How random is Microsoft's EU browser ballot?

TechCrunch put out an interesting piece regarding Microsoft's upcoming browser ballot. The ballot will allow a user to pick their computer's default browser through a well laid out menu. While the ballot is set to begin public testing this week, many have wondered just how "random" it really is.

DSL.sk decided to test the ballot's true randomness by using the site www.browserchoice.eu (which is the most recent implementation of the actual browser ballot). They, essentially, loaded the page over and over, gathering the results into a nice little table for all to see. The probablity of each browser appearing in any of of the first five spots is shown below.

EU Browser Ballot Stats

As seen from their results, Microsoft is far from playing IE favorites. The ballot seems to favor Google Chrome in the first three spots (overwhelmingly so in the third spot). However, the tests were done using Internet Explorer 8 on a Windows 7 machine. When another browser, such as Firefox, was used for testing, TechCrunch says that the results came out much different. Perhaps this could be due to the way that each browser's rendering engine handles the algorithm. It would be interesting to test such a theory.

TechCrunch notes that just because a browser shows up in the first few spots doesn't necessarily mean it's more likely to be chosen. It really depends on many other factors, such as where a person's eyes are focusing at the time. When all is said and done, the above results aren't necessarily so relevant. However, if Google's browser market share begins skyrocketing in Europe, they should probably send Microsoft a "thank you" cake.

In other news, Mozilla claims that 77% of Brits are unaware of the upcoming browser choice ballot.

Report a problem with article
Previous Story

China responds to Google's accusations as 'groundless'

Next Story

Natal possibly coming in October; 1/10 of a second lag

81 Comments

Commenting is disabled on this article.

I have heard about enough of all this crazy ballot screen business... how about working on getting a ballot screen on Snow Leopard before complaining that the one in Windows is not "random" enough...

Oh wait, did anyone else fail to realize that the ballot screen (Most likely being a small microsoft developed app) is powered by IE?

Then MS pushes an update a month into the ballot screen which favours IE by displaying the logo everywhere.. oh wait, they already do that...

/facepalm.jpg

Some people just WANT Microsoft to be cheating.
So they will have something new to bitch about.

k7of9 said,
/facepalm.jpg

Some people just WANT Microsoft to be cheating.
So they will have something new to bitch about.

Microsoft has been caught with its hand in the cookie-jar more than once. Maybe the out-of-band IE8 update of today for json (KB976662) is intended to "fix" the outcome of the ballot ?

"As seen from their results, Microsoft is far from playing IE favorites. The ballot seems to favor Google Chrome in the first three spots"

I'm not sure if the "first three places" really is a good measure of anything. If it was me I'd imagine the big places to be would be 1st, 3rd or 5th given that takes care of the extremes or the middle. 2nd and 4th positions would be the ones I imagine people don't look at quite as quickly...at least if they are only glancing at it.

Depends if you want to be one of the first ones to be read or whether you want the browsers image to stand out I guess.

That "researchers" are just so STUPID.
What about just looking at the page SOURCE! The random sorting is done on their own machine in their own browser!
A 10 years old kid can understand the code! http://www.browserchoice.eu/resources/scripts/page.js

[code]
// Genereate random browser order
function GenerateBrowserOrder()
{

var aBrowserOrderTop5 = new Array(0,1,2,3,4);
var aBrowserOrderRest = new Array();

for (var i=5; i < dataBrowsers.length; i++)
{
aBrowserOrderRest.push(i);
}

aBrowserOrderTop5.sort(RandomSort);
aBrowserOrderRest.sort(RandomSort);

aBrowserOrder = aBrowserOrderTop5.concat(aBrowserOrderRest);
}

function RandomSort (a,b)
{
return (0.5 - Math.random());
}
[/code]

RealFduch said,

[code]
function RandomSort (a,b)
{
return (0.5 - Math.random());
}
[/code]

The fact the custom sort code only takes very simple a/b item arguments and cant return if the sort is complete, I assume sort() is a bubble sort. How can bubblesort know when a random sort is completed unless it lucks a full run of all item compares returning 0 or some form of retry count being exceeded?

