This USD/JPY back-testing analysis continues the series which began with EUR/USD. Simulated track records of six best Step Two algorithmic traders are studied graphically. For a more numbers-oriented approach to performance, see the article explaining the trading system optimization process which led to the selection of these six robots.
-
Graphical analysis of trading system's simulated track record. Step Two algorithm, USD/JPY.
- Tags:
- frontpage
September 28 2009, 11:00am | Comments »
-
Graphical analysis of simulated track record of Step Two robots with EUR/USD.
Simulated track records of six best Step Two algorithmic traders are studied graphically. For a more numbers-oriented approach to performance, see the article explaining the trading system oprimization process whereby the six robots were selected. This post focuses on performance in the EUR/USD market and will be followed by similar posts dealing with other markets.
- Tags:
- frontpage
September 25 2009, 12:25pm | Comments »
-
Graphical analysis of simulated track record of Step Two robots with EUR/USD.
Simulated track records of six best Step Two algorithmic traders are studied graphically. For a more numbers-oriented approach to performance, see the article explaining the trading system optimization process whereby the six robots were selected. This post focuses on performance in the EUR/USD market and will be followed by similar posts dealing with other markets.
- Tags:
- frontpage
September 25 2009, 12:25pm | Comments »
-
Six best Step Two robots are selected for closer inspection
Having identified the range of parameters where a Step Two trading system has the greatest advantage over Step One (in the course of a Step One vs Step Two performance comparison), I proceed by narrowing down the range of adjustable system parameters even further. I select six "best" algorithmic traders for a more thorough examination, by forming a set of tighter system parameter selections, complemented with cuts on the figures of merit such as skewness of return distribution, the annualized return and the fraction of months with a negative return.
- Tags:
- frontpage
September 23 2009, 12:05pm | Comments »
-
Six best Step Two robots are selected for closer inspection
Having identified the range of parameters where a Step Two trading system has the greatest advantage over Step One (in the course of a Step One vs Step Two performance comparison), I proceed by narrowing down the range of adjustable system parameters even further. I select six "best" algorithmic traders for a more thorough examination, by forming a set of tighter system parameter selections, complemented with cuts on the figures of merit such as skewness of return distribution, the annualized return and the fraction of months with a negative return.
- Tags:
- frontpage
September 23 2009, 12:05pm | Comments »
-
Comparing security market lines for the Step One and Step Two trading systems
I continue digesting results of the recent Step One and Step Two back-testing simulation runs. Today I post the risk vs return curves for the sets of virtual traders considered in the course of the optimization, comparing what might be called (in mainstream finance) security market lines for the two modes of trading system operation. Then I zoom in on the range of trading system parameters where Step Two is seen to provide the greatest advantages.
- Tags:
- frontpage
September 16 2009, 7:12pm | Comments »
-
Trading system optimization: Step One analysis update.
The historical data for the six major forex markets have been re-processed with the upgraded back-testing code in the Step One mode (fully isolated market analysis and fully isolated portfolio management). This Step One report is useful for a side-by-side comparison with Step Two results representing the next step in the algorithmic sophistication, combining multiple markets in the same portfolio. Both reports will serve for future reference, the current goals being 1) to ensure we do make progress in the trading system development from Step One to Step Two, and 2) understand the changes in the trading style that are necessary to realize the extra benefits of Step Two.
- Tags:
- frontpage
September 15 2009, 4:56pm | Comments »
-
Return vs risk in the presence of simultaneous opportunities: Step Two optimization update.
Here is a further step towards optimizing the forex trading system performance in the "semi-isolated" mode. The "isolation" means that each market is being predicted independently of the rest, with no attempt to discover and learn the inter-market patterns. However, the money-management component of the algorithm is aware of the concurrent trading ideas form the different markets and weighs them against each another, selecting the most promising. In this report I focus on risk vs return optimization.
- Tags:
- frontpage
September 11 2009, 5:37pm | Comments »
-
Updates to the trading system optimization code.
The following changes to the optimization code have been recently introduced and are documented here for the record.
- Tags:
- frontpage
September 11 2009, 12:00pm | Comments »
-
LIBOR patterns: the story continues with CME:EM futures
With the remarkable correlation patterns seen in the logarithmic return autocorrelation data for LIBOR, the question of practical importance is to see how much of the pattern survives in the freely traded instruments related to the interest rates. This question will be addressed in several installments. I begin with Eurodollar futures traded on Chicago Merchantile Exchange (CME). This is the first article in the sequel series on the subject.
- Tags:
- frontpage
September 3 2009, 2:49pm | Comments »
-
More Step Three back-testing results. Trade entry threshold range is extended.
In the previous back-testing report featuring the "ultimate" scenario of multi-market analysis coupled with multi-market competitive portfolio management system, I noticed that one of the key optimization variables, the trade-entry threshold, does not reach the optimum value at what used to be the optimum in the single-market analyses. In this report, the range of entry-threshold parameters has been extended to reach the expected optimum, but there is still no hard evidence that the optimum has been capped within the range.
- Tags:
- frontpage
August 17 2009, 4:18pm | Comments »
-
First look at Step Two: single-market pattern recognition, multi-market money management require a separate optimization effort.
