DCG Financial Blog

Since the previous post I have been running the algorithm with a longer matched pattern length, a 50 pip trailing stop and a 100 pip stop loss, instead of the 250 pip stop loss used before. The extended pattern length is supposed to reduce the number of trades and lay a filter over the trading to remove what might be a loose method of trading. There are specific reasons why this didn’t change the performance of the algorithm very much, and they have to do with the dynamic assessment of what is the most relevant periodic at any given point in time. MEAT looks back over a specific period of time and makes a few iterative comparisons to determine what length of time is appropriate for pattern matching. What we have seen in these results hints that lengthening the period of time used as the maximum length in this specific “matching length” period of time does not change the length of time used as a relevant length of time for this aspect of the algorithm. This means two things. This time relevance part of the algorithm needs to be rebuilt or relevant patterns happen to only exist very near to each other. This means that most of the trades are happening using a six hour pattern, even when the algorithm is given the option to find patterns of longer lengths. This means that pattern matching using the relatively short periods of time I am using for analysis is subject to some inherent limitations which may be due to the fact that similar patterns happen to happen near to each other in time.
This is not a science or a proven business practice. Programming trading systems and testing them on historical data for the purpose of trader training is an art.

http://www.dcgfinancial.com/GBPMEATex.htm

Hypothetical or simulated performance results have certain inherent limitations. Unlike an actual performance record, simulated results do not represent actual trading. Also, since the trades have not actually been executed, the results may have under- or over-compensated for the impact, if any, of certain market factors, such as lack of liquidity. Simulated trading programs in general are also subject to the fact that they are designed with the benefit of hindsight. No representation is being made that any account will or is likely to achieve profits or losses similar to those shown.

I have been using the MEAT algorithm in a more limited form to examine its performance in technical analysis. The problem with using MEAT as I have in the past to study how one market effects another is that the process of determining which one market will in fact change another is more easily accomplished through reading the daily financial news. The fact that MEAT offers specific points in time where a relationship is identified is coincidental or accidental if there is not a real world explanation for the short term relationship. The MEAT market cross talk studies can be used to look back over a specific period of time and illustrate areas in time when money flowed from one market to another. Unlike Matter, money can be created from nothing. Just because we see money leave one market and enter another doesn’t mean that the same money that left one market is now the money that has entered the other. Technical analysis although limited in its usefulness to a well read and attentive trader must be mastered if we want to have a fully equipped method to trade. All of the MEAT tests seen below are backed by an autocorrelation or technical analysis version of the MEAT algorithm as a filter. This means that the MEAT algorithm is used to identify where the Pound has done something on one bar and done something else on the next, the two things it does happen to have repeated in the window of time analyzed “Cases” number of times. This is part of the reason why all of the below results are so similar. The technical analysis recommendation is filtered through the cross market analysis to produce the conditional positive that triggers a trade.
In this example we use a very large account and trade whenever a signal comes from MEAT running on short term six to 18 hour technical patterns. This results in a large number of trades and extreme success during periods of time when the market is trending. Although the results are not very positive this example illustrates the robust nature of MEAT. This test took roughly 60 minutes of CPU time on my Athlon 64 workstation. If I were to run this test several hundred times to optimize the performance of the algorithm by trying various tuning values my computer would be locked up for several weeks in testing. As a result of the cost prohibited nature of CPU power I make the decision to constantly work to improve the concept and adaptability of the algorithm instead of identifying optimal tuning conditions for a suboptimal concept.

Hypothetical or simulated performance results have certain inherent limitations. Unlike an actual performance record, simulated results do not represent actual trading. Also, since the trades have not actually been executed, the results may have under- or over-compensated for the impact, if any, of certain market factors, such as lack of liquidity. Simulated trading programs in general are also subject to the fact that they are designed with the benefit of hindsight. No representation is being made that any account will or is likely to achieve profits or losses similar to those shown.

http://www.dcgfinancial.com/GBPMEATtech.htm