News for hackers. Stuff that matters.
Re:Ask HN: Review my startup, Beansight.com by derrida on Wednesday April 20, 2011 @10:27AM (#2466710)
Just a thought: Whether a prediction is objectively good or not is different to whether the prediction is seen as good or not. If you can think of a mechanism to figure out whether a prediction was good or not after the event has happened, the system could then make a second order prediction about whether that user's a good predictor or not. Keep me up to date if you manage to solve any of these problems.
Re:Ask HN: Review my startup, Beansight.com by Oyinko on Wednesday April 20, 2011 @10:43AM (#2466773)
We worked on an algorithm that will check every prediction to know if they happened or not.

If members record a prediction on Beansight and that it happens, they will be rewarded. If they record predictions that will not happen, they will lose credibility points.

Re:Ask HN: Review my startup, Beansight.com by derrida on Wednesday April 20, 2011 @11:10AM (#2466905)
How do you check if it has happened or not? Is this based on peoples votes after the predicted time of the event? What if someone wants to make a prediction that involves when something occurs?
Re:Ask HN: Review my startup, Beansight.com by geeyom on Wednesday April 20, 2011 @11:41AM (#2467043)
Hello Derrida,

The algorithm to check whether the prediction occurred is based on each single vote. It weights more people with a great expertise score and weights more votes close to the deadline or votes that passed the deadline.

If the prediction comes to an end before the deadline, then the algorithm will wait to reach the deadline to validate it. But we will soon let the community close a prediction if they decide it's over.

Guillaume