The base price will update once a day (on average) but I'm hoping to build something based on facebook likes and twitter tweets that affects it and we can check that a few times a day without doing affecting the endpoint servers
In the study of e.g. prediction markets, the idea is that if the price is too low, lots of people will try to buy it. Then the price gets raised up to compensate. If the price is too high, lots of people will try to sell it, and the price gets lowered to compensate.
What all this means is that the traders don't learn info from you, you learn info from them. You shouldn't have to put any information into the prices. If you have a supply/demand matching algorithm, then the prices will give you the information.
So for example, what you could do is give everyone 10,000 play dollars at the start, and they can buy/sell whatever "sites" they want at the going prices. But every day (week?), there's a payout -- if you own one share of site X, and site X is #n in the rankings, you get a dividend of say 50,000 - n cents.
Then if I think that site X will be in 1000th place, I should be willing to pay up to $490.00 for a share, because that's how much I expect it to pay off for me. Something like that.