Welcome Module 1

A first look at microprediction

Collective Repeated Prediction Powering AI

Markets have crowdsourced probabilistic intelligence for over a hundred years, since the construction of the Julius Totalizator machine in 1913. Now, unencumbered by technology constraints of the past, and with many more algorithmic possibilities for combining intelligence beyond adding things up (Julius' only option) we're going to try to take competitive information aggregation to an entirely different plane.

This site is, we hope, the beginning of something analogous to the web itself. It is something you can participate and benefit from now. It is something that might be transformative later, with your help.  

Imagine awaking one day to a world in which all manner of real-time intelligence is at your fingertips. This intelligence is a view into the near future, with a horizon of minutes or hours. It is not a dashboard for a surveillance state, nor a mere catalog of things that interest many people (such as when the train will arrive, or which NFL players will be injured). Rather, it is yours to control. It includes a myriad of numbers of interest mostly to you personally, your organization and your life.

In this world you have access to a surprisingly cost effective means of predicting anything - so long as you can instrument it, and so long as the prediction task is a repeated one.



As you can see, a prediction network is not organized the same way most companies are, or most teams of data scientists. Rather, it is sewn together in the loosest possible way, powered ultimately by the action of the invisible hand of market forces - but a tiny hand, largely unencumbered by economic frictions as we are accustomed to them. This new form of information trade occurs between statistical algorithms as they play repeated games.

This video mentioned the Presidential debate analysis (Tears of Joy)



Business optimization isn't just for the largest firms who can afford teams of data scientists. A prediction network can benefit everyone when data, models and feature spaces are shared and reused.

Self-navigating algorithms

If we make it  little easier for algorithms to find their own way to problems, and solve them, we can dramatically lower the cost of bespoke data science for organizations of all sizes. 

If you'd like to create one right now, hop over to the crawling  instructions and let your creation loose on the world. 

Collective turnkey prediction

In the world we are trying to create, repeated short term prediction is free and instantly available.  See the ten minute data science project for an illustration. 

If you'd like to try it out, feel free to get started with publishing, or head to the introductory Python modules. 


In the next module we reason as to what a microprediction oracle must be