Avoiding Multicolinearity

Discussion in 'Technical Analysis' started by jumpbackjack, Aug 26, 2009.

  1. A cardinal rule for the successful use of technical analysis requires avoiding multicolinearity, i.e using two indicators that use the same data in their calculation such as closing price to get confirmation of a price move. Does anyone know of a source (book, internet site, whatever) where they have listed the indicators that work well with each other. I learned this from Blollinger's site at http://www.bollingerbands.com/services/bb/?page=6 He did state the RSI, MACD, and ROC would not be a good fit. However, RSI, OBV, Money Flow would be good. Makes perfect sense to me, I just don’t want to reinvent the wheel and I'm sure someone somewhere has this information.

    Thanks for your help.
  2. Why bother with websites that are still in the last century? Learn price action and forget about indicators. Indicators are for losers. Any indicator that is derivative of price is a means of redistributing your wealth.
  3. Thanks for your response there intradaybill (at least you did respond), but we each have our own trading styles.

    There have been almost 100 views and only 1 response. It seemed to me to be and extremely relevant question if you're using indicators for trade confirmation (aren't we all, bill excepted?).

    Come on, am I out in left field or does this make sense? If it makes sense then someone has to have figured out which indicators are compatible and posted it somewhere.
  4. Let me give you a hint.

    You want to indicators that are non correlated.

    For example, cycles, then volume/ order flow, then past price. We enough of them agree take your shot.

    Bonus. If you are trading multi instruments you want to trade non correlated. Don't trade crude, ES, and Canadian, because they are all highly correlated. They are all highly correlated to the USD.

    Why? Because if you lose you will lose real big. Pros focus on risk.

    BTW canned indicators suck for the most part as the OP mentioned.
  5. Corey


    The best thing I did was to begin to classify indicators into types (trend following, reversion), time-frame (based on parameters), and what 'derivative' they are trying to capture.

    Moving averages may plot 'position', while MACD plots the velocity. Another derivative of the MACD may plot acceleration. You should be cognizant of what derivative each indicator is attempting to plot.

    So by classifying indicators into types, time-frame, and 'derivative', you can avoid multi-collinearity, and also identify WHICH indicators will go well together.
  6. jprad


    I thought the same, until I began dissecting the random walk index. It's a trend indicator computed using highs and lows. If you take the difference of the two RWI's you come up with something that looks remarkably similar to the RSI, which is an OB/OS oscillator computed using the closing price.

    I think Bollinger's salient point is to simply eyeball your indicators. If there's a strong visual similarity then there's a good chance they're collinear, regardless of how they're computed.
  7. If you insist on using indicators, you need something that:

    a. measures the market when it is trending. the tool to use would be a moving average like indicator, the market action to look for would be buying higher highs/higher lows and selling lower highs/lower lows.

    b. measures the market when it is in a trading range. the tool to use would be some type of "bound" indicator (stochastics, etc.) the market action to look for would be selling the highs and buying the lows.

    c. meaures the market when it is in low volatility and high volatility modes. use the trending indicators when the market is in high volatility mode, and use the trading range indicators when the market is in low volatility mode.


    As people have said, most traders move past canned indicators relatively quickly in their trading careers and develop their own methodology for trading the markets.

    Good trading
  8. A. Beginner: indicators, TA, visual trading. Result: blow up sooner or later
    B. Intemediate: Automated trading system. Result: blow up, maybe ater
    C. Advanced: Algorithmic trading based on price action. Result: $$$

    Learning curve and development cost (software/hardware)

    A. A few months. $0 to a few thousands
    B. 2- 5 years, around 10K
    C. 5 - 10 years, 20k - 100K (depending on whether you do programming or hire someone)
  9. mokwit


    Hmmm, third post and you are posting a link and a cardinal rule for success plus answers.

    If you are not a shill, ask yourself, does Bollinger trade or make his money from books/CDROM/Seminar/website etc?
  10. veekay304


    #10     Sep 8, 2009