The Power of Mass Deployment

The power of statistics emerges as the sample size grows. I know, it has been repeated multiple times in Stats 101, a bunch of youtube videos, or maybe Statistics for Dummies. But has it stopped people from making judgement calls on purely empirical basis? Statements like “I’ve seen it working x times, so it’s legit” or “it’s a bad indicator because I tried on several stocks and it didn’t work” don’t really make sense when you are living in a complex realm composed by incredible amount of data, multiple dimensions of reality and endless chain reactions such as public administration and stock investments.

To illustrate, I did a back-test using a simple combination of Bollinger Bands and MACD Indicator from 2004 to 2012. It’s an end-of-day, mean-reversion strategy with a price filter and a liquidity filter. After testing it on 30 random stocks listed on TSX, this is what I got.

Nope, not impressive. But it looks quite different if we deploy it to the entire market, which is about 2300 stocks listed on TSX (1400 after using survivor filter ).

Commission is not a concern. As shown below, most of the time the strategy only holds less than 2 stocks, not 200.

The real problem for implementing this strategy, for retail investors, is computing power. Gathering latest data, completing calculation and executing trades right before market closes every day precisely is very challenging for individuals. For big players, it’s liquidity. Because the strategy targets low liquidity segments, it can’t guarantee the trading volumes will be big enough for institutional traders.

7 thoughts on “The Power of Mass Deployment

      • this is very hard for someone with only yahoo EOD data. I wonder how the result looks like if order is placed next day at open or close after signal is generated. like quantstrat does.

  1. What was the strategy? Does MACD really improve the results? My experience is that its performance is pretty dull since there isn’t much short term autocorrelation.

    I don’t think the implementation would be that hard, the program could keep track of all then stocks that would be most likely not to give a signal during the day to speed up execution.

    • It’s a combination of BBand and MACD. MACD is just a safety measure. It’s very hard to accurately find short term autocorrelation in one stock, like shown in the first graph. But we can improve it with proper filter and big sample size.

      Theoretically, it’s not that hard for individuals to acquire that computing power (Amazon can do it) and build up the system. But to keep it up and running 24/7 bug-free is the most challenging part.

  2. Is spread a problem? Sometimes I find it to be a problem with low liquid stocks (which, I’m guessing occurs within these 2300).

    • Hi, implied volatility: it’s a problem indeed. That’s why this strategy has a very conservative liquidity filter. It doesn’t trade anything with low volume or big spread.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s