Lets build a trading system

Originally posted by opmtrader
Someone mentioned earlier the Pardise Trading system at stockhits.com. I recently completed coding this system in Wealth-Lab and would like to share some of my experience with it. To explain a bit the Paradise System mainly looks for reversals indicated by various candlestick patterns. It took a while to code as there where many refinements one must make to get the pattern recognition right. Anyhow lets cut to the performance:

Testing on QQQ 1 min data for all of 2001

Before commissions and slippage the system returned $10,000 profit on a constant 800 shares of QQQ (1 emini contact):)

After commission and slippage the system returned a -$30,000 return during the same period :(

I also put the system through a number of different filters, extremely high trend values, extremely low trend values, etc. but the results were relatively the same.

While one might think that the system is completely useless based on the results I have actually been using it with some success purely as an indicator for discretionary trading. Being a pattern recognition system it basically alerts me in real time various hesitations going on in the market. I then use this input combined with S/R and other inputs to help make my decisions. While some may have a problem with this approach it works for me.

Good thread. Keep up the good work everyone.
More...
i looked into that system as well. i haven't really tried it, but if you say it is a good indicator, but bad system, how could you change it to make it a good system?
 
I am sure I coded it correctly. It took me a while until I refined the criteria for the pattterns enough. They now interpret things as accurately as you would visually.

Admittedly I coded this system when the stockhits thread was in its infancy. While I had the full disclosure on the system I'm sure there are probably some refinements that have been posted that may be useful.

For those of you interested in patterns you should really look at Boop's Bar Analysis System on Wealth Lab. This guy did a fairly robust analysis of some 3 bar patterns and chose those he felt performed better than random. After testing his system on a broad index of stocks with daily data I found the results to be unimpressive.

I'm not sure what a lot of this means. Are the best known patterns all garbage? Does it depend when and where they are applied? Are there other larger factors we haven't recognized that effect the outcomes? Probably a bit of everything. I just found it interesting to find a tool that alerts me as to when a pattern that may be important to a large number of traders appears in the market. It may not be inherently meaningful but if it apparently meaningful to enough people at the same time than it may be useful.
 
Originally posted by opmtrader
I am sure I coded it correctly. It took me a while until I refined the criteria for the pattterns enough. They now interpret things as accurately as you would visually.

Admittedly I coded this system when the stockhits thread was in its infancy. While I had the full disclosure on the system I'm sure there are probably some refinements that have been posted that may be useful.

For those of you interested in patterns you should really look at Boop's Bar Analysis System on Wealth Lab. This guy did a fairly robust analysis of some 3 bar patterns and chose those he felt performed better than random. After testing his system on a broad index of stocks with daily data I found the results to be unimpressive.

I'm not sure what a lot of this means. Are the best known patterns all garbage? Does it depend when and where they are applied? Are there other larger factors we haven't recognized that effect the outcomes? Probably a bit of everything. I just found it interesting to find a tool that alerts me as to when a pattern that may be important to a large number of traders appears in the market. It may not be inherently meaningful but if it apparently meaningful to enough people at the same time than it may be useful.
More...

I just went over the Traders Paradise with opmtrader and he has it coded differently then the one that we are looking at. He only has a small portion coded. This has probably changed and evolved since he did his coding for it.

I think that we need to continue to work on this as I believe that we are moving in the right direction.

Especially with the lagging entry similar to what I described in the beginning of page1

wdbaker
 
rcreal,

What settings are you using for your test? I ran the system on each individual contract back to the March 2000 contract. I used 5 minute data and deducted a tick slippage and $3 in commissions from each side. It only made money on two contract periods (and a measly $3 average trade at that). My equity curves looked nothing like yours. I've always been amazed at how I can test something and get completely different results as someone else testing the same thing. Although I am guessing that you didn't deduct any costs in your test.
 
The system was tested on TS 6 data (@ES.D) from 2/11/2001 to last Friday (no commission or fees deducted).

Again, this is just a baseline to start with. The code was contributed to a discussion thread at Tradestation World:

https://www.tradestationworld.com/discussions/topic.asp?ARCHIVE=&whichpage=2&TOPIC_ID=501

I simply gave an example of a system someone suggested (beginning of thread) and improve it to where it could be improved to *possibly* be traded profitably.

If you take $25/slippage & commission per r/t a ctx, you still get around $15k for the period shown. Not great, but it's better than most canned strategies in TS :)
 
You must log in or register to reply here.
Back
Top