above table shows), which is unlikely. Shortages didn't hit until Nov / Dec. A more likely picture would be using a factor of 1.2 which gives even larger discrepancies as expected. I've done the same thing for GBA / XB etc and they also seem to give this figure of about 1.2 or a little below (1.18) so it seems reasonable to suggest that if you take NPD x 1.15 - 1.2 then you get pretty close to NA actual. Using 1.2 for PS2:

Date		Sony		Sony LTD		NPD x 1.2		NPD x 1.2 LTD	Difference
Mar-01		3.01		3.01		2.55		2.55		0.46
Sep-01		5.23		8.24		2.35		4.90		3.34
Mar-02		2.35		10.59		5.18		10.08		0.51
Sep-02		7.35		17.94		3.32		13.41		4.53
Mar-03		4.27		22.21		6.99		20.39		1.82
Sep-03		5.21		27.42		2.29		22.68		4.74
Mar-04		2.11		29.53		4.85		27.53		2.00
Sep-04		0.70		30.23		1.97		29.50		0.73
Mar-05		5.12		35.35		4.31		33.81		1.54
Sep-05		3.43		38.78		2.09		35.90		2.88
Dec-05		2.21		40.99		2.74		38.64		2.35

I know I've mad a few key assumptions here, but I think they are all pretty reasonable given the timescales and so forth that we are looking at here.


Discuss

 

[Jonnyram]

Why?

 

[jarrod]

Why?

Some people can't accept that the difference in Nintendo and Sony's shipping/accounting reports have basis in anything other than PSP outselling DS? Despite the fact PSP has yet to really do that in any major region according to any major polling outlet or investment agency?

 

[monkeymagic]

A+++ for effort

But way too much time and brain cells wasted on damage control :lol

 

[ioi]

Why?

Because using some fairly rigorous maths to prove a point is better than throwing your hands in the air and claiming to know what you're taking about.

Having 3-4m units in "stock" isn't that crazy a concept going by past data that we have. In fact, it's very much a characteristic that Sony demonstrate again and again, so it will hopefully put an end to people suggesting that PSP has sold more than it has just because Sony have shipped loads of units.

 

[ioi]

Doing the same thing for GBA on the other hand, shows how Nintendo shipments track very closely to sell-through, usually with 500k or so stock at any one time.

Date		Nintendo		Nintendo		NPD x 1.2		NPD x 1.2 LTD		Difference
Sep-01		3.44		3.44		2.26		2.26		1.18
Mar-02		4.12		7.56		4.79		7.06		0.51
Sep-02		3.06		10.62		2.43		9.48		1.14
Mar-03		4.74		15.36		5.85		15.34		0.03
Sep-03		4.28		19.64		3.24		18.58		1.06
Mar-04		5.17		24.81		5.98		24.56		0.25
Sep-04		3.80		28.61		2.78		27.34		1.28
Mar-05		4.76		33.37		5.18		32.51		0.86
Sep-05		1.68		35.05		1.55		34.07		0.99
Dec-05		2.50		37.55		2.69		36.75		0.80

 

[AdmiralViscen]

A+++ for effort

But way too much time and brain cells wasted on damage control :lol

What damage control?

 

[Sho Nuff]

In the amount of time it took you to do that you could have become a master of karate

 

[Ponn]

Do you get paid for this?

 

[Dilbert]

Because using some fairly rigorous maths to prove a point is better than throwing your hands in the air and claiming to know what you're taking about.

Out of respect for all the people who really understand and use math, could you please stop using the word "rigorous" to describe what you posted? You can't call your analysis "rigorous" when it's full of unjustified assumptions and rounding "for convenience." Mathematical rigor is a very specific concept.

 

[Fishie flop oog]

The shipped numbers for Sony include units used in demo pods, giveaways, warrantee replacements etcetera.
Nintendo doesnt count those.

