Can people reliably distinguish between actual financial markets time series and randomized data? In the February 2010 draft of their paper entitled “Is It Real, or Is It Randomized?: A Financial Turing Test”, Jasmina Hasanhodzic, Andrew Lo and Emanuele Viola report the results of a web-based experiment designed to test the ability of people to distinguish between time series of returns for eight commonly traded financial assets (including stock indexes, a bond index, currencies and commodities, all given names of animals) and randomized data. Using a sample of 8015 guesses from 78 participants over eight contests conducted during 2009, they conclude that:
- There is overwhelming statistical evidence that people quickly learn to distinguish between real and randomized time series.
- The accuracy rate for the 23 of 78 subjects self-identified as finance professionals (73.6%) is statistically indistinguishable from that for other subjects (72.2%).
- Immediate feedback on the correctness of guesses may be crucial to results.
- Results may explain the persistence of the practice of (human) technical analysis.
- Novel interfaces may facilitate the harnessing of human capabilities to process and extract information from data for financial markets (and other complex systems) in ways that computers cannot.
Results also argue that financial markets time series do not follow random walks.
In summary, experimental evidence indicates that people, whether expert in finance or not, can quickly learn to distinguish financial markets time series from randomized data with high reliability via simple inference.