So lets assume the sort compares are run a whole heap of times, but it will be run a static constant figure each time.
0.5 - Math.random(), will favour positive results meaning things are pushed
random() returning 0.00-0.49~ is positive, 0.50 is 0 and only 0.51 - 0.99 is negative.
On the first sort cycle of the array the first item has a higher chance of being last than any other item.
Eventually you WILL see a pattern due to the static count of cycles of the bubblesort before it timeouts or lucks a perfect run(this is super super unlikely). The exact static cycle count will determine on paper exactly which browsers are preferred to be at the end. The end is the most forced position due to favouring pushing the first item to the end per cycle. 4th position is 2nd most favoured position and so on.

The combination of bubble sort that cant effectively end its sort(as it is never actually sorted) so requiring a fixed timeout cycle count coupled with the flawed RandomSort is exactly the end results you can see in the graph.

Actually bubble sort has a known end game count, its not a timeout. Still its a static count of compares that are returning weighted results :p

Looking at the order the browsers are in the array it goes:
IE, Firefox, Opera, Google, Safari

It's very late but my head says the results are weighted to pushing the first item to the end eg:
Firefox, Opera, Google, Safari, IE

In other news, Mozilla claims that 77% of Brits are unaware of the upcoming browser choice ballot.

This doesn't surprise me at all, and pretty much sums up the average user's feelings about the whole browser ballot debacle. The only people who really give a dam about this are the other companies. The ballot screen is only going to confuse people. Yes, some may well try another browser that they may of heard something about, but despite whatever features said browser brings to the table, I'd wager most users will go back to IE simply because it's what they know and are comfortable with.

Question: If I currently don't have IE set as my default browser, will I still see the ballot screen?

I got it the other day but it wasn't working however it's just started to work.

There's no set order in which the browsers appear. I've just opened it three times and each time the first five choices were in a different order

This whole thing makes me want to punch someone in the face. Seriously STFU Opera and all the other whiney companies that dont know how to compete in today's market. the EU should be shot for forcing a company to not bunlde their own software with their own operating system. And seriously, who the hell cares about just exactly how random randomized really is? SERIOUSLY? This remindes me of all the retards at my old job that reply to all saying "You misspelled a word in your e-mail" If you have nothing better to do than refresh a browser 100 times and formulate a percentage of which one appears in the first box, then you need to find a damn hobby.

oh good greif... yeah and random numbers arnt really random on computers... unless you get into Cryptographically secure pseudorandom number generators and those still arnt truely random... just more random then your standard rand() function would give you...

Webworldx said,
Run the test 10,000 times in Firefox. Completely different result i'm sure, and points to the IE JScript implementation.
Agreed!

Those insisting that out of whack numbers like over half of TENS OF THOUSANDS samples showing IE in one spot are somehow evidence that this is truly random is truly insane. ;)

PureLegend said,
Surely if the results were evenly distributed that would raise some warning flags.
The 50% chance of IE appearing in position #5 is more than just a statistical anomaly, unless the sample size on this is very small.

This indicates that the random number generator is rather poor. Perhaps using the microsecond of the request to the server would have made things a lot more evenly distributed.

Although the article is is great interest (despite the whiners bitching and moaning otherwise) because it shows Microsoft's tools putting their own product at a *significant* disadvantage.

Very interesting, indeed!

java2beans said,
They should have the category "Other" just in case another browser comes up.

The site does list other browsers already. They just aren't ever listed in slots 1 thru 5.

Why are they discriminating between the browsers in the first 5 slots and in the second set of slots? Isn't that oligopolistic? The 5 browser companies are colluding to make sure they keep the market share among themselves.

brianshapiro said,
Why are they discriminating between the browsers in the first 5 slots and in the second set of slots? Isn't that oligopolistic? The 5 browser companies are colluding to make sure they keep the market share among themselves.

