Forex Automaton

Our primary goal is to create a public information service providing financial markets forecasts, based on our proprietary forecasting tools: an automated trading system -- a Forex Automaton™. A live forecasting system is online since late December 2009. Our secondary goal is to quantify and monitor the very existence of sustainable opportunities for arbitrage profit-making. Or simply put, to monitor the degree to which these markets are more predictable than a "fair game" -- to a trader without access to insider information.

The Forex Automaton project was launched in April 2008 with the ambitious mission of leveraging certain algorithmic know-how to create a trading signal service useful to institutional and retail forex traders. This report highlights the noteworthy new developments that took place during the project's second year of life, from April 2009 through March 2010. The intent is to help the reader navigate what is becoming a rather complex network of research topics, concepts and results, by providing an overarching logical framework. Links to complete stories are provided.

During the month of June, the sixth month of live performance, the system went through a major upgrade. The upgrade is expected to further improve prediction quality for daily high and low (as measured by Pearson correlation coefficients between predicted and actual logarithmic returns) while its effect on the prediction quality for daily close is uncertain. This document consists of a summary section followed by 14 subsections, dedicated to the individual exchange rates tracked by the system. Those contain color-coded charts of the performance and details pertinent to the specific currency pairs. For comparison with the previous month, you may want to take a look at the May review.

This third article in a series dedicated to patterns and stability of forex spread deals with EUR/JPY. As in other studies, the data come from a popular non-ECN broker. The only surprise here is the unexpected increse of the spread in 2009, both on absolute scale and in relation to the EUR/JPY quote.

CHF/JPY resembles AUD/JPY in that its temporal distributions for high and low are heavily dominated by the early hours of the day (Central European time), reflecting morning activity in Japan.

For the first time, I address the question of how stable the optimization results are in time. While predictabilities of daily high and low show a highly stable pattern of dependence on the parameter subject to optimization, the positive results for close are mainly due to the high impact of a single period, which happens to cover the financial panic of the last quarter of 2008.

The main conclusion of the previous article was that a strategy with a position size distributed proportionally to the Kelly Criterion was found to be more attractive than the strategy were potential stop-loss would be distributed according to Kelly. A way to implement such a better strategy was understood to consist in fixing the stop-loss distance while having the position size proportional to Kelly allocation. For that, one would need to optimize the stop-loss distance. Here comes the promised development, improving the histogramming technique used to judge the "attractiveness" of a strategy, and elaborating on the choice of the stop-loss.

I compare four different position sizing strategies to use in algorithmic trading with Danica forecasting system. Two of them incorporate Kelly Criterion information. It seems that the strategy I have been using so far has been a sub-optimal one; a way to proceed is discussed.

Currently there is little doubt that the next upgrade of our trading system will introduce some form of trade idea selection into the system's output. This article reports our recent progress towards implementing Kelly Criterion for capital allocation to trades. This is our first quantitative attempt to see what Kelly does to the expected profit per trade and the shape of its distribution.

The "Kelly Criterion" in quant folklore is based on the exposition and development of Kelly's work done by Edward Thorp. Acknowledging Thorp's contribution, I find the original article by Kelly conceptually more relevant to the realities of algorithmic trading as developed by ForexAutomaton. Our forecasting algorithm, as any forecasting tool, can be very naturally considered a case of the hypothetic communication channel discussed by Kelly, and the related mathematical objects, such as joint and conditional probability density distributions for the communicated "symbols" (forecasts and real quotes), are being already monitored here, despite the fact that the Kelly Criterion for capital allocation to trades remains to be coded.

I continue developing ways of using Danica's output for profitable trading. The strategy of combining protective stop with a profit target, setting these at the extremes of the previous trading day and trading in the direction of the forecast (referred to as the Day Range strategy and reported in the previous post is modified to remove the profit target. Much better results are obtained.

In the January performance review for the Danica trading system, an idea of a day-range trading strategy capitalizing on the high quality forecasts for the direction of daily high and low was expressed and a set of what-if charts for the forex pairs tracked was provided and discussed. This post is a deeper and more quantitative discussion of the strategy. Which forecasts should be acted upon? What is the expected profit per trade? How are profits/losses distributed statistically? -- these are the questions addressed.

The intention of this post is to tie together several topics which appeared on my radar screen in the course of the trading system optimization. First, it has been understandably hard to fully rid oneself of  vestiges of the mainstream financial theory based on the postulate of market efficiency, while building a wealth-generating tool relying explicitly on demonstrable market inefficiencies. The realization that Sharpe ratio does not let one make an objective choice of a portfolio was  there from the beginning, and I recall perceiving this fact as a "necessary evil". Then came the understanding of the fact that an  artithmetic average of returns gives one a biased picture of long-term return, and consequently, Sharpe ratio is built around biased quantities.

In the recent forex/CFD data, USD/CAD is negatively correlated with light oil (WTI) CFD. This is the same as saying that CAD, one of the commodity currencies, is positively correlated with oil. This is old news. In this article I take a deeper look at the issue and analyze the shape of the correlation peak. Analyzed on the hour time scale, the correlation peak is broad and somewhat asymmetric, indicating that it is much safer to rely on the guidance of USD/CAD in predicting the oil price, rather than other way round. The necessary caveat is that this is a time-integrated picture, covering a period from late August 2008.

I've outlined the original motivation to study historical LIBOR data from predictability point of view in the USD LIBOR article. I continue with the logarithmic returns technique that proved useful in forex. Like the previous reports, this document begins with historical LIBOR charts for the Swiss Franc, continues with volatility analysis, and culminates with autocorrelations and correlations. You will see that predictable patterns in CHF LIBORs vary with duration term. Autocorrelations of short-term LIBORs show fast (about 4-day period) oscillation. For 3-month and 6-month terms, the main correlation pattern does not develop 70-day period waves on top of positive background, in contrast to USD and EUR LIBORs, but keeps oscillating between positive and negative autocorrelation values, with the oscillation period longer than that of the shorter terms. The autocorrelation of 12-month LIBOR remains similar to 6-month instead of becoming more uniformly positive as it does for JPY or more jittery as it does for USD, EUR and GBP.