I started an article on this topic in my last week's Trader's Corner by working on a promising trading 'system' applied to the Nasdaq 100 index. I got different 'signals', 1 'long' in calls versus 1 'short' in puts, when I optimized results with QQQQ versus the underlying NDX index; looks like the result I liked best (short, in puts) is winning here.

Besides my introduction to the topic of 'systems trading' that I took up last week in Trader's Corner (8/27/09), I carried on in my SATURDAY Index Trader column with my lead discussion relating to forecasting an up or down move as it was suggested by one promising indicator-based trading system:

"TAKE YOUR PICK: be long Calls or Puts."

In my weekend column I noted some recent studies with the Nasdaq 100 (NDX) intraday history I have; 2 years and 9 months currently or 1000 calendar days. I started with the idea of shorting (buying puts) when the 21-period Relative Strength Index (RSI) goes BELOW 70 on an 60-minute chart with trade entry assumed on the close of that hourly bar. Exit on puts and reversal to a long call position occurred only when an hourly bar closed above RSI 30, after being under it.

I 'applied' my starting idea to both the NDX index and to the QQQQ tracking stock. I had 1000 days of intraday history so that's what I used; my strategy idea was to use an hourly chart which is a commonly used time period for trading ideas in the indexes.

My TradeStation software, the ultimate in Systems Testing and Development (STAD) software, is set up to display a back tested P&L for individual stocks. With indexes it just shows points gained or lost. With a stock, I can enter a starting account size, a 'slippage' factor (assumes a fill that is more than or less the point when the signal was generated), and a commission. I could enter a stop-loss trigger also, but haven't so far.

I then 'optimized' as to which RSI triggers besides 70 and 30 (which 'overbought' or 'oversold' levels) produced the best profit for the nearly 3 years of intraday data I had. I ended up getting DIFFERENT trigger levels and different results in terms of profitability, for the QQQQ tracking stock versus the underlying NDX index. The QQQQ RSI system as of Saturday was long (still 'on' a buy 'signal'), whereas the NDX RSI trading system has been short or forecasting a decline.

The most profitable RSI 'trigger' values when back tested (the optimization process) for the history I had, were 72 and 39 for QQQQ, but 69 and 31 for NDX, using the same 21-hour length setting in the RSI. The point gains and percent of trades profitable were greater using the NDX optimized values than the ones showing best results for QQQQ. CAVIET: exact profitability of the NDX system can't be as figured given all the possibilities of WHICH calls and puts were bought, their costs, etc. Not so in a stock.

With QQQQ I could start with a theoretical $10,000 account but the back testing results suggested I need at least $21,000 for the period tested if I was going to allow all pyramiding of trades, given the maximum number of shares bought or shorted at any one time AND to cover the largest loss sustained on an unrealized or realized basis. I entered a 'slippage' factor of .01 per trade and $25 commission cost.

In my RSI Buy-Sell strategy/system I have allowed pyramiding of positions when there is a second, third, etc. trading 'signal' in the same direction. You would not have this occur in a moving average crossover system for example, where the system is EITHER long or short; versus a system that could have 2-3-5 or more signals in the same direction, each suggesting long or short entry.

The next charts show what the RSI system (with the different RSI 'trigger' values) looks like applied to the hourly charts of QQQQ and NDX through 9/1. A green up arrow below the hourly bar is when the last long trade was taken. How many positions are open is below the arrow; e.g., +1 (1 position or long 100 shares), +2, + 3, etc. A red down arrow is above the hourly bar at which the last short signal was generated and above the arrow is the number of open positions that were generated by EACH short signal; - 1, -2, -3, etc; i.e., hourly closes went above, THEN below, the upper RSI threshold multiple times.

Seen below is TradeStation's "Performance Summary" for the QQQQ trading system through today (9/1), reflecting a total $10,578 net profit for closed out trades. The system is currently long 100 shares. You'll note above that the 21-hour RSI has fallen to 39.5. If this indicator dips to under 39, when it comes back up above 39, it will generate ANOTHER buy signal and would be then long 200 shares. If the RSI after that dipped to below 39, when it came back up above 39, it would trigger another 100 share buy under the 'pyramiding' rules I've set up for this trade strategy.

Seen above, the Performance Summary indicates that the percent of closed out profitable trades was 64.8% and that the average losing trade was several hundred dollars more than the average winner. Since there were nearly twice as many winning trades as losing, the system produced a net gain (of $10,578) to date over a nearly 3 year period.

The 'maximum intraday drawdown' seen above says that at one time anyone trading this system would have been down $21,610 from the start point and that up to 2900 shares (not 'contracts') was held at one time as a result of taking multiple positions (pyramiding). AND, if you can stand having had a maximum of 15! consecutive LOSING trades for the period tested, this system is a winner. A string of such losses is a result of a whipsaw period with recurrent fluctuation of the indicator used to above and below the trade trigger level.

My last chart shows the Nasdaq 100 (NDX) index on an hourly basis with the 'best values' indicated for it as indicated with the below noted trading rules. As with QQQQ, there are no exiting stop points per each trade and the system is long until its short and vice versa. Multiple positions are allowed, so it is assumed that positions are added as additional signals are generated.

