twopointouch

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.