Humble Bundle Data - Android 4

I did it again.  This time, the bundle was a games bundle (The Humble Bundle for Android 4), which is Humble's normal fare.  Also this time, I was able to collect data from the very beginning.  There was a period of 7 hours where my server got turned off and no data was collected, but that time slot was in the middle of the data collection with no major events occurring near it.  I patched it up with a little bit of linear interpolation.  It shows up on the Marginal Average Price as a plateau, but is otherwise unremarkable.

One thing about this graph that I was not able to capture last time is that the average price was actually highest at the beginning, before sinking to a low, and then slowly rising.  The "event" in the middle is when games from the previous Android bundle were added as an additional bonus.  Interestingly, I was able to purchase at the point of lowest average price though I did beat it by making my purchase price a nice round number, thereby contributing to the bounce-back of the average price from its initial fall.

The initial fall is probably due to a bunch of people paying $0, or $1 immediately just to get the basic games and/or Steam keys.

The ramp-up of initial purchases is quite high, as you can see.

In this case at least, the initial "bump" was much more significant than the subsequent "blip" produced by adding more bonus content.  A lot of people already have the previous bundle games, so this is somewhat expected.

Here again is the raw data, for those who might be interested:

• humblebundle-android4.csv
• humblebundle-android4-2012-11-22-1445.xlsx

Humble Bundle Data - Results

In my previous post, I said I was going to finish collecting data for the rest of the Humble eBook Bundle at humblebundle.com and post the results here.  

The results are in.  This was probably the most successful Humble Bundle to date, based on the ending average price.  Let's look at the data for the Average Price over Time.

Average Price over Time

You may notice that bump in the middle of the graph.  That is the point in time when several PDF comic books were added to the Humble eBook Bundle.  This had the effect of pushing a lot of fence-sitters over the edge to purchase the bundle, as well as increase the margin by which purchasers were willing to "beat" the average price.

Total Purchases over Time

You may notice that the Total Purchases graph has some missing data at the beginning.  This is because initially I only collected the Average Price (and I was over an hour late in starting that collection).  The bundle started at 10:00 AM PDT; I started recording the average price at 11:30 AM, and I started collecting the total number of purchases at 5:30 PM. 

Total Revenue over Time

With those two numbers, I was able to calculate the total revenue collected.  Later, I added direct collection of this figure.

As you can see from the graph.  At no point did the average price go down much at all (there were a few times it went down a penny or two).  This answers the question I was initially asking, at least for this bundle:  should I wait for a lower price?  The answer is of course emphatically no.  If I was going to beat the average price, the time to do it was as early as possible.

I do seem to recall bundle average prices going down in past bundles, but this may have been due to abuse by people pumping the system for free and/or very cheap Steam keys.  That practice seems to have been cracked down upon with CAPTCHAs (remember: only use your scripting powers for good), and the momentum of the price and interest in the bundle seems to have been maintained by the addition of the bonus content.  I would expect similar measures in the future.

As it turned out, I when the extra bonus content was added, it was also added to the account of everyone who had previously purchased the bundle, whether they beat the average or not, so I ended up with all but the initial two bonus books.

These marginal rates were all calculated from the previous values.  It would be interesting to have better data at the beginning of the data set.  I'm curious to know how the profile of the initial wave compares to the second bump.  You can't really tell with the first seven hours missing, unfortunately.  My suspicion is that the secondary bump was sharper than the initial wave, mostly because I believe that the secondary wave was largely fence-sitters who had not bought because they thought the price was too high for the content offered.  When more content was offered (and considering the content), they immediately jumped on it.

Marginal Purchases (every 15 minutes)

Marginal Revenue (every 15 minutes)

Marginal Average Price (every 15 minutes)

It should be noted that the website data is far from perfect.  The totals sometimes went down from one reading to the next, and after the bundle had ended, the numbers were still in flux for several hours, but at the scale of these graphs, the fluctuations are insignificant.

If you would like to look at my raw data, I will provide it for download in its unprocessed CSV format generated by my script, as well as the Excel spreadsheet that I used to calculate the missing values and create these fancy graphs.

Here is the data:

• humblebundle-ebook.csv
• humblebundle-ebook-2012-10-23-1615.xlsx

Here is the final form of the script I used to create this data:

• humble-bundle-price-tracker.sh

Enjoy!

Humble Bundle Data - Collection

Today was the launch of the Humble eBook Bundle.  The way that Humble Bundles work is that the site lists a collection of downloadable items (traditionally games, but lately they have branched out into music, and now books).  You can pay any price (including free, if you choose) for the bundle, and it's yours.  You can even tweak how much of your purchase price goes (directly) to the content creators, the Humble Bundle site, and a few charities.

