Developing a Trading System Step by Step

[cnms2]

Take care!

Quote from Murray Ruggiero:

Sorry, but I really have not felt well the past week, so I am struggling to get my work done and it has not left a lot of time for Elite Trader. I hope to be caught up by tomorrow or Friday, so I can spend some time and finish this course. I am sorry for the inconvenience.

 

[Murray Ruggiero]

Quote from Xephen:

When you see system vendors use slippage and bro of $50 per contract, or maybe you've done it yourself, DON'T BELIEVE IT!! Even liquid electronic markets sometimes have crazy gaps. Or your broker fouls something up. Or YOU foul something up. It's going to happen. So you need to be very conservative with those slip numbers. By the way, if I had not used a stop in the t bonds, I think I would have actually made money that day. So it goes ...

TradersStudio 2.0 will let you set slippage and commissions on a bar by bar and market by market basis. This means you could use a slippage of .20*Average(Range,10) for example. This way slippage adjusts for volatility

 

[swingman]

Quote from Murray Ruggiero:

I really don't believe in very short timeframes for trades. I might use 5 minute bars on a chart but I really want to see trades last ideally at least 3 hours.

Remember you really only make large profits on large range days. Your goal is to predict these days.

Okay, but have you known very successful traders/managers who scalp very short timeframes?

 

[BENG]

Quote from Xephen:

As for the currency issue, MOST of the time I just cheat and use a general average value embedded in the tick value. When I am REALLY serious about precision, I convert the data manually in Excel so that the daily prices are in USD per 1 contract of the non US market and set up the tick values etc accordingly. Basically, this makes it work like an equity in the system test. How do you do it, Beng?

I use the daily spot rate to convert the data, and I don't use any commercial program out there. Currently, I'm using a 4GL based program, which can do a lot of things. Just that it's not written in OO, so it's quite hard for me to change things.

 

[BENG]

Quote from swingman:

Okay, but have you known very successful traders/managers who scalp very short timeframes?

James Simons.

 

[BENG]

Quote from Murray Ruggiero:

This technology is very powerful and one of the many features we have added.

Can it run simluations based on individual front month contract, and be able to do the roll over itself? I know simulation with perpetual data is enough, but only simulation based on real individual contract can simulate the real world situation.

Data issues... I have huge problem managing data (perpetual, ratio adjusted, individual contract, roll over date, currency conversion, tick size, tick value, news date, etc), can TradersStudio make it easy for people to import data from sources like CSI, CQG, Pinnacle, etc. I might have experience with system building, but definitely not data management. Imagine I'm new to system building, and I buy data from CSI, I don't even know how to use it properly... based currency, tick size, tick value, these values don't stay constant in the historical data, and it's impossible for me to know what has happened in the past 10 to 20 years for every product.

About simulation... can TradersStudio do something like rolling window simulation? For example, I like to take the best cases from the last 3 years to run simulation for the next 3 months, then take the best cases again for the moving 3 years window, and run simluations for another 3 months. Simulations could get really complicated in some cases, how would TradersStudio tackle this issue?

 

[Murray Ruggiero]

Quote from swingman:

Okay, but have you known very successful traders/managers who scalp very short timeframes?

What is short term, What average trade length ? I don't know any professional money manager who only trade very short term scalping, maybe before the dot-com crash but not today.

 

[Murray Ruggiero]

Quote from BENG:

Can it run simulations based on individual front month contract, and be able to do the roll over itself? I know simulation with perpetual data is enough, but only simulation based on real individual contract can simulate the real world situation.

Data issues... I have huge problem managing data (perpetual, ratio adjusted, individual contract, roll over date, currency conversion, tick size, tick value, news date, etc), can TradersStudio make it easy for people to import data from sources like CSI, CQG, Pinnacle, etc. I might have experience with system building, but definitely not data management. Imagine I'm new to system building, and I buy data from CSI, I don't even know how to use it properly... based currency, tick size, tick value, these values don't stay constant in the historical data, and it's impossible for me to know what has happened in the past 10 to 20 years for every product.

About simulation... can TradersStudio do something like rolling window simulation? For example, I like to take the best cases from the last 3 years to run simulation for the next 3 months, then take the best cases again for the moving 3 years window, and run simulations for another 3 months. Simulations could get really complicated in some cases, how would TradersStudio tackle this issue?

We do not have built in support for individual contract roll over, but I know that we can do it in script using our portfolio capabilities. I just have not had time to write the code.
The currency problem for overseas futures goes away with TradersStudio 2.0 , just supply the conversion file that you create using CSI. When CSI updates it along with the rest of your data you will always have the correct conversion without fudging it.

In 1.3.6 we wrote a walk forward macro, this does walk forward analysis which is what I think you mean by rolling windows. This is included for free in the current version of the product. In 2.0 we are integrating this technology into the product.

 

[swingman]

Quote from Xephen:

By the way, I agree about the exit strategy. I have a dozen systems that perform really great as stop-and-reverse. But I have always struggled with building really good exits. I have learned a few good tricks over the years. For a hint, see my first post to this thread. Some things you hear as common wisdom are not true once you start testing them on real data.

Xephen (or Murray when you get better),
Do you mean a stop and reversal volatility breakout system? Please tell me more about the mechanics, and where do the stops come in, if any?

 

[swingman]

Quote from Xephen:

About the short side performance, I know the common wisdom is that trades should be identical on the long and short side, but it doesn't take much testing to see that markets behave differently in down trends than in up trends.

X
Xephen Merchant Partners

Xephen,
Yes, I've had problems with the short side in building a robust system. It always ends up DRAGGING or even CRUSHING a good long-side. Any advice or clues on how to rectify this? That would greatly be appreciated, as it's frustrating.

I've obviously been guilty of using the same criteria for long and short, only to find that the short side is not working. I know this, but can't figure out how to address it. One time even, by luck (don't ask me what I did), when I finally reduced the number of short trades and used a signal more beneficial to the short side, my long-side performance completely went negative! I gave up and moved on...but am thinking of getting back into automated strategies more seriously.