Nov
5

Formula 301 – #4 Basic Consensus System

I am going to present a system that takes multiple technical indicators as part of its decision making. By combining these indicators through a simple scoring system, we get a very profitable daytrading system that works year after year.

(Lawrence – this system was designed many years ago (before year 2000) and was originally created for a trading platform that was very popular at the time. I was going through some old notes I have and adapted the code into NeoTicker. I am very glad that NeoTicker can implement the same system in 1/10th the amount of code required to implement the system. The original money management code has to be stripped due to various reasons, including NDA restrictions.)

Classic Multiple Signals System vs. Consensus Model

For many system designers, trading systems are signals, filters, and money management combinations. Especially for signals, they are very precise conditions from the crossover of 2 moving averages, the turning down of an oscillator, the breakout of a key price point, to the combination with specific trend quality. That is a pretty classical view of what trading models can be.

A more model view of trading system is that market direction changes due to many possible reasons, and if we can approximate the situation and estimate with high confidence of possible change in underlying condition, we are getting very good entry signals already.

Implementation Techniques

The main technique for implementing consensus model, is the use of scoring system. For each indicator or indicator group involved, we can assign scores to the group to identify the bullishness or bearishness of those indicators.

After scores are given for each indicator group, we can sum up the scores into a combined consensus score.

For simplicity sake, we can simply take long positions when the consensus score is bullish and go short when the score gets bearish.

The advantage of using a scoring system is that we can forget about the exact entry point signals and focus on the correctness in the directional bias, which is a way more important component that makes the system profitable.

The System

The system utilizes 3 completely different class of technical analysis to generate its score. In general, a +2 is assigned to the condition that is deemed as extremely bullish, +1 is assigned to the condition that is deemed as somewhat bullish, and vice versa. For indetermined condition, a score of 0 is assigned.

Interested readers can check out the indicators used within the script.

Here is the complete script.

formula301 part4 basic concensus script

One useful trick utilized in the script is the filtering of the signals generated at the end of a trading day. Such signals are not desirable because they force you to take on a trade at the open of the next trading day, which could be affected by the overnight events that altered the scenario completely.

The complete system is available here, Basic Consensus System.

You can now take a look at its performance.

As usual, the system is a daytrading system that always close its positions by end of day. It is applied onto emini S&P 45-minute data. The time range is the full regular trading hours from 9:30 AM to 4:15 PM Eastern Time. Commission is set at $2.50 per trade to reflect the cost.

formula301 part4 basic concensus chart

The performance of this system is way better than any of its components.

Other Considerations

I call this system a basic consensus model because it is just a starting point.

Just think about the possibilities – we have not even utilized trend filtering, or even money management techniques. There are a lot of room for further improvement over this basic system.

You can also create your own consensus model using a completely different set of indicators and scoring rules.

End Notes

I do not expect this system to perform so well after it was dropped from real-time deployment due to its weak performance in year 1999. I think it deserves its spot in trading system history as it demonstrated the possibility of combining common indicators into a better trading model.

Lawrence

Discuss this article.

6 Comments on “Formula 301 – #4 Basic Consensus System”

  1. Jay Says:

    If I built and tested this system correctly, it looks like it’s had flat performance since 12/1/2004 to 11/1/2004, on 57 trades. I guess this period is similar to the flat period in 1999?

  2. Lawrence Chan Says:

    Jay, I assume you are talking about Nov 2004 to Jan 2005. One of the components used is Stochastics SlowK. The biggest enemy of oscillators is prolonged period of sideway market. The tight range behaviour across multiple days during this period caused those whipsaw trades. The interesting thing is, the drawdown during this period is still within the system’s normal volatility in its equity.

  3. Ed Says:

    Regarding your “end notes”. It’s hard to tell why performance was so weak in 1999. Would you explain further please.

  4. Lawrence Chan Says:

    The performance of the system in mid to end of 1999 is weak for a simple reason – the rules that define the stock market was broken. Serious investigations done after the melt down in year 2000 revealed many market participants, who traded during the end of the dot com era, are not following strict margin requirement, while their brokerages did not report or stop them from doing so. Thus prices was pushed around like unlimited amount of money is available. When strict margin requirement is enforced again, the system operates as expected.

  5. Ed Says:

    Just out of curiousity do you optimize to arrive at the parameters you choose or is through chart overservation or perhaps someother perspective. Thanks for any comments you might have.

  6. Lawrence Chan Says:

    No optimization applied. Most of the parameters are just common combinations. The scoring rules are also common interpretations of those indicators.

Leave a Comment

Blog Developed
By ContentRobot