There are hundreds of indicators available for trading and innumerable combinations of indicators that may be used for trading. Every indicator has something to recommend it and much going against it. The issue is not whether an indicator is useful for trading or not, the issue is how you, the trader, mixes market inputs and indicators to produce a workable trading edge. Ultimately it is success that counts, unsuccessful traders who have failed to develop a trading edge do not last very long in trading.

I programed this indicator and tested on other timeframe. Result is not well: price is late for traiding.

What timeframes and markets are you trading? Do you use the same indicator for each timeframe? Do you use the same indicator for entries and exits?

Hi John !

No matter what time frame. The price which was calculated for opening or closing position (crossing bollinger's band and line of the price) is more different from the closing price of the time-frame .

Hi Alexander,

For the Bollinger bands, I tend to use these for extremes, a bar closing outside followed by the next bar closing inside the band.What indicators do you find to be useful?

Yes, If the bollinger band is too narrow then the only way it makes sense to enter would be if the band appeared to be about to widen significantly. If the potential isn't there don't enter not enough potential reward for the risk.

Hi John,

People get way to hung up on fancy words in trading. Simple is always better. So if it doesn't work, you can tweak it later.

If you are familiar with your first year college stats course, Bollinger bands when all boiled down, are really just confidence intervals. So the underlying assumption is that you have a mean reverting process. Basically, once you are outside of a "tail probability", the process is assumed to revert back to the mean.

This may or may not be true!

Cheers,

Elliott

#Shifmanelliott

#Shifmanlelliott

#Elliottlshifman

#Elliottshifmanl

elliottlshifman@gmail.com

Every trade boils down to the following, a security is either breaking away from the mean or reverting to the mean. Sounds easy takes a lifetime to master!!

"Investing is simple, its just not easy". Warren Buffet

i've been back-testing bollingers in daily data for more than 3 months... nothing worthy yet..

Hi: To the person who mentioned that they are confidence intervals, They are quite close

to the standard regression confidence intervals but not exactly. See the paper below for the connection between Bollinger Bands and various time-series models. Most of what is in the document is well known in the time series literature.

The interesting part to me is the proof of the relation between the duration of the trade and profit. The result is interesting because it shows that it is not possible to profit from the trade, once the trade duration passes the length used in the calculation of the moving average.

http://arxiv.org/abs/1212.4890

CRITICAL NOTE: In the paper, the proof is stated as an if and only if. I need to correct it because it turns out that it is only true in one direction: namely, if you hold the position for greater than n periods, then you cannot profit. but, if you hold it for less than n periods, you may or not profit. The proof states that, if you hold it for greater than n periods, then you definitely do not profit and, if you hold the position for less than n or equal to n periods, then you definitely profit. There is a mistake in this part of proof which makes the latter part of the proof not true. I need to clean the proof up for this correction but you may have an interest in the former part of the proof which is true.

http://arxiv.org/abs/1212.4890

Thank you Mark, nice effort. I went over the paper briefly and what I see is a theoretical model, with the proof you state being basically a proof of the model's assumptions within the stated framework of said model. I see no actual trading being done or even planned as a way to validate the model in the market. It appears that this model is projected to remain an academic exercise with a projected upward spiraling curve of proof of the model itself. This will most likely be followed with an expansion of the model into other types of theoretical trading applications using different setups and assumptions, adding yet another level of complexity and assumptions to an already complex model. The point I am making is that models are fine as academic exercises to demonstrate concepts, but the market's complexity and fluidity have a way to squash the assumptions used in the models and make them unsuitable for actual day to day trading.

I appreciate the great inputs. One of the ideas I've been trading is using 3 BBs: spaced at 1.5 std, 2 std and 2.5 standard deviations, with the simple Moving Average as the center-line. This sets up my playing field. As price moves into each of these "zones" I can adjust the size and direction of a trade. I use the width of the bands as a judgement signaling whether volatility is likely to increase or decrease and then look at the position and primary direction of the price, relative to the BBands, to trigger appropriate adjustments. The core concept is that each trade is be a mean reverting round trip trade.

Hi Oscar: I hope all is well and I appreciate the comments. As far as the relations between bollinger bands and time series models, I totally agree and the results are well known in academics.

