How Random Is Microsoft’s Random Browser Choice Screen In Europe? | Hacker News Reader

How Random Is Microsoft’s Random Browser Choice Screen In Europe?

techcrunch.com

6 pointsalexandros13 comments

The best link in the whole article is actually this one from the comments: http://sroucheray.org/blog/2009/11/array-sort-should-not-be-... which discusses the javascript array sort.

edit: this comment was not correct, while parent and reply are; apologies for the confusion =]

In an attempt to provide something non-incorrect, this is probably the simplest way in most imperative languages of doing a correct shuffle: http://en.wikipedia.org/wiki/Fisher–Yates_shuffle

Look further.

/resources/scripts/page.js, towards the bottom: function GenerateBrowserOrder()

Does it really matter how random it is? Most users will click whatever button will get rid of the message as soon as possible, which I suspect will end up not changing the default. Look at the pictures here:

http://microsoftontheissues.com/cs/blogs/mscorp/archive/2010...

Both a close button and a "select later" button, nice tempting targets for people who don't care about their browser and just want to search for 'facebook login' on Google.

I wonder if internet explorer is really the world's most widely used browser? I don't think so but I wonder if there's any data about this.

I have different data:

On one site I own, there are 50k visits, only the 12.56% were using any version of internet explorer, only 48% of this the version 8.

On another site I have analytics access, on 500k visits, 31% were using explorer, and of these, only 16% the version 8.

Both of these websites are not technical. The first one is a blog and the second one a social network website.

I have access to analytics for ticalc.org, a website frequented by high-school students who are looking for calculator programs (we rank #1 on Google for "calculator game" among other queries). I'm not sure if you'd call that "technical" or not.

The story here is similar. Of the 275,000 visits in the past 31 days: 39% came from from Firefox, 36% from IE, 12% from Chrome, and 10% from Safari.

>In fact, in over 50% of all page hits, Internet Explorer would come out to the far right spot of the five browser choices shown on the screen.

I thought for a L-R language reading right-handed person that this was the most prominent spot. Certainly I've observed most users I've done web training with to hold their cursor at the RHS of the page. Thus less effort is required to reach the right hand options.

Gaze tracking on actual users of the site would be interesting.

With 120 possible combinations, you will need about 10.000 users. Good luck.

summary of discussion: the top 5 are sorted by passing a comparison function to Array.sort() that returns a positive or negative number at random. at first glance this seems reasonable, but in fact whether it actually results in a random order depends on how the javascript sort function is implemented (and the initial order).

It's "obvious" (in a somewhat unreasonable sense of that word) that this can't give a truly random permutation of the elements of the array. Since the only randomness comes from coin-flips, the probability of any particular outcome must be a rational number whose denominator is a power of 2. But there are n! permutations of n objects, and for n > 2 this isn't a power of 2, so they can't all be being selected with equal probability.

i've been thinking about this....

take three coins, arrange them from left to right. i will show how to select one of the three coins with probability 1/3.

toss all three. repeat until only one shows heads. select that coin.

so i think you are wrong :o)

Can someone with statistics knowledge comment on whether these results are significant? With 10,000 trials they probably are, but it'd be nice to get more insight than raw percentages can provide.

An intern probably wrote the site, and MS doesn't care because no layperson is even going to read the screen.

Reply on news.ycombinator.com