 

[Pimpwerx]

Not everyone reports to NPD. OTOH, Sony reports this figures to shareholders. Let's take hald of the outstanding 8M units. Each has a mean value of $225 (rough value of [$300+$150]/2). 4M * $225 = $900M in outstanding revenue for retailers alone. That's going on half the discrepancy between Sony's shipped, and NPD's sold values. Let me make it clear NPD is not an absolute measure of sales. There was once a time when it was understood on this forum that trackers were flawed. Why this is consistently overlooked is beyond me, but it flies in the face of simple logic.

There is not capacity in the distribution chain to house all those PS2s. Even counting Canada, you can't bridge the gap. I don't think NPD's true value is even in hardware sales, which forumites love to quote. I think it exists to track relative performance of software, which comprised the lion's share of the numbers they put out. If Sony is overstating numbers, then they are lying to shareholders, period. If their shipped figures are true, and NPD was an accurate guage of sales, then retailers are currently bolstering Sony's fiscal report by over a billion dollars for no apparent reason. There exists a discrepancy for a reason. The problem lies in the stubborn reliance of regular joes using incomplete data to draw firm conclusions on hardware sales. Software sales aren't even complete, but they're bound to offer a good gauge of relative performance. You don't need the exact number of sales to tell if a game is selling well, or to know what genres are hot or not. PEACE.

 

[ioi]

Out of respect for all the people who really understand and use math, could you please stop using the word "rigorous" to describe what you posted? You can't call your analysis "rigorous" when it's full of unjustified assumptions and rounding "for convenience." Mathematical rigor is a very specific concept.

The maths is rigorous.

The fact that it is based on assumptions is the same as with all applications of maths, it's how well you can justify the assumptions and judge the effect they have on the results that counts.

 

[ioi]

Not everyone reports to NPD. OTOH, Sony reports this figures to shareholders. Let's take hald of the outstanding 8M units. Each has a mean value of $225 (rough value of [$300+$150]/2). 4M * $225 = $900M in outstanding revenue for retailers alone. That's going on half the discrepancy between Sony's shipped, and NPD's sold values. Let me make it clear NPD is not an absolute measure of sales. There was once a time when it was understood on this forum that trackers were flawed. Why this is consistently overlooked is beyond me, but it flies in the face of simple logic.

There is not capacity in the distribution chain to house all those PS2s. Even counting Canada, you can't bridge the gap. I don't think NPD's true value is even in hardware sales, which forumites love to quote. I think it exists to track relative performance of software, which comprised the lion's share of the numbers they put out. If Sony is overstating numbers, then they are lying to shareholders, period. If their shipped figures are true, and NPD was an accurate guage of sales, then retailers are currently bolstering Sony's fiscal report by over a billion dollars for no apparent reason. There exists a discrepancy for a reason. The problem lies in the stubborn reliance of regular joes using incomplete data to draw firm conclusions on hardware sales. Software sales aren't even complete, but they're bound to offer a good gauge of relative performance. You don't need the exact number of sales to tell if a game is selling well, or to know what genres are hot or not. PEACE.

I'm not quite sure what your point is there. What was in fact done above, is to use the "trend" given to us by NPD and fit it to the data we have for shipments to take into account of these discrepancies that you have mentioned.

 

[fronn]

Are you actually trying to prove something with fudgy numbers?

Surely you cannot see the irony in this? Using fudgy numbers to prove something about percieved fudgy numbers? Silly man.

Leave the fudgy math to accountants.

 

[ioi]

Are you actually trying to prove something with fudgy numbers?

Surely you cannot see the irony in this? Using fudgy numbers to prove something about percieved fudgy numbers? Silly man.

Leave the fudgy math to accountants.

Did you not understand the original post?

 

[fronn]

Did you not understand the original post?

Yes, I understood. But you cannot just go and base your entire point on something that was entirely controlled by your rounding of .03 to account for stock. It has no roots in reality -- it may well be correct, but that is hardly a way to prove a point.

It's called fudgy math.

 

[Dilbert]

The maths is rigorous.

The fact that it is based on assumptions is the same as with all applications of maths, it's how well you can justify the assumptions and judge the effect they have on the results that counts.

The fact that you can make these two statements back-to-back in the same post proves my point.

