Betting on Search

My post on Saturday about pre­dic­tion markets being a useful way to access col­lective intel­li­gence brought a response from Gary of Tall Street. Tall Street is a new search engine which operates a form of stock market on search results. You search for and add sites to the system and invest pretend money in the sites you like or own. If other users click sites you’ve invested in, your stock increases in price.

They haven’t been Techcrunched or boing­boing’ed as yet so the number of sites in the system is low, which limits the utility of the service. However, a search for ‘Search Engines’ has been pre-​​populated. It would be useful if there were a way to seed the system with results from another service. I also think real money would be an inter­esting addition. As it is, only the ongoing addition and betting on sites by suf­fi­cient users will reveal whether the model actually works. But I have to say.. it looks good on paper. Would I invest real money in Tall Street? No. But then I’m not an investor on the lookout for innov­ative new ideas.

Yes, it’s just another student project, and it looks a bit rough. But haven’t we seen some­thing fitting that descrip­tion before…?

Gary was also good enough to answer a few ques­tions for me:

Could you explain briefly how tall street works.

Traders make invest­ments on sites belonging under keywords. They get rewarded (their invest­ments increase) if when people who search those keywords find their sites useful (based on if the link gets clicked and any ratings the link gets) and punished if not.

Can pre­dic­tion markets work for search when there are so many ‘correct answers’?

This is a problem any search engine struggles with.

The ideal situ­ation is:
The user has some idea about what they want, they express this in “search terms”, the search engine returns the best answer for what the user wanted.

A search engine works by:
Returning all the pages that contain that term (or had that term in the anchor text to that page). Then it ranks those pages based on some algorithm which takes into account the number of pages linking to it, (and linking to those pages) among other things.

Tall Street works by:
Returning all the pages that someone thought belonged under that search term, and the rankings are influ­enced by

1) how many people thought that site belonged under the term (if lots of people invest in a site it gets lots of invest­ment then it gets a high ranking),

2) how much they thought the site should be ranked in com­par­ison to other sites under that term (some unpop­ular terms might only have one investor in, and they could com­pletely control the rankings for that search term, this is fine since if someone dis­agrees then they are able to invest more heavily and change the rankings),

3) how much influ­ence (money) the investors investing in the sites under the term have over the dir­ectory i.e. their net worth (if that investor is good investing in popular sites then over time they will get more influ­ence since they will accu­mu­late more money)

Why is this better?

Rankings are up-​​to-​​date — the link struc­ture of the web that search engines analyse to determine rankings isn’t dynamic — links remain long after a site moves. What could happen with Tall Street is if some site added a new cool feature that no other site under the keyword had, then invest­ment in that site could increase and it’d end up at the top of the results straight away. You wouldn’t see some­thing like this in the link struc­ture until long after. ( i.e. The feature would have to be very cool that people decided to link to the site and then the search engine would have to recrawl the web, and then update the results)

Users can find sites by type and easily find similar sites — Sites are under keywords because people thought that the site belonged under the key words. With search engines you get lots of results many aren’t relevant. They just appear because they contain the term you are searching for on the page. We have a feature where you can drill down to find pages under keyword1 and keyword2, which lets you easily find similar sites e.g. http://www.tallstreet.com/view/PHP/tutorial/
Fair Ranking system — Everyone gets an input into the rankings. If you think a site doesn’t belong, you can mark it as spam. If you find a site that is better then what is showing as the best then you can sign up and invest in it. If you’re right and other people agree, then you’ll make more money and have more influ­ence over the directory.

At a later stage, we could allow traders to group together and publish their port­folio. This means you could search for ‘what do New Zealanders think is the best news site’. Or ‘what do web developers think is the best ref­er­ence site’. This isn’t some­thing a search engine can do as easily.

Is this a web 2.0 product, and how so?

My aim in the project is to create a cool fun website that I think it’ll help make finding good sites easier. I guess if you follow the defin­i­tions on

http://twopointouch.com/2006/08/17/10-definitions-of-web-20-and-their-shortcomings/

It has to some degree:
The wisdom of crowds
Shared Web Applications
User Participation

But mainly because I think these elements will help produce the best result.

I try to allow the site to work even if you don’t have javas­cript (we do not rely on AJAX) and provide an AJAX way if I think it can help usability.

I would argue that you should narrow down a little on what you search for, to gain a little traction in a few limited ‘best of’s. What’s your answer?

My view is to provide some value to people as early as possible. You are right. Limiting would show exactly how the product works. But one way we give users value is to allow them to add their own site under keywords. This is fine since if they if they are spamming they will be punished, and if they want to earn lots of money to invest in their site they will need to play the game well and make good invest­ments in other sites. We could publish some of the popular search terms on the front page, or if people want to recom­mend some cat­egories to promote we could do that.

Who are you and where are you from (pro­fes­sion­ally and geographically)?

I’m a software engin­eering student from the bottom of the world (New Zealand). I have been working for an eco­m­merce store which is where I picked up the skills to do this.

How are you funded currently?

Currently the project is exper­i­mental rather then an actual company, it’s self funded.

What are your plans for this going forward?

The real problem is gaining users, the site isn’t useful unless people use it, and people won’t use it unless its useful. This is made harder by the fact we have $0 mar­keting budget.

The key things we are con­cen­trating on at the moment are:
– We are trying to add in content to make the site more useful.
– Getting users.

Share this post:

Possibly related:

• Yahoo! 2.0
• What is Non-​​​​Linear Search?
• Googling for Answers About Web 2.0