Forex Automaton

During the month of May, the fifth month of live performance, the system kept running on complete autopilot, with no code upgrades or parameter changes. 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 April review.

Changes in the algorithm

No code upgrades or parameter changes took place during the month.

Figures of merit

The figure of merit used, Pearson correlation coefficient between the forecast and real logarithmic returns on a day scale, is a measure of forecasting quality.

By construction, the Pearson correlation coefficient is a quantity bounded between -1 (the forecast move and reality are total opposites) and 1 (the forecast is perfect). A success or lack thereof on every trading day makes a contribution to this quantity.

In order to make a large positive contribution, one needs a coincidence of a large move in a currency pair with a large forecast move in the same direction. Since a hypothetic rational operator of the system will not pursue small forecast moves, understanding this to be a noisy system, a forecast with large magnitude is more likely to result in a successful trade.

By the same property of a product, an impact on the Pearson correlation coefficient of even a large forecast move in a wrong direction can be moderated if the actual price move turned out to be negligible -- again what we want, since in this case the effect of the system's imperfection on the operator's portfolio is likely to be similarly insignificant.

Likewise, an impact of a wrong forecast move of negligible magnitude will be negligible both for the Pearson correlation and for the operator's portfolio since in this case, the operator is likely to ignore the trading idea.

Finally, the worst case is the one when the system predicts a large movement in the currency pair and a large movement does materialize -- but in the opposite direction. In this case, a large negative contribution to Pearson is recorded.

Why use Pearson correlation coefficient instead of running a model portfolio? This independent figure of merit, characterizing the forecasting component regardless of the money management strategy used, allows one to split the complex problem of trading system development into independent tractable pieces, and solve each pieces of the problem separately, eventually combining a winning forecasting system with a winning money management one, having an independent quality assurance process for each.

Performance tables

In the tables, data for the "month" column are taken from the "month" column of the last day's forecast, which reports 24-trading-day running average of the Pearson correlation figure of merit. End of May 2010 data are taken from the May 31 update. End of April 2010 data are taken from the April 30 update.

This month's effect on the performance since inception was, as Table 0.1 indicates, positive for AUD/USD, USD/CHF, USD/JPY, AUD/JPY, EUR/AUD, and EUR/CHF (6 pairs in total) -- their performance since inception improved compared to end of April. These are the same pairs that showed positive Pearson correlation coefficients for the month. It must be noted that the "month" column is for rough orientation only as it gives the average over 24 most recent trading days which may not necessarily be the month. In contrast, the conclusions on the basis of the "since inception" columns are strict. Overall, the mean over markets for the Pearson correlation coefficient for the month was negative but consistent with zero, with a negative change to the since-inception figure.

In all three tables, the precisions of the mean "since inception" improved (uncertainties went down) compared to April. It is remarkable that at the same time, the mean values for low and high went up, while the one for close went down. For day low and high, we again attribute the improvement to the v0.5 upgrade performed in January. A decrease in prediction quality for close is worrisome: since the live launch in late December 2009, February 2010 remains the only month of solid positive performance for close. By the way, recent study where the entire track record (including back-testing, which nominally starts in April 2006) had been split into 5 non-overlapping intervals, demonstrated that the overall positive performance for close was due to essentially one out of five intervals. So one may conjecture that a similar pattern might be expected to be repeated (as if in a fractal manner) over smaller time scales.

The figures of merit for the low and high play the role of a canary in a coal mine, ruling out the possibility of some of the less subtle problems. The continued (indeed, improving) high quality of daily low and high forecasts indicates that the canary is alive.

Usage strategies

Continued lackluster performance of forecasting for daily close makes us switch research attention to strategies relying more heavily on the stably high forecasting quality for daily high and low, and not so heavily on daily close. Any newsworthy developments will be as usual reported in the Forex trading system: are we there yet? section of the site.

System performance in the individual markets

The main event of the month was of course the panic on May 6th. The panic created a very interesting pattern of an oversold market on May 6th-7th, and for some currency pairs (USD/CAD, USD/JPY, EUR/CHF) the system was able to read the pattern correctly and change its stance, literally turning on a dime, and going from a winning bearish attidute directly to a winning bullish one (or from bullish to bearish in case of USD/CAD, which is equivalent). For USD/CHF, the system switched gear in the same manner even earlier, on May 7th.

The following 14 subsections are dedicated to the specific currency pairs, useing a particular method of charting to illustrade performance. Two charts will be presented for each. The first chart will show all trade ideas. This corresponds to trading in every pair every day.

The second chart will highlight the days when forecasts for all three components of a day's candlestick were pointing in the same direction. I call this level zero (L0) requirement or L0 trigger.

Table 0.4 shows little or no effect on the quality of trades, from the L0 trigger this month, as judged by the simple tally of wins and losses. Note that such a simple counting approach ignores the problem of the relative impact of these wins and losses. Therefore it can only serve as an illustration to supplement more quantitative studies.