 

[ioi]

The fact that you can make these two statements back-to-back in the same post proves my point.

What?

All applications of maths are based on assumptions. Everything. Engineering, computer science, physics, biology, economics, everything.

It's what assumptions you make and how you justify them that matters.

You cannot say the maths behind it is not rigorous because it is. You multiply a number by another and it equals another number. It is a fact and you cannot dispute it. What you can dispute are the assumptions which is fine, but that isn't what you said. You said the maths is not rigorous which is wrong.

I made my assumptions at the start and then went on to justify them and why making them has no effect on the overall results, just as you do when approaching any mathematical model whether it is the modelling of wind loading on a bridge or breeding rate of bacteria or interest rates.

What exactly is it about the assumptions that you don't agree with?

Yes, I understood. But you cannot just go and base your entire point on something that was entirely controlled by your rounding of .03 to account for stock. It has no roots in reality -- it may well be correct, but that is hardly a way to prove a point.

I don't think you did understand, because nothing is controlled by rounding of .03?

 

[galoisky]

If that mathematics had been rigorous, it would have contained the word proof about 40 times.

 

[Flo_Evans]

hey ioi, you wanna do my 2005 taxes? I could use some of that fancy rounding!

The NPD numbers are projections based on a sampling of 25% (or possibly less, I am not sure exactly how much they sample). The Sony shipped number includes Canada, which the NPD doesn't track at all. AFAIK Sony doesn't give a number for south america separately either, so thier 'NA' shippments might include them aswell.

 

[Yoshi]

Although he may not be 100% accurate, I think, what he says is valid, because he uses a point of time to compare shipped vs npd-sold, where actual sell through was nearly the same as shipment. As this date was many months later than PS2's launch, his numbers seem to be a good approximation. His posting isn't about telling us, those numbers are 100% correct, but he shows, what shipment VS actual sell through circa is like

 

[jett]

hey ioi, you wanna do my 2005 taxes? I could use some of that fancy rounding!

The NPD numbers are projections based on a sampling of 25% (or possibly less, I am not sure exactly how much they sample). The Sony shipped number includes Canada, which the NPD doesn't track at all. AFAIK Sony doesn't give a number for south america separately either, so thier 'NA' shippments might include them aswell.

Just fyi SCEA doesn't have a distributor or representant in south america(unless you meant Mexico for some reason ). Everyone buys their PS2s from import stores. Sony just doesn't give a shit.

 

[ioi]

Although he may not be 100% accurate, I think, what he says is valid, because he uses a point of time to compare shipped vs npd-sold, where actual sell through was nearly the same as shipment. As this date was many months later than PS2's launch, his numbers seem to be a good approximation. His posting isn't about telling us, those numbers are 100% correct, but he shows, what shipment VS actual sell through circa is like

My god, an intelligent poster!!!

What I could have done is just compared Sony shipped to NPD raw numbers and said "look, Sony have shipped 40m but NPD say they have only sold 32m as of Dec 05" !

What I have in fact done is taken a worst-case situation for my argument (best case situation for PS2 / everyone who disagrees with my point).

I've made two key assumptions:

1) That when a system is sold out, shipped = sold. REASONABLE

2) That discrepancies in LTD numbers are likely to be consistent from month to month. So if LTD is 20% higher then it's likely that each month is about 20% higher (give or take some random variations)- so month 1 may be 25% higher, month 2 15%, month 3 26%, month 4 16% and so forth, with the average over 2-3 years being 20%. It's highly unlikely that one month numbers will be 80% higher, then 40% lower the next and so forth, this is an unreasonable fluctuation. NPD isn't fact but it's accurate to +/- 10%. REASONABLE


So I have found the HIGHEST POSSIBLE factor to increase NPD by without having a case where more has been sold than shipped at any one point, and the point at which they become equal fits in well with the period that we know PS2s became quite scarce. That's the HIGHEST factor possible, presenting the PS2 in the best light possible. And yet people are still not happy !!

