Hi, I happened to come across this in my travels and thought you lot might be interested. Have fun!

I'm pretty sure someone already won. There was an article on slashdot maybe 2-3 days ago announcing it..

maybe they allow multiple winners?

-me

The contest runs until Oct 2, 2011. There are progress prizes though.

I find this sort of competition reprehensible. Basically, what they want you to do is provide to them an algorithm that will form the basis of their business for the forseeable future. If it's better than their current one by 1%, they'll give you $50k; if better by 10%, they'll give you $1000k. What do you have to give them? All rights to use the algorithm in any way they see fit... for ever. If you could achieve either of these improvement milestones, your algorithm would be worth far more than 1 million. Indeed, the sorts of money that the company would make from these improvements would be rated in the 10s to 100s of millions over the course of just a few years. Personally, if I achieved either of these milestones, I wouldn't submit my entry for final judging. I'd offer it to their competitors on a royalty basis!

Cheers,

Timkin

Quote: Original post by Timkin
I find this sort of competition reprehensible...

I haven't quite understood the legalese surrounding the "non-exclusive license." Here's what Netflix says:

Quote:
Why the non-exclusive license?

First, we want to verify for everyone that the code did what was claimed; that means looking at it. And then we want to use it if we can. We’re a business and we want to make sure we can capitalize on the discovery. But we don’t want to impede the winner’s ability to capitalize on it as well. Actually, we hope they can build their own business and license it to others as well. That is the point after all.

It sounds like on the one hand a non-exclusive license allows Netflix to do anything they like with the code, without ever paying another dime, from here on after. On the other hand, it sound like the original creator also has the right to use the algorithm for whatever purposes they see fit (they just can't sell it again to anyone else, I think).

Anyway, legalese aside, who's going to win the prize? Probably some very bright kid with much time on their hands (read: no heavy job), who would never have worked on this algorithm if it weren't for the prize. So he walks away with a million smackers for a few months work.

So the idea is given to him, he doesn't need to worry about licensing it and selling it around for royalty, and he doesn't need to worry whether the company really will make an extra 10 million off of his algorithm in several years, as they would need to if he wanted to make his million with a 10% royalty (probably would be more like 3%).

Personally, I think it would take quite some time for Netflix to gain 10s of millions, and certainly not 100s of millions, for a modest improvement in their recommendations system. They currently make a profit of 55 million a year, according to InfoTech. A 10% improvement in the recommendations system might earn them, say, 5% more sales (optimistically). Doing the math, I wouldn't sell this for royalty if I created the algorithm.

So I think it's a good idea. If I were a real wizkid and had a lot of time on my hands, I'd try it myself.

[Edited by - Asbestos on October 13, 2006 10:21:24 AM]

...except that a 'whiz kid' has very little chance of winning this. The teams that have already shown improvement over Netflix's classification method are comprised of data mining experts and academics. These people do it for the kudos more than anything and the publications that result. The likelihood that they would be pipped at the post by some kid working for a few months in his backroom on a home PC is very low.

As to the profit projections, you're ignoring market growth. Home delivery of DVDs is taking off around the world and is driving the local video store out of business at an unbelievable rate. Within 10 years, it is unlikely that you would find a mainstream video/dvd rental store occupying a local retail outlet. It simply wont be economically feasible.

One final comment...on the licensing... the license you give to them is to look over all of your work...and you agree to provide to them a very detailed account (including all mathematical proofs, code, algorithmic descriptions, etc) of the method so they can validate it. The license you agree to give them also permits them to create any and all works that they see fit based on residuals; that is, the knowledge they gain in their brains from looking over your work... and you agree that you will never charge them anything for this, now or in the future. They also require you to publish a detailed description of the method to the public at large (which legally speaking places the methods in the public domain and removes any and all IP rights you have over it). Basically, yes, they don't mind if others have access to it... but they want it for themselves principally so they can make more money from it. Why else would they be doing this? Certainly not for the betterment of mankind!

Quote: Original post by Timkin
The teams that have already shown improvement over Netflix's classification method are comprised of data mining experts and academics... The likelihood that they would be pipped at the post by some kid working for a few months in his backroom on a home PC is very low.

Good point.

Quote: As to the profit projections, you're ignoring market growth. Home delivery of DVDs is taking off around the world and is driving the local video store out of business at an unbelievable rate. Within 10 years, it is unlikely that you would find a mainstream video/dvd rental store occupying a local retail outlet. It simply wont be economically feasible.

That's true, but I still wouldn't want to stake my retirement on the royalties that may or may not be generated by my algorithm. For starters, it would be very difficult to prove exactly how much revenue would be generated by the algorithm. A 10% improvement in the recommendation system would not equate to a 10% increase in sales. Sure, you and I could think of some rational system (the percentage increase in the number of selections from the recommendation list, say), but Netflix would probably be able to launch a convincing counter-argument to its being less.

Also, relying on royalties is essentially placing a bet that no one else will ever come up with a better solution. Unless I can mathematically prove that there can never be a more optimal solution, that isn't something I would bet on.

Quote: Basically, yes, they don't mind if others have access to it... but they want it for themselves principally so they can make more money from it. Why else would they be doing this? Certainly not for the betterment of mankind!

Right, as I said above. Indeed, the notion that they "hope you can build your own business" from the algorithm is a little silly, as so can anyone else.

Residents of the province of Quebec in Canada are ineligible to participate.

Booh :P We're always left out hehe.

Quote: Original post by xEricx
Residents of the province of Quebec in Canada are ineligible to participate.

Booh :P We're always left out hehe.

Join the club. Try entering almost any competition conducted by a U.S. company that sells a product internationally and you'll find that 999 times out of 1000 you cannot unless you live in the U.S. Something to do with their stupid tax laws, or so I've been told.

This contest cracks me up. One million dollars? I'm sorry but if you develop a system to provide recommendations better than what Netflix has presently it would be worth way more than a million dollars. You'd have no problem at all finding financing to found your own corporation. Licensing to clients in the first year would be at least $10 million in revenue.