Sunday, February 26, 2012

Crowdsourced stock market trading

Stock market trading has become a dirty word, or if not that, at least uninteresting. Wall Street excesses and the 2008 crash have led to little recent opportunity for financial return (non-existent interest rates for saving, and flat stock markets for equities (the S&P 500 return in 2011 was 0% (S&P 1257 at 12/31/10, 1258 at 12/31/11). Gold has been one of the only asset classes to realize real return (142% five-year return, $632 as of 12/28/06, $1531 as of 12/29/11). The particular subjective day trader gave way to faceless high-frequency computer algorithms as one of the only means of squeezing profits out of the stock market.

One thing that could turn this around, and have the dual benefit of bringing more transparency to markets and market practices is crowdsourcing. The enormous amounts of clean, freely available, computable, straightforward-to-understand data without privacy issues are ideal for crowdsourced manipulation.

Earlier attempts at applying crowdsourcing to stock market trading (for example, Yahoo Prediction Markets with leaderboard-style tracking of traders’ mock portfolios) fell by the wayside with the 2008 crash, but the concept could be reincarnated. There are several obvious ways to deploy crowdsourcing in stock market trading startups:

  1. First would be a direct implementation of crowdsourcing as from the Wikinomics,, eteRNA model: making usable web-based datasets available to the wisdom-of-crowds to apply diverse ideas from different disciplines, often resulting in better results than those produced by the ‘experts’ in any field. Leaderboards, competition, leveling-up, forums, badges, and other gamefication techniques would be expected.
  2. Second would be a platform where real-life traders can open source their trades, either before or after execution. Interested traders would grant open access to their trade logs, inviting crowd review to find winning trades, strategies, and traders, and conduct meta-analyses like what strategies work well in a high-volatility environment, a down economy, etc.
  3. Third would be prediction markets 2.0, a more social gamefication implementation of prediction markets for stock trading, sales forecasting, movie hit projections, elections, and flu outbreaks through platforms like Iowa Electronic Markets, Intrade, etc.

Sunday, February 19, 2012

Black Swan thinking – there’s an app for that!

As mobile apps increasingly mediate human interaction with the outside world, possibly eventually becoming a full buffer layer, there should be an app for Black Swan thinking, or more broadly, for bias reduction.

A Black Swan is an event that is rare, has extreme impact, and is retrospectively (but not prospectively) predictive. As humans with story-based not statistics-based evolutionary-relic perceptual systems, we should think more black swannishly or at least have mechanisms for minimizing exposure to downside black swans (e.g.; stock market crashes, terrorist attacks, health situations), and maximizing exposure to upside black swans (e.g.; startup investments, knowledge, parties).

Antibias App: a decision-making tool based on personal bias
The Antibias App, an on-board bias reduction coach (an extension to the Siri 2.0 personal virtual coach), could improve human perception by allowing randomness to be seen, statistics-based thinking, and a focus on the unknown (antiknowledge) as opposed to the known.

The Antibias App could list the top 5-10 bias areas (e.g.; confirmation bias, decision-making, belief, and behavioral biases, social biases, and memory errors and biases) with your personalized score for each one and a composite score as applied to different contexts (e.g.; personal, professional, political, economic). Even determining personalized biases is valuable; this could be accomplished through automated data collection, sentiment analysis of social media droppings, and online tests.

An advanced feature of the Antibas App could be a click-through to see the top three pro/con arguments on any issue and where different composite bias scores lie (e.g.; your own, your social network, your professional peers, your neighborhood, your nation state, etc.).

The Antibias App could be viewed in different modes such as story mode, statistics mode, graphics mode, and data visualization mode. The meta goal of the Antibias App is to increase liberty and choice by opening up more ways of thinking about bias and improved action-taking as a result.

Sunday, February 12, 2012

Detroit 2.0 – cultural transformation

Certainly cultural transformation occurs but to what degree can it be actively catalyzed?

Creative class cities bloom and their opposites become walking Detroits.

How to revitalize your city into Rochester 2.0, Detroit 2.0:

  • Free houses for artist communities (the aesthetic future starts now); stimulatory homesteading initiatives
  • The post-ecotourism fad: ghetto tours; hip hop music and dance classes
  • Favorable tax policies and free trade zones like Paul Romer’s Charter Cities program (example: Hong Kong in Honduras)
  • Singapore/Korea-like targeted industrial policy (Welcome Stem Cell Research!)
  • Social policy liberalization: immigration amnesty, gay marriage, euthanasia, decriminalized marijuana use

Sunday, February 05, 2012

The big data era's flux and pulse

Big data is an important contemporary trend but what does it actually mean?

What is big data?
Big data refers not just to the absolute size of a body of information (which currently can be on the order of terabytes, petabytes, and exabytes), but its usability and manageability. Some of the defining parameters of big data are its large size, high velocity activity (incoming, processing, outgoing), heterogeneous nature (a variety of structured and unstructured data types like video and images), and requirement for real-time analytics.

What is the process of working with big data?
The process of working with big data involves several steps. First there may be an exploration of the data using tools for classification, visualization, and summarization. Then there is the detailed step of data cleaning to make the data consistent and usable. The next step is data reduction, for example defining and extracting attributes, decreasing the dimensions of data, representing the problems to be solved, summarizing the data, and selecting portions of the data for analysis. Then, the steps of predictive analytics, scoring, reporting, publishing, and quality validation and maintenance can be applied.

What are the applications of big data analysis?
Some of the benefits of big data analysis are the ability to summarize information, make predictions, identify trends (for example, consumer spending patterns), and rank and prioritize information. Some of the specific algorithms employed include for summarizing: clustering and associations; for making predictions: tree-based methods, neural networks, and k-nearest neighbors; for identification: anomaly detection, similarities and matches, and change detection; and for ranking: logistics and frequency detection.

Excerpted from an Association for Computing Machinery (ACM) talk on Big Data & Predictive Analytics (slides).