Just look at the data. Sony ship huge amounts in Mar-Sep to build up for the xmas rush, then ship less in Sep-Mar as the stock is used up. You can see it 2-3 years in a row. Whether the factor is 1.23 or 1.2 or 1.15 or anything else it makes little difference, you would still have 4-5m PS2s that have been shipped and not sold at some points.

Nobody is saying that the numbers I have shown are exactly what the PS2 has sold. It's merely a trend, a comparison using a bit of logic to account for NPD discrepancies and the lack of inclusion of Canada in data, that was done in a way to present the PS2 in the best light possible.

 

[Frankfurter]

The NPD numbers are projections based on a sampling of 25% (or possibly less, I am not sure exactly how much they sample).

NPD covers a bit more than 60% of the market.

Excellent work @ ioi

 

[SolidSnakex]

And we went through all this because you're upset about people saying the PSP is close the DS sales in certain territories?

 

[Ghengis Cohen]

And to think after that that work you went through people are still going to play videogames on Sony systems.

 

[ioi]

And we went through all this because you're upset about people saying the PSP is close the DS sales in certain territories?

Not at all. I like maths and find it interesting- it was a half an hour well spent this afternoon between lectures, and was actually more fun than playing any games I've bought in the last couple of weeks so don't worry about me

 

[Pimpwerx]

My computer is wigging out. DP. I'll post a reply later. PEACE.

 

[Cpl_Cain]

I don't understand all of the numbers mumbo jumbo. What exactly are we saying here?

 

[Trident]

The maths is rigorous.

The fact that it is based on assumptions is the same as with all applications of maths, it's how well you can justify the assumptions and judge the effect they have on the results that counts.

You don't want to argue with jinx about math.

 

[Vennt]

And to think after that that work you went through people are still going to play videogames on Sony systems.

:lol

 

[Ignatz Mouse]

There are too many assumtions with too much of a potential swing for these numbers to mean anythign to me. Especially if the end result is to suggest that there are 2.5 million units on shelves in Europe. That seems pretty far-fetched to me, and I'm more willing to believe one of the assumtions about ratios or estimations is wrong.

 

[xaosslug]

So in short, what you're charting is that a company that reports on Sony's ship/sold numbers must know more than Sony themselves. Oootay... Good luck with that. lol

 

[RuGalz]

A+ for the effort but the logic flawed simply because you can't just use the number comparison for one system and apply to another just because it's from the same company. The production rate is different and the demand in the market is different between systems.

 

[Pimpwerx]

Like I said in the other thread, you are extrapolating figures using already extrapolated numbers. Your percent error is compounding. Comparing sell-through to shipped is wacked for the fact that NPD is extrapolated. Thus, as your sample size increases, the disputable range of possible outcomes scale proportionally.

The math is not rigorous. This is business math at its most basic, and that's basic arithmetic. I don't think anyone is having problems following the numbers, it's the justification that's the problem. When one set of numbers is already flawed/questionable/hokey, then the second set of numbers, derived from the first are also gonna be flawed/questionable/hokey. It's a closed group in that regard.

The thing that sets off the alarm bells everytime is that gap between the numbers. On the PSP/DS scale, it might not be as monumental from a hard sales pov (percentage wise, it's a train wreck). But when you scale to the PS1/PS2 levels, now you're looking at numbers that just don't fucking work in any reality you care to live in. And even the percentage of error (20% difference) is massive. Are retailers subsidizing SCEA? Are they just ordering systems blindly with no regard for demand? OR...would the more logical conclusion be that numbers we already know are extrapolated and thus flawed are not a reliable gauge of performance. Shipped aren't either, but at least there is some accountability to the shipping figures provided. AFAIK, trading laws require Sony to report accurate numbers, unless they're misleading investors. The truth lies in-between somewhere, and no amount of guesstimation will generate acceptable results. PEACE.

 

[Dilbert]

All applications of maths are based on assumptions. Everything. Engineering, computer science, physics, biology, economics, everything.

It's what assumptions you make and how you justify them that matters.