But the theorem-result about duration and profit is not an academic exercise and has useful applications. The only assumption is that there is no slippage. And if there is slippage, the result will still hold approximately. The result is the following in case you didn't get to that part: ( it's near the end so, if you skimmed, you could have missed it ).

Suppose one trades a position using the standard "go short at the top band

and go long at the low band" and exit when you revert to center line" Bollinger Band strategy Also, suppose that the period used in the calc of the moving average is n.

Then, assuming that the position is held for greater than n periods ( i.e: we didn't reach

the center line by n periods ), then it is NOT POSSIBLE for the trade to have

an overall positive return. This is a useful result, because after n periods, you know

that the trade you are in cannot be profitable overall. This result is independent of the band width used !!! Note that the result is not saying that you cannot make $ back

after n periods but rather that the OVERALL profit cannot be positive.

Another useful related result ( which is not in the paper but I plan on adding it when I correct the proof ) is that, after n periods, and at each period after that, there is a closed form expression for the maximum $ can be made from that period on.

So, for example, suppose, n = 30 and you're at period 32 and you've lost 10K so far in the trade so far. Then, there is a closed form expression for the maximum amount of $ that can possible be made from period 32 to when the trade closes. Suppose this amount is 1K. ( and I have

an expression for it but it's not included in the paper. ). Then, since you're already

down 10K and you know you can only make 1K max going forward, then this means

that the least you can lose overall is 9K so you may want to exit immediately. I would think this knowledge would be quite useful for someone using Bollinger Bands as I have described.

So, in summary: I agree with you about the relations between Bollinger Bands and

the various time series models. But the duration result is quite applied and I think

quite useful and I think not well known. The only assumption is that the trader is using Bollinger Bands as I described and that there is no slippage. If there is slippage, the result is still quite useful but the relations-expressions will only hold approximately. I hope that helps to clarify the result.

Someday I hope to get the result in some kind of technical analysis magazine

but I need to clean the proof up and also remove all the connections between Bollinger

Bands and time series models. Also, a lot of thanks should go to John Bollinger for creating Bollinger Bands. Without his Bollinger Bands, none of the results would be true. BB's are an interesting animal for sure.

All the best and thanks for your comments. If anything wasn't clear, you can always email

me at markleeds2@gmail.com.

The proof was tested using tradestation ( Bob Fulks from the markets list helped me with the testing. I don't have tradestation ) and was shown to hold empirically. The other part of the proof was tested empirically ( the part where I said that if you exit before n, then the trade has to be profitable ) and was shown not to be true which led me to find an error

in one of the steps of that part of the proof.

Note that the reason I did not include the result about the closed for expresion for the amount of $ that can be made each period after period n was not because it was

propietary. I only realized that a closed form expression could be obtained while

working with Bob on the testing and I have gotten back to the paper since to add

that derivation-expression.

Hello Mark and thanks for clarifying the situation with the model. I would like to suggest a small addition to your work which I think might increase the validity of the conclusions: I would add a paper trading run of the model to the work you are planning to do. Choose any security, any time frame, use a pre specified amount of imaginary capital to be used in the run, and specific rules for triggering your trading signals with your model. Keep records of entries and exits, costs and profit/loss. Let's see what a run like this shows about the performance of a trading system based upon your model when faced with real market conditions. I do not think you will be able to simulate slippage but you could include a nominal amount of variation throughout the run to simulate slippage. Since you have someone dealing with the model in tradestation, presumably creating a trading system out of the model, this would be both a relatively simple and possibly very valuable addition to the work you are doing. You could obtain results for the same trade completed before and after n number of periods as long as you maintain a tight control over how entries and exits are determined by the model. After all nothing says a model would indicate only one exit point for a given trade. I would be very interested in the results of such a paper trade run. All the best.

Hi Oscar: Definitely that would be a good thing to add when I improve-clean up the paper.

All I can say in this email is that we did exactly that and proved the statement of the theorem empirically. All one needs to do this is

For various values of n where n is the period of the moving average, do the following

( using any band width ).

A) generate Bollinger Band scenarios ( using real market data prices ) where one takes a short ( long ) position due to reaching the upper ( lower ) band. and exits when the price eventually hits the center line.

