(It is Sat. night, it is cold and rainy, and generally miserable outside, so rather than watch George Foreman demostrate his new weenie cooker i decided to haunt these forums.) I want to throw in a little comment here regarding the interpretation of Bollinger's bands. It is often said that when prices stray two standard deviations from their mean, that "there is a only a five percent chance of the price moving further away from the mean." Or other equivalent and equally incorrect statements. Bollinger, himself, to my knowledge, never made such an error of interpretation. I read his book ages ago. That business of assigning probabilities comes straight out of statistics for a normally distributed population with a fixed and known population standard deviation. In that case, it is true that 95% of observations that might be drawn from the population would fall, on average, between plus or minus 1.96 standard deviations of the population mean. But i have never seen any data that would suggest that individual stock prices observed for finite periods are normally distributed about the mean. anyone in this forum have any idea what the distribution function of prices about the mean actually looks like? I would be most curious to know, and of course delighted if the distribution actually did obey Gauss's function. Surely there are some Ph.D. physicists toiling late into the night for Goldman that could shed light on this issue. Or perhaps one of you guys with access to raw time and price data will plot the actual distribution out so we can see what it looks like. In the meantime, we ought not to attribute specific % probabilites to price interceptions of the bands. Not anyway until we hear definitively from the physicists. There is nothing at all wrong with calculation of the standard deviation, however, since in any case that calculation is going to give us the most efficient estimate of the price scatter. Whoever said here that the bands give us a measure of volatility was certainly correct -- but it is not correct to attach specific probabilities without knowing what the form of the distribution function is. Now, there, don't you'all wish i'd go back to Foreman and his weenie roaster?
depending on the market as a whole the bollinger band buying breakout strategy (buying when the price touches the upper or selling when touches the lower band) is one of the most difficult strategies. I did research for about 6 months before I started trading real money using the system. I began using real money on system March 2005 and i was wiped out by October 2005. I still use bollinger bands but i tend to find myself using them less and less. They are much more of a supplementary tool for me.
no idea why someone would not use a BB over MA...BBs at least give you a stats measure against MA price move then you can decide from that move to bet on mean reversion or not.
not in this market i realize you are being humorous but this is an incredibly low volatility period, ESPCEIALLY considering we are in a strong uptrend when volatility returns, traders who have gotten too complacent are going to get hammered
I am compelled to correct my earlier post where i wrote: "....there is a only a five percent chance of the price moving further away from the mean." This should be 2.5% not 5%, since in the normal, symmetrical, unskewed distribution 95% of the area is incorporated between plus and minus 1.96 sigma leaving just 2.5% under each tail.
Whitster. I was deadly serious. I have spent many a dark night attempting to devise a volatility-based reversal indicator. Perversely (or perreversally), every one I tested was better if recast as a continuation indicator. Big runs draw old ladies away from their knitting and attract mischievous cats to unattended keyboards.
Combine your BB's with the RSI and you will know when the price is going to pop over or under the bands.
Deere stock, symbol DE tests profitably if I change the parameters a bit. Deere stock might not be the best choice for a bollinger band system. Other stocks might perform better with a bollinger band system. Using 25.04 years of daily price data from 4 January 1982 to 26 January 2007, 50 day moving average, 0.1 standard deviations, and 10 % heat I obtain these results: Number of trades 216 Total profit $ 401623 Profit after subtracting $ 10.00 commission, slippage per transaction: $ 397303 Heat is 10.00 per cent of equity. Greatest drawdown is 0.2028 (20.28 per cent). Cumulative Annual Growth Rate (CAGR) is 15.87 per cent. CAGR / Drawdown is 0.78 Instantaneously Compounding Annual Growth Rate (ICAGR) is 6.41 per cent. Annually Compounding Annual Growth Rate (ACAGR) is 6.62 per cent. Information Ratio is 0.13 Initial capital is $ 100000 Long trades only. Growth rates are calculated after subtracting commission & slippage. In comparison buy and hold shows these results: Cumulative Annual Growth Rate (CAGR) is 119.83 per cent Greatest drawdown is 50.15 per cent. CAGR / greatest drawdown = 2.39 Instantaneously Compounding Annual Growth Rate (ICAGR) is 13.71 per cent. Annually Compounding Annual Growth Rate (ACAGR) is 14.70 per cent. === I find that buy and hold growth rates are often much greater than growth rates obtained using any trading method that I recall testing. I also recall greatest draw down of buy and hold is typically larger than corresponding values from mechanical trading systems. I suspect that trading systems have more to do with reducing volatility than maximizing growth.
=========== I like John Bollinger better than his bands; I am sure you could find some use for them, but barchart.com, on 43 trades test of 20 day /ma bands averageing 3 days each on SPY had a loss . In contrast; SPY 100 day ma swing trade is profitable