You cannot say the maths behind it is not rigorous because it is. You multiply a number by another and it equals another number. It is a fact and you cannot dispute it. What you can dispute are the assumptions which is fine, but that isn't what you said. You said the maths is not rigorous which is wrong.

I made my assumptions at the start and then went on to justify them and why making them has no effect on the overall results, just as you do when approaching any mathematical model whether it is the modelling of wind loading on a bridge or breeding rate of bacteria or interest rates.

What exactly is it about the assumptions that you don't agree with?

Rigor is about more than just being able to multiply two numbers together. I know this might blow your mind, but math is LOGIC. The bottom line is that your entire modeling process was not logical, and therefore lacks rigor.

I know a few things about modeling situations using math -- I was a physics major, after all. Since you asked, here is my list of complaints with your "proof."

1) First, you want this to be seen as a case of "my equations are rigorous but the numbers have assumptions," but that's not true. You have made a couple of very basic assumptions that the ratio of sales in the United States and Canada is constant, and that the month-to-month error in the NPD sales is constant. These are non-trivial assumptions which affect the rest of your calculations -- what if the relationships are periodic? linear? higher-order polynomial? uncorrelated? -- and I see absolutely NO justification for why you think modeling them as constants makes sense.

2) You round off in the middle of your calculations, which is a no-no. Even high school lab students know that you should report all of your data to all of its significant figures for calculation accuracy, and then round off to the lowest number of significant figures to report the final result.

3) Also from lab -- high school or early college, depending on which track you were in -- is the notion of error estimates. Each value has an uncertainty, which needs to be propagated through the calculation to come up with the uncertainty in the final result. When you claim that you might have 2.5M units on a shelf in Europe, it begs the question: 2.5M...plus or minus how many? Are you certain within a hundred thousand, or within a couple of million?

4) Of course, the problem is that you have no way of estimating what the likely error is in ANY of your numbers. For example, you make assumptions that the error in the NPD numbers is roughly constant over time and that 2-3 years is a sufficiently long period to come up with the mean, but provide no justification for either statement. Unless you have some way of comparing several NPD reports to ACTUAL, EMPIRICALLY VERIFIED values, I don't see how you know WHAT the margin of error is.

5) Finally, you use the phrase "simplify for simplicity" WAY too much. Using that phrase means that a) you understand the detailed phenomenology that you are modeling and b) can justify that you are ONLY dropping higher-order effects with your simplification and that your results will be within x% of the full calculation. I'm sorry, but claiming that you're doing things "for simplicity" is simply bogus. (No pun intended, of course.)

Honestly, I think the only thing you can show with the raw data is that shipped > sold...but that falls squarely into the realm of "no shit, Sherlock."

 

[ioi]

There are too many assumtions with too much of a potential swing for these numbers to mean anythign to me. Especially if the end result is to suggest that there are 2.5 million units on shelves in Europe. That seems pretty far-fetched to me, and I'm more willing to believe one of the assumtions about ratios or estimations is wrong.

That's the whole point tho. The 'assumptions' don't make any real difference. There is no way to get around the fact that for example:

As of Sep 01 Sony shipped 8.24m PS2s and even the most generous estimates say that 5m could have been sold upto that point. So that would suggest 3m+ in "stock".

But we also have 10.59m shipped as of 6 months later (so only 2.35m shipped in between) whereas in reality, that period (Sep 01- Mar 02) was the real boom in PS2 sales in NA with at least 4.3m sold according to NPD. So in that period Sony sold 4.3m (+ Canada + NPD inaccuracies) but only shipped 2.35m? So where did the other 2.3m+ come from? Maybe the 3m+ Sony had in stock as in the above paragraph.

And the pattern continues over and over. Do people need me to draw a pretty graph to be able to see it properly? :lol

 

[Vennt]

If your graphing skills are as lacking as your mathematical skills I'd suggest you find another outlet for your frustration

 

[DarienA]

If your graphing skills are as lacking as your mathematical skills I'd suggest you find another outlet for your frustration

<rimshot>

 

[Ignatz Mouse]

That's the whole point tho. The 'assumptions' don't make any real difference. There is no way to get around the fact that for example:

As of Sep 01 Sony shipped 8.24m PS2s and even the most generous estimates say that 5m could have been sold upto that point. So that would suggest 3m+ in "stock".

