wciszak

You have made two interesting points. 1. You have observed that the difference between a profitable candlestick system and an unprofitable one is trade management. If so, then could it be that it is the trade management and not the candlestick pattern that contains the money making logic? After all, it can be shown that it is possible to produce profitable trading results from a random entry signal coupled with a logical exit signal that has its own predictive value. In other words, the entire predictive algorithm is in the exit part of the system. Could the candlestick pattern be the proverbial stone in the stone soup? 2. With your race driver analogy you have argued that trading success lies not in the method itself but in its skillful execution. It is true that some gifted people have exceptional instincts that allow them to quickly make the right decisions in confusing situations that would confound others (I am not one of them and, if you are, then my hat is off to you with a deep bow). That is why I prefer not to rely on my own fickle judgement but to have well defined and reliable rules instead.

I appreciate your reply. I use the following method in testing trading signals. The entry signal would be a single criterion (such as a particular pattern). The exit is after a fixed number of bars, with no stops. If the results are breakeven after a large number tests, this indicates that the pattern has no predictive value. If a group of test results shows a bias to one side or the other (indicating a possible predictive value) I would then compare these results to results obtained with a random entry signal. (Even random trading signals tested on random data can produce profitable subsets - clearly created by chance). The "real" signal not only has to show bias but this bias has to be significantly stronger than the best results produced by random signals, This is similar to testing drugs in the pharmaceutical industry. The candidate drug not only has to work but has to work better than a placebo, The idea that candlestick patterns must be tested together with other patterns (such as gaps) also resembles some claims in the drug industry. For example, a common remedy for joint pain contains glucosamine and MSM. Extensive studies were undertaken which showed both ingredients to be useless. The manufacturers hit back with the claim that the tests were not valid because the ingredients were tested separately and not together. Perhaps, if they had been tested together, with the same results, the objection might have been that the pill should have been taken 5 times a day and not 3 times a day, as tested. Clearly, there is no end to this and what it amounts to is overoptimization. For more views on this, you might want to want to take a look at my article on optimization.

The patterns described here seem to work only on well chosen charts. I have tested numerous candlestick patterns on 10 years of data in various markets and have not found a single pattern that performs better than a random signal. If you have actually tested any of these patterns and found them effective, please publish the results and prove me wrong.

Interesting thread but I see a lot of references to insights, beliefs and opinions about TA. What about actual testing? It seems to me that if you discover a valid technical pattern (one that repeats and is predictive) than it MUST work, by definition. If things are not working out for you, perhaps you are relying on false, untested patterns.

Technical analysis is simply a toolbox. Some tools are more useful than others, all need to be used properly. What we use the tools for is to build something. What we build with technical analysis is predictive signals. We cannot blame tools for building a leaky roof and we cannot blame technical analysis for building a trading system that loses money.

I understand your view, shared my many, that using stops seems to save money, especially if the market keeps going further into adverse territory after you were stopped out. However, if you use stops routinely (money-managements stops and trailing stops) you do place yourself at a disadvantage because you will force yourself to always give up a chunk of your equity before every exit. Testing any valid system first without stops (only using some logical exit signals) and then with stops will show that using stops impairs system performance. A good discussion of this is found at Does using stop loss orders improve trading results? | Skeptical Trading Strategies Regards