B) Consider all of the scenarios in A) where the trade duration was greater than n.

C) Check if the scenarios in B) ever resulted in a + profit. If there's no slippage, they won't. If there's some slippage, they might slightly but it still holds pretty well. ( I realize

pretty well is vague ).

Bob Fulks and I did A-C together and convinced ourselves that it's true. ( Bob did most of the work. His simulations assumed no slippage ). But I absolutely

agree that, in a re-write, I should include a simulation to back up the proof. But I do

know it's true.

In fact, we also checked the closed form expression for how much can be made going forward after n, and that worked out also. So, it's a useful result for anyone who engages in that type of strategy. Of course, whether to engage in that type of strategy is an altogether different issue :). It's more of a risk management result for people who do engage in it.

I hope one day to get around to improving the paper ( by getting rid of the part of the

proof that is not true, namely that, if the duration is less than n, you definitely profit.

this is NOT TRUE ) and including simulations to back up the proof. There's also a

way to simplify the proof which will be useful also. Thanks for the recommendation.

Good thing Mark, sounds very promising. All the best

The problem with all Indicators & Oscillators is that they are lagging. The price and Volume are actually the only thing presenting reality minute by minute....

Hi Iris, I agree with you. Going into a trade, the trader is in full control and can use their selected indicators to enter a trade on their terms. Once in a trade, the market is in control, and from this point the only indicator that really matters is price.

no it is the bad one .... if you have any useful usage tell me plz

Hi John, great topic!

I personally like using Bollinger Bands, although mostly for confirmation of trades, since they do lag a bit. I believe that the use of indicators/oscillators and technical analysis tools in general depends on the personal preferences and goals of each trader. Thanks!

I know this topic is not relevant to our discussion, but, I would like to change my Charts Platform. I am currently using ESignal

Anyone has any reliable user friendly Charting platform to recommend???

Hi Despoina, Thank you for your insights. I agree that with all the tools available to the trading community, it is important to know the right tool for the job at hand. Bollinger Bands are a great resource to judge the whether volatility and prices are stretched to extremes.

Indicators give us the confidence to place trades. As traders, we have our views of what the price should be at some future point in time and we place our bets. As time passes we see how our price forecast performed against the judgments of all the traders who also voted with their capital. What a great system for discovering the fair price of an asset at any point in time.

Cheers!

@Iris, you are right this is not the topic of this discussion. Your answer depends on the type of trading you are doing.

I have done a lot of testing with Bollinger Bands. The interesting part is that the indicator seems to work better if you trade it the right opposit then what the author says. Anyway, I have found several usable secondary applicationst, however I would never trade just simply by Bollinger Bands.

I agree with Tomas. That's the case with most technical indicators. In addition the problem with bollinger bands is, it is not normalized. Would be a good idea to normalize it before using it.

And to forecast the whether? LOL!!! Technical analysis.... hilarious!

Dear all please share if you have any actionable thoughts about B. bands thanks !

Hi John,

Unless I'm mistaken, Bollinger Bands are really just a fancy way to say "confidence interval". Based on all of the probabilistic laws that you expect from a process that follows a Normal Distribution.

Seeking out tail end probabilities is not a bad thing to do in trading, but more attention should be focused on what variables you are measuring to seek out those tail end probabilities. I love to say "Garbage In - Garbage Out". If your statistic is a function of price action, you are unlikely to extract the predictive power you would want for trading.

Just my 2 cents. Cheers,

#Shifmanelliott

#Shifmanlelliott

#Elliottlshifman

#Elliottshifmanl

elliottlshifman@gmail.com

https://elliottshifmanbayesiantradingmodels.blogspot.com

Hello,

I use Excel to create some 7 charts all about the same company and symbol. 1 of the charts is Bollinger Bands. The Close is graphed along with the Bollinger Bands. There is a graph with closing prices and 3 moving averages, Dividends with growth rates, RSI with close, Several different pivot point calculations several with colored Support and Resistance points, 10 year Revenue graph, 10 year EPS and Dividends graphs, and a 10 year percent cost of goods sold/sales graph.

None in isolation is as meaningful as the combination of all of them, including the Bollinger Bands. The Bollinger Bands seems more meaningful with the close included in the graph.

Thanks, David