But we also have 10.59m shipped as of 6 months later (so only 2.35m shipped in between) whereas in reality, that period (Sep 01- Mar 02) was the real boom in PS2 sales in NA with at least 4.3m sold according to NPD. So in that period Sony sold 4.3m (+ Canada + NPD inaccuracies) but only shipped 2.35m? So where did the other 2.3m+ come from? Maybe the 3m+ Sony had in stock as in the above paragraph.

And the pattern continues over and over. Do people need me to draw a pretty graph to be able to see it properly? :lol

Jinx describes the issues far better than I could.

I wold leave it at this, simply-- when multiplying factors, small changes make for great variation in the result. Coupled with the fact that we aren't really sure what any of the numbers really mean. I've seen it speculated that "shipped" for Sony means everything from "produced and left the factory" to "warehoused, shelved or sold by retailers" and we have no concrete info on which is true. We don't know real numbers for Canada. We don't know the methodolgy by which NPD extrapolates numbers.

We do know that you prefer Nintendo.

Given all that, I don't give you numbers any creedence. I could fudge them to the degree you have to show Sony's sold almost 100% of their shipped units.

 

[xaosslug]

"When nintendo fans have too much time on their hands.."

 

[ioi]

Rigor is about more than just being able to multiply two numbers together. I know this might blow your mind, but math is LOGIC. The bottom line is that your entire modeling process was not logical, and therefore lacks rigor.

I know a few things about modeling situations using math -- I was a physics major, after all. Since you asked, here is my list of complaints with your "proof."

1) First, you want this to be seen as a case of "my equations are rigorous but the numbers have assumptions," but that's not true. You have made a couple of very basic assumptions that the ratio of sales in the United States and Canada is constant, and that the month-to-month error in the NPD sales is constant. These are non-trivial assumptions which affect the rest of your calculations -- what if the relationships are periodic? linear? higher-order polynomial? uncorrelated? -- and I see absolutely NO justification for why you think modeling them as constants makes sense.

2) You round off in the middle of your calculations, which is a no-no. Even high school lab students know that you should report all of your data to all of its significant figures for calculation accuracy, and then round off to the lowest number of significant figures to report the final result.

3) Also from lab -- high school or early college, depending on which track you were in -- is the notion of error estimates. Each value has an uncertainty, which needs to be propagated through the calculation to come up with the uncertainty in the final result. When you claim that you might have 2.5M units on a shelf in Europe, it begs the question: 2.5M...plus or minus how many? Are you certain within a hundred thousand, or within a couple of million?

4) Of course, the problem is that you have no way of estimating what the likely error is in ANY of your numbers. For example, you make assumptions that the error in the NPD numbers is roughly constant over time and that 2-3 years is a sufficiently long period to come up with the mean, but provide no justification for either statement. Unless you have some way of comparing several NPD reports to ACTUAL, EMPIRICALLY VERIFIED values, I don't see how you know WHAT the margin of error is.

5) Finally, you use the phrase "simplify for simplicity" WAY too much. Using that phrase means that a) you understand the detailed phenomenology that you are modeling and b) can justify that you are ONLY dropping higher-order effects with your simplification and that your results will be within x% of the full calculation. I'm sorry, but claiming that you're doing things "for simplicity" is simply bogus. (No pun intended, of course.)

Honestly, I think the only thing you can show with the raw data is that shipped > sold...but that falls squarely into the realm of "no shit, Sherlock."

I also know a lot about modelling using maths, I'm doing an Engineering degree after all.

The math is rigorous. Rigor is about stating something that is correct. And one number times another that equals a result is correct. There is no way of getting around that. It's not my maths that you are questioning it's my assumptions, which you have admitted in a round about way.

