Statistical Analysis of MacBook Pro Releases Dates

[eeMattS]

Like many of you, I am very excited about the MacBook Pro refresh. While waiting, I decided to do a quick statistical analysis of the MacBook Pro releases dates. I found the following.

Average days between releases = 197.75 days
Standard Deviation = 71.61 days
(The average number of days the release date deviated from the average days between releases, 197.75 days)

This makes 3/4/10 the date at the first standard deviation (197.75+71.61) and 5/15/10 the date at the second standard deviation (197.75+2*71.61). What this tells me is that the release is *almost* guaranteed before 5/15/10. I know that most of you assumed this as well but I thought I put some data behind it.

Attached is the excel spreadsheet I made for this data. It also contains a table and graph for the probability mass function. The probability mass function determines the probability that the release will on occur on X date.

Something I would like to see is the standard deviation for release dates in the MacRumor's Buyer's Guide.

Thanks,
Matt

[Moderator Note: This thread is for the stated topic only: looking at MBP release dates mathematically. It is not for general speculation about upcoming MacBook Pros.]

 

[vant]

Like many of you, I am very excited about the MacBook Pro refresh. While waiting, I decided to do a quick statistical analysis of the MacBook Pro releases dates. I found the following.

Average days between releases = 197.75 days
Standard Deviation = 71.61 days
(The average number of days the release date deviated from the average days between releases, 197.75 days)

This makes 3/4/10 the date at the first standard deviation (197.75+71.61) and 5/15/10 the date at the second standard deviation (197.75+2*71.61). What this tells me is that the release is *almost* guaranteed before 5/15/10. I know that most of you assumed this as well but I thought I put some data behind it.

Attached is the excel spreadsheet I made for this data. It also contains a table and graph for the probability mass function. The probability mass function determines the probability that the release will on occur on X date.

Something I would like to see is the standard deviation for release dates in the MacRumor's Buyer's Guide.

Thanks,
Matt

You can do all the analysis you want. At the end of the day, it comes down to Apple releasing it when they are ready. All this analysis is just pointless and a complete waste of time.

 

[iamamonkey]

What is the......... im not going to bother.

Apple will release it when Apple decides that it is ready.

To me all releases in the past is coincidental. Apple do not follow trends. They break trends.

Following a trend is everything that Apple is against, this goes from hardware, software and surprise surprise, release dates.

Yes, historically speaking the MBPs should already have been released, and yes it is late. Trying to predict when Apple will release the new iXXX MBPs is like trying to predict the lottery.

So to recap, what do we know for sure?
1) MBPs should have been released in Feb, but it wasn't.
2) Historically, they are late.

That is about it. Nothing we know is for sure. For all we know Apple may not even update the MBPs. Maybe it is not worth their effort. Maybe the market is saturated with laptops (PCs and Macs). Maybe they see the iPad as the new cashcow. Only Apple themselves know what they are doing, when they are doing it and how they are going to do it.

All this statistical analysis of uncertainty is getting tiresome.

 

[Capt Crunch]

Like many of you, I am very excited about the MacBook Pro refresh. While waiting, I decided to do a quick statistical analysis of the MacBook Pro releases dates. I found the following.

Average days between releases = 197.75 days
Standard Deviation = 71.61 days
(The average number of days the release date deviated from the average days between releases, 197.75 days)

This makes 3/4/10 the date at the first standard deviation (197.75+71.61) and 5/15/10 the date at the second standard deviation (197.75+2*71.61). What this tells me is that the release is *almost* guaranteed before 5/15/10. I know that most of you assumed this as well but I thought I put some data behind it.

Attached is the excel spreadsheet I made for this data. It also contains a table and graph for the probability mass function. The probability mass function determines the probability that the release will on occur on X date.

Something I would like to see is the standard deviation for release dates in the MacRumor's Buyer's Guide.

Thanks,
Matt

You didn't do a fitness test of a gaussian distribution. While it is a reasonable assumption, you should do something like a chi-squared test. You may want to try to fit the data to exponential or Erlang distribution.

 

[Skuman]

To me all releases in the past is coincidental. Apple do not follow trends. They break trends.

This is the perfect example of poor understanding of science that exists in society today.

 

[iamamonkey]

This is the perfect example of poor understanding of science that exists in society today.

Me or the OP?

 

[rata911]

Someone interested in calculating a Maximum-Likelihood estimation?

Yeah, me neither.

 

[Skuman]

Me or the OP?

