Talk:Bollinger bands

From Wikipedia, the free encyclopedia

Contents

[edit] John Bollinger comments on this article

If I were allowed to write this article, I would follow this template:

A short introduction to trading bands with a link to the trading bands article.
A description of Bollinger Bands including formulas
A comparison of Bollinger Bands to other types of trading bands.
A description of the definitions of high and low on a relative basis that Bollinger Bands provide.
A description of the two indicators derived from Bollinger Bands, %b and BandWidth, with formulas.
An example of a lower-band trading signal.
The definition of the Squeeze and an example.
A discussion of the use of Bollinger Bands in pattern recognition.
A short note on the amount of data found inside the bands from a theoretical perspective with a practical example.
A section on using Bollinger Bands in trading system development.
Some ideas for future research and development of these and related concepts.

However, it has been made absolutely clear that this sort of contribution is not welcome here. Perhaps some other intrepid soul might follow this template and, in doing so, provide the sort of article on Bollinger Bands that would serve to inform users and make both Wikipedia and I proud.

John Bollinger, CFA, CMT -- Bbands 18:43, 13 June 2007 (UTC)

[edit] Chebyshev's inequality

Cherkash's assertion that "the bands by Chebyshev's inequality must contain at least 75% of prices, is incorrect." is itself incorrect. Chebyshev's inequality applies to all distributions, and for each value in an N-period sequence, the mean and standard deviation are fixed parameters of the distribution that generated that value. The fact that the simple moving average is a biased estimator of the mean only increases the estimated standard deviation.

The "Counter-example to which is a case of strong directional trend in prices accompanied by low volatility (the bands are narrow in this case and prices don't fit into the channel formed by the bands around the strongly-lagged moving average)." is self-contradicting. By the computational method used, a "strong directional trend" is itself a condition of high volatility, widening, not narrowing the bands, especially if "strongly-lagged." A simple moving average is not linear regression estimation. --216.77.225.171 19:38, 4 February 2007 (UTC)

Chebyshev's inequality doesn't apply, since only the last value of the N-period window is compared to the mean and standard deviation of that window, and each of the other points is compared to a different distribution (some earlier sample). For example, consider the sequence 1,1,1,...,1,20,400,8000,160000,... where the first 19 values are 1, then after that each value is 20 times the previous. Then every point after the 20th is more than 4 standard deviations away from the mean based on the trailing window (e.g. for the first 20 points the mean is less than 2 and the standard deviation is approximately 4.2) and so is outside the (20,2) Bollinger bands. While you might describe this pathological example as a "strong directional trend," it doesn't really look like a price series. I suspect the original claimed counter-example was confusing the standard deviation of prices (used to determine the bands) with standard deviation of returns (normally used in calculating volatility). Since both sides of the argument are flawed, I'm just removing mention of Chebyshev. Zander 00:57, 3 May 2007 (UTC)
The sequence you describe is not a random variable, but a deterministic concoction of your own. Chebyshev applies to random variables, like price and return. The extent to which they have trends which are determinable, only biases the moving average as an estimator of the mean. -63.98.135.196 17:23, 25 May 2007 (UTC)
The sequence I describe could be the realizations of a random variable. A deterministic series can be considered the realizations of a random variable that has a probability distribution conditioned on the history that is always a Dirac delta function. Or if you have trouble seeing that, my series is also a possible realization of independent samples from a normally distributed variable with zero mean and variance 1 with some positive probability (or, say with mean 1 and variance 20 with some larger probability) (at least for any finite length). In any case, I was only trying to provide an illustrative example, but the main point is that Chebyshev's inequality only allows you to a) bound probabilities given a distribution or b) bound the number of outliers in a sample based on the sample mean and variance of the exact same sample. Specifically, a) does not allow you to say anything about a finite sample, because with some small but nonzero probability you will get only samples far from the mean. And b) does not let you say things specifically about the last sample in a rolling window, only about all of the samples in the window simultaneously, specifically the last element could often or always be one of the ones that is far from the sample mean. Zander 03:19, 30 May 2007 (UTC)


This is ridiculous that it hasn't been fixed yet. Zander is clearly right. Suppose we are looking at a group of 20 data points on a 20-period Bollinger Band. Chebyshev's inequality will say that 75% of those values in that group will lie between two standard deviations from the mean, but that does not mean 75% will lie between the Bollinger Bands. The difference is that the mean and standard deviation at each point is calculated over a different group of data points than the mean and standard deviation at point 20.Image:Bollinger_Bands.jpg As an example this is a random variable whose does NOT have 75% of the point between the red Bollinger Bands. Chebyshev's inequality says that at least 75% of the last 20 periods will lie between the mean +/- 2 * stdev (shown in green) at the last point.
Also, Chebyshev's inequality works for static data sets in addition to random variables.--Colinc719 (talk) 04:04, 26 February 2008 (UTC)