This is the first report on the multi-market trading system performance in the "isolated" mode. The "isolation" means that each market is being predicted independently of the rest, with no attempt to discover and learn inter-market patterns such as those we observed and reported. The money-management component of the algorithm is aware of the concurrent trading ideas form the different markets and weighs them against one another, selecting the most promising. Initially seen as a "boring" step (so-called Step Two) towards a smarter algorithm that does take full advantage of those patterns, the isolated analysis also provides a benchmark against which the value of the inter-market analysis will be measured. One might expect that for the same values of the parameters the profitability of the Step Two algorithm would be comparable to that of Step One, while the volatility of the performance (investment risk) would be reduced thanks to portfolio diversification. This preliminary study reveals that, all the algo parameters being the same, the competition between trading ideas boosts the profitability considerably, but also increases the risk. The fact of the "interaction" between trade ideas questions the original assumption that the best money-management parameters for the isolated markets are the best in the multi-market scenario, as already noted in the preliminary Step Three report.
- Tags:
- frontpage
August 13 2009, 6:53pm | Comments »
-
First peek at the simulated trading results in the simultaneous multi-market mode.
The first peek at the simulated trading system performance results, obtained in the multi-market mode of operation, reveals some exciting early results but also highlights the amount of work ahead. These results are obtained with the forecasting algorithm handling the different markets jointly, attempting to learn and use cross-market patterns. The main question is: what do we gain from the combined multi-market analysis? Theoretically, the benefits are to be expected in the area of pattern recognition (where new classes of patterns, those involving different markets, may appear and be detected) as well as in the area of money management (cf. asset diversification along the lines of the modern portfolio theory). We are not ready to address this question, as it turns out that having switched to the combination of the markets, we have landed in the trading regime which is different from that seen for the nominally equivalent parameter values within any single market. The next task will be to sort out the trivial and not-so-trivial differences.
- Tags:
- frontpage
August 3 2009, 9:04pm | Comments »
-
Summary of the trading system optimization results. Step One.
I've gone through three rounds of historical data optimization (back-testing) for the major forex pairs, performing simulated trading in each pair independently. This mode of the market analysis, with the various exchange rates being treated by the algorithm in isolation from one another, is not the way the production trading system will operate. However, understanding the markets in isolation and optimizing the trading system in this simpler problem setting is seen as the first step towards optimization of the more complex algorithm, where the amount of information at every point in time will be radically increased by combining analyses of multiple exchange rates within the same algorithm. This study is a cross-check to see how the indvidual optima hang together, trying to define a coherent dry residue from the combined experience of the 150,423 simulated trading histories.
- Tags:
- frontpage
July 21 2009, 5:28pm | Comments »
-
Summary of the trading system optimization results. Step One.
I've gone through three rounds of historical data optimization (back-testing) for the major forex pairs, performing simulated trading in each pair independently. This mode of the market analysis, with the various exchange rates being treated by the algorithm in isolation from one another, is not the way the production trading system will operate. However, understanding the markets in isolation and optimizing the trading system in this simpler problem setting is seen as the first step towards optimization of the more complex algorithm, where the amount of information at every point in time will be radically increased by combining analyses of multiple exchange rates within the same algorithm. This study is a cross-check to see how the indvidual optima hang together, trying to define a coherent dry residue from the combined experience of the 150,423 simulated trading histories.
- Tags:
- frontpage
July 21 2009, 5:28pm | Comments »
-
Forex trading system optimization status
I've gone through three rounds of the trading system optimization studies with simulated trading in the individual forex markets (major exchange rates) and I am working on a synthesis report. After that, the plan is to switch to the optimization of the multi-market version of the system. This is likely to proceed in two distinct steps: first, the money management part becomes aware of the concurrent trade ideas in the different markets and learns to treat them properly. The thing to check is the reduced "risk" or fluctuation in returns. Second, the forecasting algorithm begins to handle the different markets jointly, effectively learning to utilize cross-market patterns. PR: wait... I: wait... L: wait... LD: wait... I: wait...wait... Rank: wait... Traffic: wait... Price: wait... C: wait...
- Tags:
- frontpage
July 14 2009, 7:23pm | Comments »
-
EUR/USD optimization: the old optimum looks good enough
This study completes the third round of optimizations done with individual (isolated) time series for the major forex exchange rates. What could be a better trading system parameter optimum for EUR/USD turned out to be a disappointment. While investigating the reason, I discovered a new high risk, high return area in the parameter space but ignored it.
- Tags:
- frontpage
July 6 2009, 5:02pm | Comments »
-
USD/JPY optimization: the old optimum looks good enough
I am continuing another round of optimizations, revisiting the six popular currency pairs one after the other, extending the range of the trade-entry parameter further into the "conservative" area. In the case of USD/JPY however, the "old" optimum seems good enough so that no improvement is brought about: even though the "new" returns, the "new" risks do not look justified. The "landscape" of the optimization problem here looks very clean and understandable.
- Tags:
- frontpage
June 30 2009, 5:51pm | Comments »
-
With trade entry parameter range extended, GBP/USD optimum moves closer to the rest
With the trade-entry parameter range extended into the more conservative area, the GBP/USD optimization results begin to look more like those of other currency pairs, with arguably a better optimum.
- Tags:
- frontpage
June 29 2009, 12:40pm | Comments »
-
Optimizing the forex trading system: USD/CHF revisited
This study covers an extended range of the trade-entry threshold parameter, the one that controls the "patience" of a trading system or the amount of "excitement" about a trade idea needed to enter the trade. With the trade-entry parameter extended into the more conservative area, the USD/CHF optimization results begin to look more like those of other currency pairs.
- Tags:
- frontpage
June 24 2009, 5:03pm | Comments »