You, and l'll explain why. It was not meant to be a personal attack, just an observation. If you have Excel, and have a look at the figures in the spreadsheet, you'll see that updates in the last ~7 years have been pretty consistent, around approximately 220 days, with a few exceptions of less than that (in some cases, much less). OP decides he wants to find out when the next release will be, within say x% of reliability, based on what has happened in the past. He works out the average and standard deviation. All pretty simple stuff, nothing too complicated. Yet you chastise him for trying to predict when a refresh will happen. He has done what scientists have been doing for the last 400 years. Trying to predict the future by looking at the past.

 

[iamamonkey]

You, and l'll explain why. It was not meant to be a personal attack, just an observation. If you have Excel, and have a look at the figures in the spreadsheet, you'll see that updates in the last ~7 years have been pretty consistent, around approximately 220 days, with a few exceptions of less than that (in some cases, much less). OP decides he wants to find out when the next release will be, within say x% of reliability, based on what has happened in the past. He works out the average and standard deviation. All pretty simple stuff, nothing too complicated. Yet you chastise him for trying to predict when a refresh will happen. He has done what scientists have been doing for the last 400 years. Trying to predict the future by looking at the past.

No personal attack taken from you. And I hope the OP does not take my reply as a personal attack either. I do have Excel and I do understand stats. (kindda).

Statistically the OP is spot on about his analysis. Maths and numbers do not lie (unless it is manipulated, climate-gate anyone?) I guess I am rather fed up with reading loads of predictions about the next MBP update and I took it out here. For that I am sorry, and I didnt mean to chastice the OP.

But I still maintain, Apple will release it when they are ready, not before. And to try (not the OP, but others) and hang, draw and quarter Apple for not following the "trend", based on stats is rediculous.

Stats help us predict the future (weather forecasts spring to mind) but in most cases, reality tells a different story. History teaches us lessons on how to deal with the future, but no one knows what the future brings. Similarly, statistically (and historically) Apple releases updates in Feb (or whenever they do) and on a Tuesday, so what do we do? We start saving up to buy a MBP in Feb. But then get dissapointed reality comes and Feb comes and goes and nothing happens.

What the OP did was to base his prediction on a statistical analysis of the past and all kudos to that. This is much better than "I gotta feeling" style predictions.

I guess I should have taken what the OP indended this post to be, i.e. an interesting statistical approach to predicting when Apple will release the MBP.

I am also waiting for the MBP refresh, but I have resolved to the fact that Apple will do what Apple deems best. Plus the longer Apple take to release it, the "newer" my MBP will be when I buy it during the summer. (lets hope they do not release it in the Autumn!!)

 

[runebinder]

Probably means absolutely nothing in the great scheme of things, however noticed in the UK the 15" 2.53GHz and 2.8GHz models are no longer available on the PC World/Currys websites . First time I've seen this.

Both 13" and the 17" models are available though. Make of that what you will.

 

[NJuul]

You are forgetting that to predict the future you have to look at the relevant data, which in this case *might* be days since last update, but is probably also related to the availability of new CPU's and/or GPU's, and for how long apple has had assemble them into new working prototypes.

 

[mrblack927]

Everyone here who says that statistical analysis is completely useless in this case clearly has a poor understanding of what statistics is...

Statistics does NOT tell you anything "for sure". It's a model for estimation. How accurate that model is, is dependent on what variables you factor into it. For example, if you were to do a multi-variable regression with day of the week, days since last update, proximity to other releases, economic climate, etc... you would probably get a pretty decent predictor of the mbp release date. That is because these variables ARE relevant, and they DO affect the probability of release.

I mean no offense, I'm just saying that those of you who say "you can't predict apple! they defy all trends!" are speaking out of turn. If that statement were true, then every day would have an equal probability of release. That means that it is equally likely that new mbps get released 8 months later, 25 years later, or 3 days later. Do you really think that is the case? Do you really think that apple might wait till June 2034 to release new mbps? Of course not, that date is less likely than April 2010. Doesn't mean it will happen in April... just means that the probabilities are different.

Those of you who are saying "apple will release it when they are ready" are missing the point. Yes. that is true, but "when they are ready" can be estimated statistically! It is unlikely that they will be ready 3 days after the previous release, just as it is unlikely that they will be ready 25 years after the previous release. So take your condescending attitude elsewhere, because OP was simply analyzing the probable interval of time that it will take apple to be "ready".

If you know what a normal distribution curve is, than you know what I mean. Apple absolutely does follow trends! In this case, the mean (center of distribution) for macbook pro refreshes is 8 months. Some of you obviously think that if apple doesn't release new notebooks exactly 8 months apart, then they don't follow a trend. That's completely false. The OP was spot-on with his analysis and I personally would be very surprised if apple went past two standard deviations without releasing the new models. So thank you OP, some people do appreciate your work.