Note below that if the 21-hour RSI falls below 31, when it then rises back ABOVE 31 the system would indicate closing out puts ('short' positions) and buying calls (going 'long'). Until that triggering event, this system remains in NDX index puts.


Using the NDX strategy applied to the above chart, with history going back the same 1000 calendar days, results indicated for CLOSED OUT trades from 12/22/06 through 7/24/09 (with a multiple short position still OPEN) showed 55 wining trades, versus 21 losers: 72.4% were profitable!

The total net NDX realized point gain to date (9/1) was 1,197 points. An account trading this system had to sustain a 'maximum drawdown of 4,212 points in the underlying index but spread across multiple positions as the strategy led to price averaging a number of calls or puts during different periods.


Although the terms are used somewhat interchangeably, a system is 'mechanical' in that it has pre-set buy and sell trade entry 'signals' so has an objective set of rules that can be quantified; e.g., buy when a certain moving average crosses above/below another specific moving average.

When you want to 'overlay' a trading signal with your own judgment call, then it's no longer what is thought of as a trading 'system' unless further back testing and development of the trading rules suggest a new trading rule(s). A trading system takes the emotional component out of the trading equation.

You may well know that, as a trader, you are pitted against some quick reacting fellow traders that seem to place orders in a heartbeat. Ever wonder why the order flow at market turning points is so fast? Besides the fact that there are a number of sharp professional traders (e.g., so-called 'scalpers') who trade for small index option moves, there is also an order flow from trading system 'signals' that are generated in an instant.

Suppose you sense that market momentum MAY be starting to shift. Sooner than you sense it, let alone react to it, buy or sell orders are streaming in. Ever wonder why? Well, one reason is the number of computerized trading systems that are being used.

Technical trading systems are a product of the computer age, in that computer power is what has allowed System Testing And Development which I refer to with my acronym 'STAD'. Trading systems are typically composed of a set of entry and exit conditions that are set up in specialized software. The most widely used software for this purpose to my knowledge is TradeStation, a company long in the forefront of this activity. Moreover, there are a substantial number of TradeStation user groups out there that meet in person or in online forums and swap ideas.

When was the last time you met to share trading ideas with a like-minded group of fellow option traders? Probably never, but these groups are out there and are very dedicated to constantly improving their trading systems. There are also books on devising and using trading systems; e.g., by Martin Pring. TradeStation (the company) has developed some of their own resource materials; for example, basic booklets on designing trading systems. The company has evolved into TradeStation Securities, an online broker.

The basics of what 'systems trading' is will be the subject of this column and one more. While learning how to set up trading systems takes the right software and some dedication, the basics are not all that difficult given some dedicated study and work, particularly such tools as optimization. This subject is something that well-informed option traders should at least know is out there as its part of the 'competition' so to speak.

System entry and exit conditions must be 'back tested' on historical market data to see how profitable these 'trigger' conditions for entry and exit would have been in the past. Testing then needs to be redone over time. Back testing in turn, allows refinement of the technical rules and is another key part of STAD. Without computer applications to handle this, trading systems could have not have evolved as a popular means for systematic trading.

Finally, use of the optimization method going forward must be monitored to prevent a serious drawback of the systems approach, that of 'curve-fitting', which is finding a set of rules that worked perfectly with the benefit of hindsight and past events but will NOT necessarily be effective going forward in new types of market conditions and cycles.

I will concentrate on demonstrating some basics of trading systems and not so much on the pitfalls and shortcomings of developing and using trading systems. There are strengths and weaknesses in the 'systems approach', but this discussion can be left to a specialized study of trading systems should you become interested in applying technical analysis in this manner.

The obvious appeal and a strength of trading systems very much relates to the common investing and trading pitfalls: lack of a plan and discipline in carrying out a plan that includes rigid risk management principles and the difficulties in taming negative emotions like fear, greed and 'fantasy' or seeing market conditions in a way that reflects a bullish or bearish bias rather than objectively what's happening in the market.

The software that allows trading system development and testing doesn’t look that different from ordinary charting applications. There is a real time (or end-of-day) data feed, which charts and applies chart markings like trendlines, etc, indicators and other studies like fibonacci retracements. The similarity stops here since there is a vast amount more involved and which makes the software a relative memory hog. The faster computers and more memory (RAM) are very desirable.

Trading systems can be broadly broken down into ones that use:

1. technical indicators

2. chart patterns

3. a combination of both methods.

The validity of each approach stems from a basic principle of technical analysis: the knowledge that market cycles repeat and are identifiable. The sample systems I employ and I described recently might be 'applied to' or tested on intraday data but are not any different in construction than ones applied to daily or weekly charts although the input values for indicators will vary. The concept is that systems calculate a certain number of 'bars' or trading periods, whether this is 5, 10 or 60 minutes, a day or a week.

At the risk of trying your patience, I'm going to write one final systems trading article (part 3) that just goes in some depth into the basic TYPES of trading systems: indicator-based, ones using certain chart patterns and ones using a combination of both methods.