Well, not all of the bundle is yours at any price.  There are bonus items, which are typically the best of the bunch.  To get these items, you are required to beat the average price of the bundle so far. 

Well, I like bundles, and I like e-books, so I decided to get the bundle.  But I wasn't prepared to pay the price at the time to get the two bonus books.  They looked good, but the average price was then just over $9, and I thought that if I was going to spend that much money on an e-book, did I really want the ones offered?  I would be willing to bite--but only if the price went down.  Humble Bundles typically hover around the $5-7 range for unlocking the bonus content, which for my money is a better impulse purchase for content I didn't get to pick out myself.

Funny thing about having to beat the average price:  a lot of people do it, and that will have a tendency to create a constant upward trend in the price to unlock the bonus content.  It got me wondering, though.  I fully expected the price to trend up for the first couple of days, but then does it dip down again?  My gut told me that there would be an initial spike, a dip, and then a spike at the end of the two-week window when the bundle was offered.

But I couldn't find any data on the subject.  There are several websites where data on the various Humble Bundles can be found, but a graph of average price over time was nowhere to be found.  "Well," I thought, "I'll make one then."

So I did.  Or rather, am.

My data collected at the end of day 1, showing average price and total purchases over time.

I whipped up a quick script on my Linux box that uses links -dump to grab a text-only version of the web page.  It then successively greps this page for the data I'm after, and appends a line of data to a CSV file.  Initially, I only collected the timestamp and average price, but I decided that the total quantity sold at that time would also be a valuable and relevant piece of data.

I added a line in my crontab file to run the script every 15 mninutes (which is plenty of granularity over two weeks, and also very reasonable to the remote server), and viola:  data!

It will be interesting to see how this little experiment turns out.  It may be that the price will almost always trend upward.  I know that with certain past gaming bundles, games from previous bundles were added to the current bundle as bonus items in order to motivate people to pay the higher price and keep the average up.  I suspect that these items were added at times when the average price was dipping in order to bring it back up.  I don't know that there are any books in reserve for this bundle, so that may not be an option.

If it does happen, that will be interesting to watch in the data.

Side note:

• Another bundle site of note that just got started specifically for e-books is StoryBundle.  StoryBundle is slightly different, in that they set a minimum price ($1), and the bonus books can be unlocked at a constant price ($7). 

I will post an update to this blog when the bundle is over with the full results of the data I collect. 

Update: Here are the results.
Enjoy!

Ladies and Gentlemen, the Ansible

I hate sensationalist newspeddling, especially when it's scientific news. Reporters say the silliest things. No, Doc won't be able to get that flux capacitor working in his De Lorain, but that doesn't mean that the experiment described in this article is any less cool.

It all has to do with the speed of light. Nothing, incuding information, can travel faster than the speed of light, but apparently, some scientists are going to try to get information to travel backwards in time along a beam of light.

The reason this might work is that according to Einstein's Relativity, from the perspective of a photon, no time ever passes. Photons don't experience time, because they are travelling at the absolute speed limit of the universe. Absolute speed means infinite time dilation. From the perspective of a photon, zero time passes between the time it is emitted and absorbed. In fact, from the perspective of a photon, the photon doesn't travel any distance either: it's just a bit of energy that doesn't happen to be matter. It hops from its emitter to its final destination in a single moment. If its destination forces its spin to be a certain direction in order to be absorbed, then that's what the spin will be throughout its entire existence: no matter how far or long from our inertial reference frame we think that photon travels.

So, if we can affect the properties of the entire existence of the photon at the absorption point, and if we can entangle a pair of photons so that they share properties, then it is possible to construct an experiment where we cause entangled photons to follow different paths. If the path from which we cause the property to be fixed is longer than the path from which we observe the effects of the property, then it is possible for an event to be caused by an event in its objective future, at least, from our reference frame. (From the reference frame of the pair of photons, everything will be happening simultaniously.)

So, what's it good for?

One thing it's not good for is changing events in the past. At best, what you would get is a way for information to be passed instantaniously over great distances, or to place-times in the past that couldn't have affected you with the information yet. It takes several seconds for light to travel the distance between Earth and Mars. Imagine being able to control and monitor a Mars Rover in real time from Earth. Imagine being able to teleconference with your friend on a planet orbiting a star dozens of light-years away. What you'll be doing is monitoring the state of photons that were emitted decades ago by a waystation roughly half way between the two of you, but the state of which was fixed moments ago on the other end. The beam would have had to have started transmitting decades before it reached you, but you would be able to effect the states of the photons received on the other end in the present.

Imagine getting WoW ping times in the 100 millisecond range on an overseas server. Oh, so now I have your attention!