 

[iamamonkey]

I have already apologised for jumping the gun and missing the point (something i do quite often). And I have also stated my reasons for jumping the gun.

I have also said that upon reflection I agree and have taken what the OP is trying to do here. which is look at it from a statistical point of view (which I appreciate), not another thread for started to predict the release of the MBP based on nothing more than a feeling or someone saying something to someone else.

Whilst statistically analysing this is interesting statistically, no one knows when the MBPs will be released except for Apple. We can predict it with some accuracy, and give it a probability of x% that is will be on "Y" of month "Z" in year "K", but my point it, no one can say for sure, not until Apple themselves says so.

 

[Dozerrox]

Everyone here who says that statistical analysis is completely useless in this case clearly has a poor understanding of what statistics is...

Statistics does NOT tell you anything "for sure". It's a model for estimation. How accurate that model is, is dependent on what variables you factor into it. For example, if you were to do a multi-variable regression with day of the week, days since last update, proximity to other releases, economic climate, etc... you would probably get a pretty decent predictor of the mbp release date. That is because these variables ARE relevant, and they DO affect the probability of release.

I mean no offense, I'm just saying that those of you who say "you can't predict apple! they defy all trends!" are speaking out of turn. If that statement were true, then every day would have an equal probability of release. That means that it is equally likely that new mbps get released 8 months later, 25 years later, or 3 days later. Do you really think that is the case? Do you really think that apple might wait till June 2034 to release new mbps? Of course not, that date is less likely than April 2010. Doesn't mean it will happen in April... just means that the probabilities are different.

Those of you who are saying "apple will release it when they are ready" are missing the point. Yes. that is true, but "when they are ready" can be estimated statistically! It is unlikely that they will be ready 3 days after the previous release, just as it is unlikely that they will be ready 25 years after the previous release. So take your condescending attitude elsewhere, because OP was simply analyzing the probable interval of time that it will take apple to be "ready".

If you know what a normal distribution curve is, than you know what I mean. Apple absolutely does follow trends! In this case, the mean (center of distribution) for macbook pro refreshes is 8 months. Some of you obviously think that if apple doesn't release new notebooks exactly 8 months apart, then they don't follow a trend. That's completely false. The OP was spot-on with his analysis and I personally would be very surprised if apple went past two standard deviations without releasing the new models. So thank you OP, some people do appreciate your work.

Thanks, I was reading the jabbings above and was about to reply but you did the job much more concisely than I could.

 

[Sawbuck]

Everyone here who says that statistical analysis is completely useless in this case clearly has a poor understanding of what statistics is...

Statistics does NOT tell you anything "for sure". It's a model for estimation. How accurate that model is, is dependent on what variables you factor into it. For example, if you were to do a multi-variable regression with day of the week, days since last update, proximity to other releases, economic climate, etc... you would probably get a pretty decent predictor of the mbp release date. That is because these variables ARE relevant, and they DO affect the probability of release.

I mean no offense, I'm just saying that those of you who say "you can't predict apple! they defy all trends!" are speaking out of turn. If that statement were true, then every day would have an equal probability of release. That means that it is equally likely that new mbps get released 8 months later, 25 years later, or 3 days later. Do you really think that is the case? Do you really think that apple might wait till June 2034 to release new mbps? Of course not, that date is less likely than April 2010. Doesn't mean it will happen in April... just means that the probabilities are different.

Those of you who are saying "apple will release it when they are ready" are missing the point. Yes. that is true, but "when they are ready" can be estimated statistically! It is unlikely that they will be ready 3 days after the previous release, just as it is unlikely that they will be ready 25 years after the previous release. So take your condescending attitude elsewhere, because OP was simply analyzing the probable interval of time that it will take apple to be "ready".

If you know what a normal distribution curve is, than you know what I mean. Apple absolutely does follow trends! In this case, the mean (center of distribution) for macbook pro refreshes is 8 months. Some of you obviously think that if apple doesn't release new notebooks exactly 8 months apart, then they don't follow a trend. That's completely false. The OP was spot-on with his analysis and I personally would be very surprised if apple went past two standard deviations without releasing the new models. So thank you OP, some people do appreciate your work.

Amen! It never ceases to amaze me that folks will look at two or three data points and call it a trend.

I was a Windows guy for YEARS - mostly out of habit - bought an iMac and there has been NO looking back. Got my iPod Touch a couple of weeks ago and carry it everywhere, looking at an new MBP in a month or two and potentially (eventually) an iPad just because the darn thing looks and works so slick.


