D'Anconia is my independent study project for Fall 2013. It is intended to be a web application with which one can build graph-based macroeconomic models of an equity's performance, and then use these models to predict the equity's value in specific scenarios.
Max wants to make money trading ExxonMobil (XOM). He researches the oil industry and related industries significantly before he creates a causal model of the macroeconomic effects which influence the stock's price. Armed with knowledge of short-selling and options trading, Max knows it doesn't matter whether XOM goes up or down; he only needs a confident estimate of its percentage gain/loss to make money.
In D'Anconia's Model Builder tab, Max drags out each macroeconomic effect onto an empty directed graph template, connects the nodes and assigns edge weights according to his perceived model.
Then in the Events tab, Max can drag out various macroeconomic events, set their parameters, and immediately see their effect on XOM according to his model. This tab is the primary focus of the project: a laboratory to quickly test different ideas about the future, given a subjective (but ideally well-researched) causal model of the stock's drivers.
- Two panes: Model graph and Events list
- Drag events in Events list to graph; they transform into nodes
- Click and drag to create edges between nodes
- Click edges to set edge weights
- There will probably be a more simplistic "Multiple regression" option as well, which would automatically generate a model from regressing all chosen events on the chosen stock's price
- Three panes: Chosen events, Events list, and Outcomes
- Can include multiple models to show multiple price outcomes
- Click an outcome -> links to corresponding model in Model Builder tab
- "Outcomes" are percentage gains/losses computed by generating all "leaf node" parameters of a model by sampling uniformly at random from past data, and pushing these values through the model
- A simple list of all the user's models
- Click one -> links to corresponding model in Model Builder tab
- This necessitates a login system, and a method of serializing each graph sensibly to generate persistent URLs to graphs
- d3.js - JavaScript library to render nodes and edges of each graph
- Bootstrap - frontend framework which makes overall layout painless
- Ember.js - web framework for convenient templating and DOM updating
- Node.js/Django - backend frameworks (I'll pick one.)