Bollinger Bands® Message Board

[apashaei@gmail.com]

Dear John,

I was wondering if you've done any studies regarding the slope of the bands & their directional relationships to one another.

The bands can either (a) diverge (b) converge (c) move in parallel. There are 3 variations for each of those state, creating 9 total states. For example, bands could both move (i) up, (ii) down, or (iii) sideways. And converging bands could either (i) head toward one another, (ii) lower band moves sideways, upper band move down, or (iii) upper move sideways and lower move up.

In your fine book, which I've read a number of times, you do cover some aspects of this when discussing bandwidth (chapter 8), but bandwidth removes the directional element of these various possible states, since its function is to simply show expansion and contraction in volatility.

Here is a scenario: In an early part of an up trend, we might get upper band curving upwards and lower band curving down (covered in your book) as a impulse move to the upside. Once the lower band curves back up, that suggests the momentum is gone but now both bands moving up in parallel suggest the trend remains up (this is in my experience). Once the upper band curves down while lower band still moving higher, we have converging bands and this suggests a sideway market. This allows for the systematic categorization of the state of the market at any point in this. Assuming we're not after continuous advice, as you so wisely point out, it can still be used as a good filter for building systems. I'm an option trader, so it also helps me to know that the trend is possibility over for now, and I should increase the 'theta' in my positions and reduce my long 'gamma'.

The divergence/convergence of the bands, in this respect, seem to reflect momentum. I've done some comparison with some popular momentum indicators such as MACD/H, ROC, and TRIX and have confirmed that there is a strong correlation there. And yet, I've also noticed that the bands are smoother while at the same time more responsive (less lag).

In order to filter out any noise and smooth out the turns in the bands, I've applied Kaufman's filtering technique that he uses to remove the noise in the slope of his KAMA indicator (Smarter Trading - chapter 8). Interestingly enough, he uses standard deviation to do that. I find his method quite effective.

I've started doing some testing & am seeing some encouraging results so far.

I was wondering if you've any thoughts on this or have done any in-depth studies regarding this?

Thanks in advance.

Alex

[bbands@bollingerbands.com]

You’re obviously an advanced user. Congrats on getting beyond the first veil.

The relationship of the bands is an often overlooked, but extremely valuable, area. Some aspects are easy to see in BandWidth, which I addressed in BoBB. Other aspects are easier to see on charts, but can still be coded for systematic trading after one becomes familiar with them.

The key to this analytic approach is the rate of change of the middle band coupled with the rate of change of volatility, trendiness and noise.

I’ve been teaching a few of the patterns one sees at tops in recent talks. For example consider this four-part topping sequence: a turn up by the lower band, a turn down by the upper band, a failed try at a tag of the upper band and a sign of weakness.

I am afraid I haven’t any proper studies to present, but I am sure that you will find this to be an area of sustained value.

Please report back with your findings and observations from time to time.

Regards,

jab

[apashaei@gmail.com]

[bbands@bollingerbands.com]

Since the middle Bollinger band is selected to be coincident with the intermediate-term trend, the rate of change of that band is a trend indicator of sorts.

I agree that s/n is a worthwhile area of study. So is entropy. Have a look at the Information Theory entry on Wikipedia.

I write about patterns such as that four-part topping pattern from time to time in the Capital Growth Letter.

jab

[l.v.spanje@hccnet.nl]

Dear Mr. Bollinger,

After reading your book/articles I have been an BB addict for several years despite the people around me who cannot understand why I like those swinging lines

I found myself very happy with a combination of

on the chart: BB 20,2 with 1 STD lines too (usually coinciding with ema8 in trends). price within 1 STD lines warns me of non trending situation

BBsqueeze indicator as John Carter and Hubert Senters (Tradethemarkets.com) use: BB 20,2 inside Keltner together with an overlay of Momentum to show direction. In this way the signal is "underlying" specific.
Purple dots represent the squeeze signal. If it turns blue again BB are expanding. Momentum direction (eg up when under 0) shows the way.
I projected %b as the second overlay as a flat MA20 shows me better price behaviour at the 1 and 2 STd as well as the ma20.
(no time yet to reprogram the indicator to project the squeeze indicator on the 0.5 %B line=ma20)

[img=http://img189.imageshack.us/img189/6807/gbpusdbb.gif]


What I am still missing is the influence of the MA20 slope on the appearance of the bands:
I often miss the opposite band curling back again (Volatily no longer outpaces MA rise). As I understand it it's because either the MA is rising slower with "constant volatility" and/or volatiliy is dropping relative to the rising MA. Up to now I haven't found a proper visual presentation that reflects these stages.
Bandwith rules out the influence of the MA slope, thereby missing the "opposite Band signal" .

Do you have some ideas/suggestions?

thank you for you inspiring articles
fixer

[bbands@bollingerbands.com]

Actually BandWidth does capture the effect you are looking for to some extent. Consider for a moment the BandWidth formula:

[l.v.spanje@hccnet.nl]

thank you for your answer.

I cannot "see" it yet, but I will study it

fixer

[bhouse1273@gmail.com]

I use a widget for the MetaTrader4 platform based on the Bollinger Bands indicator that comes with the platform. It signals buy or sell when the price intersects the lower or upper line... the classic "Bollinger Bounce" strategy for ranging security prices.

However, I added a slope-based feature that detects a strong trend by measuring the current bar lower or upper line with the corresponding value n bars in the past (typically, only 2 bars). When a difference greater than a user-supplied slope parameter is detected, the buy or sell signal is reversed, allowing the trading bot to follow the trend instead of the "bounce".

Ideally, the slope detection parameter would be dynamically adjusted to market conditions, but hey -- one feature at a time.

Bill