1) I have made assumptions and I said they were assumptions and I have attempted to support them and provide some basis for them. The ratio between US and Canada is largely redundant in the end as this is encompassed in the overall ratio. The month to month ratio in NPD being consistant to a degree of error is a very reasonable one to make. We are not even talking month to month tho, but taking 6 month periods so any random fluctuations are likely to be averaged out even more over these 6 month periods. Even so, if the ratio is 1.5 some months and 0.9 other months (from a 1.2 average that is highly unlikely statistically) then it would actually make little difference to the argument. Looking at the numbers and the regular pattern of the "overshipping" in the Mar-Sep periods also supports this well.

2) Of course. The obvious problem here is that the data we are using is not accurate so rounding off is irrelevant. Despite that, I'm not sure what exactly has been rounded off. The data is all from tables in excel which calculate the figures exactly. I have rounded them to present them here at the END of the calculations made by excel to 2 decimal places for ease of reading. However this point is largely redundant since the data is so inaccurate in the first place and the aim of it is to show basic trends.

3) Again, largely redunandant in this situation.

4) Less than 4m units, that's the main thing. But again I think this point is largely redundant.

5) Not at all, I am understanding that anything that I am simplifying will have little effect on the overall point and am making a rational decision to do so. Again largely redunandant. I only mentioned the word simplicity once and it was in regard to NPD data being consistenly wrong compared to actual data. I have explained this in 1) above. We don't have any exact data so it is a simplification, but a reasonable one in my opinion.

What I have in fact shown is the pattern which Sony often ship their units in and that it's not uncommon to have 3-4m units that have been shipped and not yet sold through.

 

[ioi]

I could fudge them to the degree you have to show Sony's sold almost 100% of their shipped units.

I'd like to see that because it's impossible.

 

[Dilbert]

Maybe I made a mistake by typing so much, since you seem to have missed my point. I think this phrase from your response sums it up: "....the data is so inaccurate in the first place and the aim of it is to show basic trends."

Since you admit that the data is of completely unknown accuracy and you're making assumptions with no more justification than saying they are "reasonable" (which is circular), why should anyone believe your conclusions? Your "proof" is completely unconvincing for those reasons.

 

[Version 3.0]

ioi, I'm afraid you're just NOT going to convince Sony fans of anything. But, if you'd like to try, you're going to have to do this same thing with every system that has numbers available. Then you can show that the discrepancy between manufacturers' shipped numbers and NPD's tracking of sold units is always higher with Sony's systems than with others.

Although you shouldn't have to--it's blatantly obvious.

And then you can speculate why, and kick off an arguement about it for the millionth time on GAF. :lol

 

[Ponn]

"When nintendo fans have too much time on their hands.."

ZING!

Damn, don't even have to try anymore...just let them post and sit back and watch.

 

[ioi]

If your graphing skills are as lacking as your mathematical skills I'd suggest you find another outlet for your frustration

Can you understand that any better?

 

[HomerSimpson-Man]

ioi, I'm afraid you're just NOT going to convince Sony fans of anything. But, if you'd like to try, you're going to have to do this same thing with every system that has numbers available. Then you can show that the discrepancy between manufacturers' shipped numbers and NPD's tracking of sold units is always higher with Sony's systems than with others.

Although you shouldn't have to--it's blatantly obvious.

And then you can speculate why, and kick off an arguement about it for the millionth time on GAF. :lol

Or just read Jinx's informative analytical breakdown of ioi's approach. One of the things I've come to comprehend is never get into a science/math debate with -jinx- on GAF, you'ill get ass slapped again and again.

 

[sonycowboy]

Can you understand that any better?

Brosh,

I honestly think your posts would go over better if you weren't so darn arrogant. I should know, it happens to me as well (too often lately, unfortunately)

I think you've made a pretty decent point, but it's only one possible explanation and it's not particulary proof and yet you seem to get upset when people tell you that you've got quite a few assumptions that aren't supportable by any facts.

You really need to allow for the fact that this is simply your opinion (which has some analysis behind it) and that others opinions will vary. Neither is right or wrong necessarily.

And, please don't tell me that you actually believe that there was ANY POINT in the past 5 years where you believe that 4.2M PS2's were sitting on store shelves in the US.