The other browsers are simply wrappers. Who cares?

brianshapiro said,
Why are they discriminating between the browsers in the first 5 slots and in the second set of slots? Isn't that oligopolistic? The 5 browser companies are colluding to make sure they keep the market share among themselves.

They aren't discriminating. They decided to pick the five biggest browsers.

The other ones are mostly just wrappers anyway, and some of them aren't even translated into most European languages. Only the top 5 browsers are properly translated, supported and developed.

This makes me want to cry a bit inside. Random numbers are not completely random. There have been many articles on this before. This is NOT something new.

http://en.wikipedia.org/wiki/Pseudorandom_number_generator

Louis M. said,
This makes me want to cry a bit inside. Random numbers are not completely random. There have been many articles on this before. This is NOT something new.

http://en.wikipedia.org/wiki/Pseudorandom_number_generator


I also personally love that people think that just because its of thousands of requests that that means theres any more change of each browser being distributed equally than there is of Chrome showing up every time

The code makes simple use of Math.random(), a function that's implemented by the browser, not the author of the page. Given the differences in browser javascript implementations, it's obvious why there are differences. What a non-story. Shame it's on Neowin's front page.

It would have been a lot easier for them to skip quickly through the source, see that it uses a random function, and be done with it. Waste of time.

The (pseudo) randomising is done with javascript so this test was a total waste of time. If the ballot screen was truly random would we be expecting each browser to come out an equal percentage of the time anyway?

Perhaps they should disable javascript and write an article about how IE8 is always first too. ;)

ziadoz said,
If the ballot screen was truly random would we be expecting each browser to come out an equal percentage of the time anyway?

Actually, by definition of random, this would not be the case. What you are talking about is the myth that equal distribution is the same as randomness. If Chrome comes out first 8 out of 10 times, this says nothing.

They always will.

That ballot screen could have Opera listed 25 times in a row and I still wouldn't pick it.

There was an article on here some months back where Opera admitted that they didn't have the funds necessary to advertise the browser, or make deals with OEMs to have it included on new machines. If you can't afford to do business, you shouldn't be in business.

Edited by iamwhoiam, Feb 23 2010, 5:25pm :

warwagon said,
It might be stupid, but you also have cry babies like opera who whine and complain that they are last on the list :)
Opera started all this, and has the lowest market share.. I would have just not included them at all.. See if the EU would really push it.

rm20010 said,
So the other shells shouldn't be on there? Right.

IMO, no.

News flash...just because something is included within the OS, it does not mean that you have to use it. Anyone that has any common sense knows there's usually more than one of any particular item. So it would stand to reason that if there's one browser, there's bound to be more. If someone fails at decision making when it comes to what web browser to use, then the OS maker should not be held liable for it.

Edited by iamwhoiam, Feb 23 2010, 6:11pm :

warwagon said,
It might be stupid, but you also have cry babies like opera who whine and complain that they are last on the list :)

It's stupid to make it random? Why?

Ryoken said,
Opera started all this, and has the lowest market share.. I would have just not included them at all.. See if the EU would really push it.

Opera does not have the lowest market share. At least not in Europe.

And Opera didn't really start anything. They reported a crime to the authorities, and the authorities took it from there.

And remember, it was Microsoft who proposed the ballot screen in the first place. So according to your logic, IE should have been left out.

Yawn.

Soldiers33 said,
the user picks what he wants, not whats first on the list
I wish this were true. The browser ballot is really aimed at people who don't realize that there are other browsers out there besides IE. Because these people are clueless about other options, the placement of them on the screen when giving a choice DOES, believe it or not, make a difference.

There have been plenty of studies done on election ballots that show that people do tend to select options based on their location on the list.

