I've decided to show some results of popular methods, and simple systems using Metastock over the Dow, excluding dividends, over the years 1857-2001, starting with a $10,000 value, and adding $15 expenses for each transaction.Buy & Hold Dow 26528.52%Trend following methodsParabolic (.02 Step,.20 Max step) 113.56%Parabolic (.02 Step, .01 Max Step) 22133.56%50 DMA cross 28130.20%close out of 20 day price channel 5125.43%close out of 40 day price channel 3465.31%high above 40 day price channel high, orlow below 40 day price channel low 1473.21%high above 20 day price channel high, orlow below 20 day price channel low 923.7714 day+DI above -DI and ADX for longs,and 14 day -DI above +DI and ADX for shorts 79514%Same as above, but not reversal system, exits on +/-DI crossover in opposite direction 88643.5%3 consecutive closes above/below 65 DMA 76124.60%10 DMA and 80 DMA rising/falling over prev. 3 days 17830.29Overbought/Oversold MethodsBuy when close has been below bottom Bollinger Band, then rises above it, sells when hits top band. Shorts are opposite: Bankrupt by 1910Excluding commissions; -97%CCI crossover -100/+100: Bankrupt by 1910Excluding commissions; -14.36%Stochastics, 14 day w/3 day smoothing( %K) and 3 day smoothing of %K (%D)Buy when %K crosses over %D, when %K is above 70Short when %D crosses below %K and %D is below 30Exit when stochastic cross in opposite direction of long;Bankrupt by 1921Excluding commissions; 2836.52%14 day RSI above 70 for sell, RSI below 30 for buy; Bankrupt by 1920Excluding commissions; -100%Adding RSI overbought/oversold levels to the Bollinger Band method described above doesn't wipe you out until 1929. Probably just because it tends to lessen the trade frequency of inevitable losses.Most of the time, I am working on futures systems, and I do not consider volume in such systems. I have been working on stocks and realize the importance of volume, I have mostly been working with OBV, but have not really found a system incorporating OBV in a beneficial manner. One of the problems with OBV is that a stock could be trading heavy on a flat base, suggesting distribution, but the OBV continues up. Conversely, the stock could be at the bottom of a base, trading down slightly on low volume, but OBV will continue to set lower lows. I think these are too defective traits of OBV to overcome going on the direction or level of OBV. Also you'll note in the trading systems above that trend trading indicators outperformed oscillator/overbought/oversold indicators, for the most part. Several of the former literally went bankrupt (lost the ability to trade due to inability to afford commissions, or deficit) by trading, and even with out including commissions, these methods didn't stack up. For other examples you could look at the book, Ultimate Trading Guide, I believe the title is correct, by John Hill, who is author of Futures Truth, a trading systems tester. He reached the same conclusions, albeit testing certain overbought/oversold indicators on futures. These oscillators do not make money, although Bollinger Bands can be bought when futures go above the top band, or shorted when they go below the bottom band. I have tested this but found the profit was not worthwhile. Losses are inevitable, the difference between these overbough/oversold indicators and trend followers is that trend followers make much more on occasion in a long trend, which make up for the losses. O/O indicators, on the other hand, take small profits and don't make up for the small losses. Another important consideration is that commissions will eat up what little profits you earn. I only included discount broker commissions in my tests, not even slippage from spreads. Some people believe Bollinger Bands are logical, but they may not always be used logically. They say 2X standard deviations should make the stock stay in the bands 95% of the time. First of all to make this statement you must prove stocks follow normal distribution, then you must realize that when you are talking about that 95% number, you are talking about 95% of all sample sizes of 20, not 95% of each individual days. The moving average of a Bollinger Band is 20. 95% of the time, the moving average will be between the last standard deviation. Not price. All of the tests were done based on opposite signals for longs than entrys. This will make it more difficult to be positive in an upward bias Dow. In other words, it should, theoretically, go short as much as it goes long. You could expect the tests to make less money by going short than by going long due to the upward bias in the Dow. You could compensate for this by making the entry rules for shorting more strict, therefore less frequent, but I have not done this.In any case, if the market were truly random, and that TA doesn't work, then I don't see how any of these tests would have been profitable. One would expect the systems would give negative results, due to shorting an upward movign market, not always being long an upward moving market, and including commissions. Such wasn't the case here.Also, not one of these tests showed an unnoticeable drawdown. In the case of Parabolic SAR (.02,.20), the drawdown started in 1975, and the equty line remains 99% off it's peak. The system was on a tear in the early part of the data, but after 1975 the system should have been faded. Most of the drawdowns have been for a few years at least in some occasions. For example, the SAR (.02,.01) peaked in 1937, passed that peak in 1964, and met the same level again in 1974 before continuing higher again.The reason I am using a maximum step of .01 is because I have found that way more profitable than letting the SAR get too close to the stock price, which almost always gets you out in the beginning of middle of a trend.
Best Of |
Favorites & Replies |
Start a New Board |
My Fool |
BATS data provided in real-time. NYSE, NASDAQ and NYSEMKT data delayed 15 minutes.
Real-Time prices provided by BATS. Market data provided by Interactive Data.
Company fundamental data provided by Morningstar. Earnings E