For at lot of reasons I trust the folks at Apple to to what is best for the shareholders - and as a customer, I am okay with that since they meet my needs at the same time.

if there is a sinister plot to the MBP release date it is merely the marketing folks trying to get the maximum sales bump - and once released I will likely help them succeed. For me, have bitten the Apple? No Windows product will suffice.

 

[Mamdouh Medhat]

Like many of you, I am very excited about the MacBook Pro refresh. While waiting, I decided to do a quick statistical analysis of the MacBook Pro releases dates. I found the following.

Average days between releases = 197.75 days
Standard Deviation = 71.61 days
(The average number of days the release date deviated from the average days between releases, 197.75 days)

This makes 3/4/10 the date at the first standard deviation (197.75+71.61) and 5/15/10 the date at the second standard deviation (197.75+2*71.61). What this tells me is that the release is *almost* guaranteed before 5/15/10. I know that most of you assumed this as well but I thought I put some data behind it.

...

First, I need to declare that I strongly support the points in mrblack927's reply to the ones criticizing the use of statistical techniques in estimating the likelihood of a release on a certain day based on earlier trends - it was well put and captured the essence of the arguments. So, I'm not gonna dwell further on this, and move on.

I checked the numbers, and they seem solid enough from a computational point of view - however, and here's the real problem in my opinion, the calculations are based on a normal approximation to the distribution of the days between the earlier releases, and this approximation is, in fact, very poor.

Try to make a QQ-plot, or even just a simple histogram of the data - the empirical distribution is very skewed at best and bimodal at worst. We only have very few observations, but a simple bootstrap would confirm that the distribution is actually far more related to something bimodal, like a shifted Beta, for example, with one small mode at around 85 days and another larger mode at 235 days. Hence, the normal approximation is indeed very bad for this particular data, and so the estimated probabilities for the release dates (which are based on the quantiles of the fitted normal approximation) are a bit unreliable.

I'm really not sinking your boat, 'cause I really, really digg the idea of estimating the likelihood of a near-future release, but I'm just commenting based on my work with the data. I'm still trying to figure out a more proper way to do the estimation.

- Mamdouh

 

[silexh]

Stats

A higher resolution (1680 x 1050 or something) would indeed be nice, however, my current pc has a 1024x768 screen so everything is an improvement really...

But to get back on the topic of statistics and release dates:
According to the statistical data provided by eeMattS a couple of pages back (nice job btw), here are some probabilities for people who do not know that much about statistical analysis (and as a kind of recap):
on or before March 4: 84.2% (first standard-deviation from mean)
on or before March 9 (rumored date): 86% (educated guess)
on or before March 16(rumored date): 89.5% (educated guess)
on or before March 23(rumored date): 92% (educated guess)
on or before May 15: 97.8% (second standard-deviation from mean)
on or before July 26: 99.9% (third standard-deviation from mean)

Keep in mind these are still probabilities, so there is not much use to them... It might just help you to pass the time... (I might calculate the guesses later on)...
I've been following these threads for a while now and I just couldn't help myself.
(I do hope it's tomorrow...)

 

[savar]

Everyone here who says that statistical analysis is completely useless in this case clearly has a poor understanding of what statistics is...

Somebody else pointed this out already, but I'll repeat it since this thread is a million miles long and I doubt many people are reading more than 5-10% of it.

It sounds like somebody was assuming that the periods between Apple product refreshes are normally distributed, but it is pretty clear that they are not.

Without knowing anything about the underlying distribution

Also, when performing statistical analysis, one needs to be very clear about definitions. In this case, what counts as product refresh? What counts as an MBP? Because the MR buying guide includes PowerBooks in its MBP data series, and it also includes refreshes that only affect one of the 3 models. So when we base numbers off that data, its already got a built-in bias towards frequent updates of a fragmented product line.

Now that Apple's lineup is solidified, I would expect few if any partial refreshes, which means that the mean time between refreshes will be significantly longer than the historical average displayed in the MR buying guide.

If correct, that assumption would mean that historical data doesn't have much use in the current situation.

 

[Zipties]

Now on to something I would like to say about the statistical analysis of apple releases.

...
It sounds like somebody was assuming that the periods between Apple product refreshes are normally distributed, but it is pretty clear that they are not.

Without knowing anything about the underlying distribution
...

I have to say that the OP's work, though genuine, is flawed. Like savar said, the only way to assume that the distribution is normal would be to use the Central Limit Theorem (CLT), but the problem is that for the CLT to stand your sample size must be "sufficiently large." The sufficiently large typically means 65+ samples. We have 10 at best if we are to assume that the Power Book releases follow the same trend.