I wanted to share some insights I gained as I developed my technical analysis system which is featured on the TradingHelpDesk forum. I started my journey in Technical Analysis about 5 years ago when I read my first book on Candlestick Charts. It all seemed so easy. Find a Hammer at the bottom of a downtrend and the stock is going to go up. Better still, find a Bullish Engulfing pattern at the bottom of a downtrend and you’re on to a winner. Even better still, find a Bullish Engulfing pattern at the bottom of a downtrend, at a time where the 4 day moving average has just crossed the 9 day moving average (exponential of course) and you can write your own pay check. Well that was the theory, but let’s face, if it was as simple as that then anyone who read a candlestick book would be a millionaire and all the candlestick book authors would also be rich beyond their wildest dreams!
Now please understand I’m not knocking Candle Charts or Candle Chart practitioners, but what I will say is as you delve deeper into the subject you find more and more convoluted recipes for deciding when a stock is going to go up. Reading Candle Charts starts to fall into the realms of art rather than science. Dedicated practice in looking at thousands and thousands of charts definitely gives you an edge, but it's very difficult to pin down the exact process a good Candle Chart technician goes through in successfully reading the chart. One thing’s for sure, it’s not a simple recipe that you can write down in a list of five instructions. I personally find the challenge of translating the ‘art’ into a computer algorithm to be very interesting and stimulating.
For me the first question to answer was quite simply this: “Is there any information in the stock chart that allows us to predict its future move?” There are, of course, people out there who think Technical Analysis is complete nonsense and Warren Buffett is one of them [see “Talent Is Overrated” by Geoff Colvin]. Now, who wants to argue with Warren? I therefore set myself the task of answering this fundamental question. At one extreme the avid supporters believe you can make long term predictions and at the other extreme there are those who believe that previous price action has no bearing on the future price. They believe that trying to predict future stock direction is as futile as trying to predict the trajectory of a gas molecule in a bell jar!
In order to make a stab at answering this question, I designed an experiment that gave some very interesting insights. Testing over 1 year of data I set up a simulation that would look at every stock on the NASDAQ and only buy when the stochastics were in the oversold region AND where there was a strong Bullish Engulfing pattern. I arranged for my simulation to sell the stock within 15 days when the profit was at its highest. I compared this with a similar simulation where I bought the stocks at random points over the year and also sold within 15 days when the profit was at its highest. In reality you wouldn’t know when the profit was at its highest but for the purpose of the experiment I wanted to see if buying on a Bullish Engulfing pattern and selling at the maximum profit within 15 days gives a better profit than buying at random and again selling at the maximum within 15 days.
The results were very interesting. I found that irrespective of buying, Bullish Engulfing or Random you get roughly the same profit. The main conclusion that I could draw from this is that the Bullish Engulfing pattern, in conjunction with oversold Stochastics doesn’t give you a better performance than simply buying at random.
Does this spell the death of technical analysis? Well for me it was just the beginning. I decided to forget all perceived wisdom and start again. The system I’ve now developed doesn’t use candle charts techniques (developed 400 years ago) or conventional Western technical analysis. Instead, it takes advantage of 21st Century Technology and uses a modern computer and sophisticated statistical modelling techniques to identify stock candidates that are more likely to move up than down. The Pawlewski System does of course make mistakes but on average over a year, 75% of its trades are profitable, each making at least 1% profit and usually more. The system finds several trading candidates over each month and has been shown to make consistent profits over the last 15 years of test data.
A common question I get asked is about the dangers of inadvertently fitting the algorithm to the data until you get a good result, then mistakenly expecting this ‘over optimised’ system to perform just as profitably in the future. Well I am very aware of this type of problem and have designed the system not to over specialize on past data.
I believe, and have shown, that technical analysis does work in the short term and is ideal for swing trading. I’m not so sure about long term timeframes but I have an open mind. I also believe there is a lot of hype around the subject and I haven’t seen many of the claims backed up by hard evidence.
Marek Pawlewski http://www.tradinghelpdesk.com