Benjamin Rubenstein said,
I wish this were true. The browser ballot is really aimed at people who don't realize that there are other browsers out there besides IE. Because these people are clueless about other options, the placement of them on the screen when giving a choice DOES, believe it or not, make a difference.
I thought this was a free marketing ploy by Opera and the hope that they can gain 5 more users. Huh, silly me.

Edited by zeke009, Feb 23 2010, 8:01pm :

artfuldodga said,
funny, thats not how i or anyone i know vote lol

The studies were more aimed at people who were undecided, I think. Even so, the position of options on a ballot can influence how often they are selected.

Literal translation of what was written - "HEY U KNO WAT DA ODER DAY I FLIPPD A COIN AND IT LANDED HEADS ALL 4 TIMES I MEAN WTF GOD IS PLAYIN WIT ME MAN"

Edited by unity+, Feb 23 2010, 4:25pm :

Trance. said,
Literal translation of what was written - "HEY U KNO WAT DA ODER DAY I FLIPPD A COIN AND IT LANDED HEADS ALL 4 TIMES I MEAN WTF GOD IS PLAYIN WIT ME MAN"

I just did this, and on the 4th time it finally came up as tails.

There just isn't any way a coin could land heads 4 times. Let's face it.

Harreh said,

I just did this, and on the 4th time it finally came up as tails.

There just isn't any way a coin could land heads 4 times. Let's face it.

Mathematically speaking, there's a chance of 0,0625 (0,5^4) of tossing heads four times in a row. So yeah, there is a way. :P

Trance. said,
Literal translation of what was written - "HEY U KNO WAT DA ODER DAY I FLIPPD A COIN AND IT LANDED HEADS ALL 4 TIMES I MEAN WTF GOD IS PLAYIN WIT ME MAN"
No. It is not.

Have you ever taken any statistics classes? If so, do you mind telling me what grade you got in them?

A sample size of "tens of thousands" *IS* statistically relevant in marking the location of 5 objects (browsers). Your sample of 4 as an equivalent to this is a laughable fallacy.

markjensen said,
No. It is not.

Have you ever taken any statistics classes? If so, do you mind telling me what grade you got in them?

A sample size of "tens of thousands" *IS* statistically relevant in marking the location of 5 objects (browsers). Your sample of 4 as an equivalent to this is a laughable fallacy.

I was mocking them with a hyperbole.

markjensen said,
No. It is not.

Have you ever taken any statistics classes? If so, do you mind telling me what grade you got in them?

GOSH it's just a joke. Get over it.

Edited by Ricmacas, Feb 23 2010, 7:45pm :

Ricmacas said,

GOSH it's just a joke. Get over it.
Missing joke tag. Or winky face. Or any other method of separating out one ridiculous post over another said in seriousness.

Sorry that my reasoned response bothered you enough for you to tell me to "get over it".

Keine Lust said,

Mathematically speaking, there's a chance of 0,0625 (0,5^4) of tossing heads four times in a row. So yeah, there is a way. :P

Exactly... its a long time since I studied statistic but I think you only have to flip a coin 7 times before it becomes equally likely you will get 4 in a row as you will not get 4 in a row at some point in the "flipping".

Also btw... flipping is a flawed equal game for the same reason toast always lands butter side down, "flipping" by a human is not equal. In fact a well trained person could flip a coin almost every time the way they want it, its actually a thing which can be controlled quite precisely. To make it more even you could try tossing it down the stairs, rather than catching it.

Edited by lt8480, Feb 23 2010, 10:39pm :

Solid Knight said,
Flipping a coin isn't random. It's the result of various inputs.

qft

+1 for determinism theory

not knowing the result beforehand doesn't mean magic or randomness is involved.

then again, a "randomness generation" in a pc, just is the same. the result of operations to determine a result based on a bunch of given (programmed) parameters (can be anything, for example the system time, etc...). well, cheers, here goes your randomness

... and the magic is gone.

Glassed Silver:mac

Redz0ne said,
This is just getting ridiculous now!!!

research on how random they are!!!! really they need to stop this... ballot is pointless..