TREND PREDICT ( Trade Station )

[mohsinqureshii 10]

Trend Pattern Prediction System

 

If a discretionary trader expects the market to rise tomorrow, he would probably look for opportunities to be a buyer. Similarly, if he expects the market to drop, he'll probably try to short the market. Oddly enough, trading systems rarely work this way. Generally speaking, trading system don't try to predict the market. They merely try to follow it. In fact, it's usually considered counter-productive to have an expectation for market direction when trading a system. Whether your personal prediction is for an up trend or a down trend, you're usually well-advised to follow your system, regardless of whether it goes long or short.

 

This doesn't mean there's no role for market prediction in the world of trading systems. A common technique is to predict the price change N days forward. If the prediction is for the market to be higher in N days, you would buy today and sell in N days. Likewise, if the prediction calls for the market to be lower in N days, you would sell short and cover in N days. If the prediction indicates the market will not change enough to exceed transaction costs, you would stay out of the market. The prediction itself can be achieved via any one of a number of methods, one of which I'll develop below.

 

While this approach may be profitable, it tends to emphasize the prediction part of the strategy over trading system logic. Most profitable trading systems probably succeed in large part because of the way they combine entry and exit rules to best capture profits while minimizing losses, following the classic dictum "let your profits run, cut your losses short". Moreover, certain entry rules tend to favor specific market conditions, where particular exit rules tend to work best. Simply buying today and selling N days from now because the prediction calls for a higher market in N days neglects to take into account the basic tenets of trading system design.

 

An alternative approach, which I'll develop here, is to predict a specific tradable pattern and, when found, apply a set of trading rules specifically tailored to that pattern. In this way, the prediction logic is combined with trading system logic, giving us the best of both. I'll illustrate the approach using a trend pattern, but other patterns, such as reversal patterns, could be used as well.

TRENDPREDICT.ELD

trendpredict.txt

[BlueHorseshoe 10]

Hello,

 

An interesting post, thanks!

 

I have seen Bryant use the inbuilt optimisation capacities of TS in a similar way before in an article on Neural Nets. The spreadsheet sorting aspect of the process can also be conducted within tradestation using arrays.

 

Cheers,

 

BlueHorseshoe

[AndrewJackson 10]

mohsinqureshii ,what are the settings? Tried to reference more bars than allowed by the current Max bars error.Thank you

[schris87 10]

You have to change the # of bars back to reference by doing the following:

Format Strategies->Properties for all -> Maximum bars study will reference (bottom left) and change to higher value (I usually have it between 300-500)

[AndrewJackson 10]

Thank you schris87, do both strategies work together.This is what i got shown in picture.