The First Annual Summary of Forex Automaton Research Progress, April 2009 - Forex bipolarity

User Rating: / 7
PoorBest 
Written by Forex Automaton   
Friday, 03 April 2009 13:22
Article Index
The First Annual Summary of Forex Automaton Research Progress, April 2009
Forex bipolarity
Leading indicators
Periodicity or oscillation
What did the crisis change?
Black-box algorithm paper-trading
Conclusions. Future.
All Pages

Pattern 1: "bipolar disorder"

The initial study of autocorrelations and inter-market correlations in forex was done within the data sample of the same fixed time span, from 00:00 2002-08-20 to 00:00 2008-02-01 (New York time). There is no fiddling with the choice of time coverage for the different sub-samples -- these are apple-to-apple comparisons.

AUD/JPY bipolar disorder autocorrelation AUD/USD bipolar disorder autocorrelation GBP/JPY bipolar disorder autocorrelation USD/CAD bipolar disorder autocorrelation
CHF/JPY bipolar disorder autocorrelation EUR/AUD bipolar disorder autocorrelation EUR/CHF bipolar disorder autocorrelation EUR/GBP bipolar disorder autocorrelation

Fig. 1.1. The most stunning predictive autocorrelations on the hour scale among the forex exchange rates. Click on the panel to get to the article reporting the measurement. Top row, left to right: AUD/JPY, AUD/USD, GBP/JPY, USD/CAD. Bottom row, left to right: CHF/JPY, EUR/AUD, EUR/CHF, EUR/GBP. The noise (red in the figure) is obtained from martingale simulations based on the historical volatilities of the forex rate in question for the period under study. The noise is presented as mean plus-minus 1 RMS, where the RMS characterizes distribution of the correlation value obtained for this particular time lag bin by analyzing 20 independent simulated pairs of uncorrelated time series. The RMS is a measure of accuracy in the determination of the correlation values, and is dependent on the amount of data and the time scale.

The so-called "bipolar disorder" is a tendency to form quickly alternating rises and falls, more pronounced than in a fully unpredictable time series of the same volatility, shows up as negative deeps surrounding the zero-time lag peak. The expression "bipolar disorder" follows B.Graham who used to attribute to the stock market the behavioral features of a metaphoric manic-depressive patient, Mr Market. The manic-depressive successions of rises and falls in the price time series, whereby rises seem to trigger the falls and vice versa, creates the negative auto-correlation feature in the corresponding series of logarithmic returns -- recall that the product of two returns with the opposite sign is negative. Being next to each other on the chosen time scale, they create the negative deep at the unit lag. The feature often persists on more than one time scale (see AUD/JPY analysis as an example).

The statistical concept of bipolar disorder is the exact opposite to the concept of a "trend", the latter being a positively self-correlated sequence of price movements.

AUD/JPY 1-hour delayed autocorrelation history
AUD/USD 1-hour delayed autocorrelation history
CHF/JPY 1-hour delayed autocorrelation history

Fig. 1.2. History of the 1-hour lagged autocorrelation magnitude in some markets with "bipolar disorder" effect. Click on the panel to get to the article reporting the measurement. Time axis is labeled in MM-YY format. The number of bins (24) equals the number of quarters in the six-year period. The noise (red in the figure) is obtained from martingale simulations based on the historical volatilities of the forex rate in question for the period under study. The noise is presented as mean plus-minus 1 RMS, where the RMS characterizes distribution of the correlation value obtained for this particular time lag bin by analyzing 20 independent simulated pairs of uncorrelated time series. The RMS is a measure of accuracy in the determination of the correlation values, an uncertainty dependent on the amount of data and the time scale. From top to bottom: AUD/JPY, AUD/USD, CHF/JPY.

To be of practical value for forecasting and algorithmic trading, these features have to be expectable in the future, at least in principle. A rare but huge event and a frequent one of a moderate magnitude may leave the same trace on the autocorrelation. At the very least, one must ensure that these time-averaged signals are not merely diluted residues of certain once-in-a-lifetime events. If they are merely that, a trading system with regular decision-making and execution, such as the one being built here on the Forex Automaton site, is not the optimal strategy to take advantage of them. Time histories of the effect shown in Fig.1.2, demonstrate that we are not  dealing merely with consequences of just a few rare events in the time-integrated autocorrelations, Fig.1.1. The inefficiency in question ("bipolar disorder") is historically continuous.



Last Updated ( Thursday, 31 March 2011 10:06 )