This exact fraction in the plot above is 0.742857, and you don't show the data from which your bands were calculated; there may be an error. Chebyshev's inequality can be used to show that at least 75% of the data lies between the Bollinger Bands. This is applicable, because this probability statement applies separately to each pair of points of the band which are calculated at one point in time for which the moving average and moving standard deviation are fixed quantities, at that point in time. Non-random counter examples can be concocted, but it is to random variables, like price, to which this applies. —Preceding unsigned comment added by 63.98.135.196 (talk) 17:10, 17 April 2008 (UTC)


Please look through a proof of Chebyshev's Inequality to understand how it works better. It does not just apply to random variables, but rather applies to any data set. It does not make sense to say that "Non-random counter examples can be concocted, but it is to random variables, like price, to which this applies" because that would imply that if a price happened to move in one of those non-random counter examples, 75% of the data would not be within the Bollinger Bands.--Colinc719 (talk) 22:29, 4 May 2008 (UTC)


The article makes opinionated statements about what the bands can and cannot be used for.

Name one.

[edit] But what *is* it.

This entire article attempts to describe what Bollinger bands are *not*. Virtually every sentence is some sort of negative. Bad writing at its worst. Would it make anything worse just to delete the entire article? 70.132.12.137 18:21, 27 April 2007 (UTC)

[edit] Interpretation

The interpretation section was written to be general and to describe generally how bollinger bands are interpreted. What was written in the intepretation section was general trader "lore", how people have historically used the bands. Trade Like a Hedge Fund by James Altucher and The Volatility Course by George Fontenelis. It is my personal belief that the scientific spin this article has taken is the incorrect viewpoint, this seems to correlate with the user who talked about "astrology". Now the buy the lower bollinger band and sell at the 20-day MA does have scientific validity (Trade Like a Hedge Fund citation, ch 4) which shows 5 year profitability of the 100 Nasdaq 100 stocks. Backtesting options is nearly impossible, so the general description of bollinger bands (Volatility Course) and options will have to be taken at face value, but is in my experience discussed among traders as a strategy, hence general trading lore. This goes along with Pleclech's "as long as it doesn't make claims for their scientific basis". Again the interpretation was written in a general tone.

Furthermore, I don't think the section should be deleted but should be added to. For instance if someone has a problem with the buy lower bollinger band sell upper bollinger band strategy, it shouldn't be deleted just because there is another interpretation called bollinger band breakouts that states that stocks rest, volatility (hence bollinger bands) contract. After a period of stock price resting, prices will break through the bollinger band and "ride the upper or lower bollinger band". This is an apparent contradiction of the mean reverting strategy (buy lower bb, sell upper bb), but nevertheless both strategies are used and discussed by traders, thus both should remain. Technical Analysis is more an art than science and therefore should be presented as such, incorporating various uses of the bollinger bands by the "craftsmen" aka traders that use them. TraderMatt May 27

[edit] Mr.Bollinger Quote

Adds nothing to what has already been written in the article. Mainly, interpretations of bollinger bands is diverse (stated in the interpretation section). The comment on creating rigorous trading ideas/investing approaches would probably best be discussed in a strategy creation forum. I would suggest that in your many years studying the bollinger bands, that you have general yet specific interpretations of the bollinger bands. Additions of those intepretations would best benefit the user rather than philisophy and vague instructions on building a technical indicator from the ground up (My interpretation of what your quote was saying). TraderMatt May 27

[edit] lack of example

Bollinger bands can be showed on a weekly or monthly basis , with examples when the full bar exits over the top band , usually an indicator for a technical correction to follow.

Another example could be the whole travel from the upper band to the lower one.

IF Mr. Bollinger could rewrite the whole article , it would be a huge bonus for all the readers.

I do not agree that there is a conflict of interest. IF the inventor of anything is ready to write an article about it - why not?

My English is not good enough, but on the Hebrew wikipedia I rewrote some articles on fields that I am an expert and it was welcome.

--YoavD 14:08, 11 June 2007 (UTC)

[edit] It's OK by Me, but John Bollinger could do a better job

I have no beef with the article as it stands. I don't see any major incorrect statements, but i know nothing of Chebyshev's inequality. It seems fairly neutral, and factual to me. But i do think it's a pity that Mr. Bollinger's contribution would not be welcome. He is afterall the expert with regard to the practical application in trading. There is nothing at all wrong with calculating the bands using the standard deviation formula. But it seems that not a few people have drawn incorrect statistical inferences from Bollinger bands. So far as i know though, Mr. Bollinger has not been guilty of that. I read his book and found it enlightening with regard to applications in trading. —Preceding unsigned comment added by Piezoe (talk • contribs) 00:02, 13 January 2008 (UTC)