How do you use a regression channel??

Discussion in 'Technical Analysis' started by stocon, Jan 19, 2003.

  1. stocon


    I'm curious if people might share their thoughts on the use of this tool and how they may use it and any pointers or tips or drawbacks as well as opinions.
    thx Steve
  2. no one knows? i would too be interested if anyone could shed some light on this
  3. stocon


    Like anything I guess sometimes it works like a charm and other time not. I wonder if there might be some filters, fits etc that might be helpful.
  4. gordo


    If you are speaking about a channel like the Raff regression channel, it can be viewed somewhat like a set of price bands ( IE- Bollinger bands). What I mean by that is you have a set of lines (trend lines) that represent trend support (the lower line) and trend resistence (the upper line).

    If you have an application like Metastock or Tradestation, they contain regression channels as an indicator option that you can draw from various price points in a chart. Last that I remember, Metastock also contained a breif synopsis on how to apply/interpret them. But basically, you draw from the lowest low to the highest high in an uptrend, and vice versa in a downtrend.

    Essentially, they help you identify where most of the price movement is likely to occur in a trend. From there you can use them to guage acceleration in price movement (upside breakout from the channel) and reversal of trend (downside breakout from channel in uptrend).

    If you project the channel to right, that would be into the future, on a chart, it can serve as reference for future price targets. Let's say that you have been trading a stock with a regression channel drawn from the low of $5 to the high of $35 (it's peak). Over the next few months the stock corrects and works it's way down to around $20. After a little work, the stock appears to be ready to move higher again. If you drew your regression channel so that it would extend out into the right (future) on the chart, the slope of the trend might suggest that say in 6 mos. the stock would be trading around $50. Note however, that the bottom trendline in the channel would become resistence (another potentially useful monitoring tool).

    I hope that this helps you guys. It takes a little practice to get familiar with. Some of the past issues of Stocks & commodities and also Active Trader have done articles on regression channels. Can't recall with issues, otherwise I would post them.

    Good luck,

  5. You can also use one or more dynamic channels to gauge the current direction and relative range and enter/exit based on channel breaks.
  6. stocon


    When you say dynamic, do you mean Keltner or bollinger Bands?
    Could you explain a little more?
    thx Steve
  7. The subject was about regression channels. I was referring to using a dynamic regression channel - one that moves as time passes (rather than just drawn from one point on the chart to another manually) giving you a picture of the current trend channel for whatever timeframe you're interested in (e.g., maybe the last 20 bars). A breakdown/breakout of the moving regression channel can help identify a change in direction.
  8. man


    I just wrote a little abstract about the subject since I want to get into testing channels of that kind in summer. My thoughts are basically:

    1. key point is the starting point of the channel. Since I want to completely atomize the process, i need to define that sharply. It is not as easy as to say "take the high or low of the last x days as a starting point", since one major strength of the idea is that the channel is flexible in its length.
    MA crossovers do not help either, since they are either too late when you take two longer MAs or cross over to often if you take short MAs.
    We might use some kind of advanced statistic to define bottoms and hills in historic prices.

    2. I would like to calculate a regression line from the recent extreme (found in 1.) to the most recent price. I would like to make two additional, parallel lines to this regression line, one up and one down. These are the channel borders. I want to allow them to have different distance to the regression line, by drawing them at the distance where the number of outliers (which is a parameter) is the same for each side.

    3. now at any point of time I have the channel, describing behavior since most recent extreme value. The channel has three main features: length, slope and width. Now the shape of the channel can go into a pattern recognition system, which waits for a specific shape to appear. For example a narrow downsloped channel with a length of more than x. If the upper channel border is broken a long trade is executed.

    4. since the source of the whole thing lies in 1., where the starting point is defined, it is possible to calculate channels with different length, by looking for other starting points. This means that I can apply two different channels at any point of time. The longer one gives me the big picture, eg US stocks are still in a downward channel, whereas the short channel gives me most recent description of events, eg we had an upward channel of several weeks. Now I can use the longer picture as a filter and only take short-signals from the short channel.

    These are the basic ideas. It is not easy to test, so it will take some time. comments appreciated.

  9. I think that you are better of if you just plot 15period MA of lows and highs and see how the price is behaving. You will have to make adjustments to accomodate different conditions in different markets.
    #10     Jan 21, 2003