StrategyTrading1

First, let me define “micro-trading”. This is a term I use to describe trading bar intervals shorter than one day in length. For example. one could  format symbol settings to intra-day then in the minutes category any number less than 390. Why 390? Because, there are 390 minutes in a normal trading day. A common number is 15. Fifteen minute bars are a popular time interval for intra-day trading.

Why intra-day trading? My testing, and the testing of others, has shown that it can be more profitable to trade intra-day bars than daily bars. There are many important market moves that happen intra-day that can’t be captured with daily bars.

So, I call this type of trading “micro-trading”. A similar term is “day trading”.  However, this implies closing out one’s positions at the end of the trading day, and carrying no positions into the after market hours.

I used TradeStation to backtest AAPL (Apple Computer) over the period 10/27/2008 through the close of today (2/13/2009). In my previous post (”Is This Market Tradeable?”) I noted that there had been little price change  in the SPY over this period of time, while there had been much volatility, with the market moving violently up and down within a loosely defined “trading range”.

The system I applied to this test  I named Simple_LS_01. It’s a simple buy and go long, then sell and go short system using a moving average, that can be optimized.

The logic is: if price (close) goes above the moving average the system buys (goes long) and covers any short positions. If the close goes below the average, the system sells, and goes short. The trades are made at the market on the next bar open. It’s probably the simplest system one could create, but it’s a good test of bar length optimization.

Because the trading day consists of 390 minutes, these minutes are evenly divisible into a number of bar lengths. For example: 195 min bars, 130 min, 78 min, 65 min, 39 min, 30 min, 26 min, 15 min, 13 min, 10 min, 6 min, 5 min, 3 min, 2 min, 1 min, and of course tick by tick data. All of these numbers divide evenly into 390.

I set my beginning date to 10/27/2008 - the beginning of the current trading range market.

I assumed a $5,000 position in AAPL on each trade.

Let’s first look at a simple “buy and hold” strategy. A buy of AAPL on the morning of 10/27/2008 would have resulted in the purchase of 52 shares at 95.20.  AAPL  hit 112.19 on 10/30. The maximum intraday drawdown occurred when AAPL printed a low of 78.20 on 1/20. This was a drop of  -33.99 points from the high, times 52 shares for a total of $-1,767.48 or -35.3% on the initial $5,000 investment. 

The ratio between the maximum gain and maximum drawdown (ignoring negative sign) is 16.99/33.99 or .50, or 50%. I would consider this a poor reward/risk ratio. AAPL closed today at 99.12. So, that position gained +4.1% to date.

For each intra-day bar period I optimized for the moving average that produced the best gain and lowest drawdown.

I adjusted each test so that the first trade occurred on 10/27/2008. If you are using TradeStation to test and you want to make sure all the tests are comparable, you have to adjust the “maximum number of bars study will reference” and the date range, so that the first trade occurs on 10/27/2008, or whichever date you want to begin with. This post is a revision of one I published earlier today, and now includes this adjustment.

The best results were obtained with bar lengths that varied between 39 minutes and 5 minutes. The 39 minute bars made $2615 profit with a $-1003 drawdown, for a Net Profit/MaxDrawdown ratio of 260.7%. 40 trades occurred and 35% of them were profitable.

The 10 minute bars made $2674 profit with a $-640 drawdown, for a Net Profit/MaxDrawdown ratio of 418.1%. However, this system generated 70 trades and only 23% of them were profitable.

Profits were below $2000 on bar lengths above 39 minutes and below 5 minutes.

If one is thinking of “micro-trading” and using short duration bars, such as 1 ,2 3 min or tick bars, a very common practice, make sure to test this bar length first. Don’t assume a short bar length will be more profitable than a longer bar length!

Also, the shorter the bar length the more bars in a day, and the more one’s computer resources are taxed. Short bar lengths take longer to optimize and generate a huge number of trades, particularly if one is trading a number of stocks. This generates a lot of accounting aggravation at tax time.

Of course, this test was only on one stock and over a choppy market environment. Any system should be tested over different market time periods on a basket of stocks, and a simple system like this one is just a starting point for any serious system trader.

However, it does demonstrate a basic idea of system testing: start with an idea, in this case “micro-trading”, then find out what does and doesn’t work, and begin to eliminate the failed strategies, and keep the successful strategies. It’s similar to “natural selection” although this is “tested selection”.

If you keep testing ideas, you’ll gradually develop better and better systems, and hopefully, will improve your profits, and minimize your losses.

One additional note: I don’t just optimize systems for the highest profit - I look for the max profit as a percentage of the max drawdown, because I want to maximize profits and minimize losses. Remember, controlling drawdown and losses is critical if you want to become a successful trader.