You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an alternative browser.
You should upgrade or use an alternative browser.
Chabah on Automated Trading
- Thread starter Chabah
- Start date
OK good, thanks for the feedback everyone. It is true we are about to turn the corner and get into code and stop being so theoretical. The reason for the delay is that I am awaiting my TradeStation account to be opened so I can access the platform and start building and testing the system.
I love the golf metaphor comment. I acknowledge it is largely true, although a key difference is that rather than having a $20k trading station set up, I am working on a shoestring. I have plenty of obstacles that I am working to overcome.
One example is that I will initially have to avoid pattern day trading because my account isn't large enough (on TradeStation). My (larger) brokerage account is locked inside a 401K account administered by AmeriTrade, so I have to manually place my orders and can't use certain types of orders. Surmountable, but it takes some planning and adjustments.
With regards to stop losses, I have one more largely theoretical lesson to write up, and then it will become clear how the system handles volatile stocks. Essentially, we have a portfolio of unwinding strategies that we use depending on a number of factors. It's pretty cool - I will post within a day or two.
Anyway I am glad to see that there are some readers. This next week we will start featuring actual trading strategies and code and by the end of October we should start having some backtesting results.
Thanks for the comments, more coming soon.
Chabah
I love the golf metaphor comment. I acknowledge it is largely true, although a key difference is that rather than having a $20k trading station set up, I am working on a shoestring. I have plenty of obstacles that I am working to overcome.
One example is that I will initially have to avoid pattern day trading because my account isn't large enough (on TradeStation). My (larger) brokerage account is locked inside a 401K account administered by AmeriTrade, so I have to manually place my orders and can't use certain types of orders. Surmountable, but it takes some planning and adjustments.
With regards to stop losses, I have one more largely theoretical lesson to write up, and then it will become clear how the system handles volatile stocks. Essentially, we have a portfolio of unwinding strategies that we use depending on a number of factors. It's pretty cool - I will post within a day or two.
Anyway I am glad to see that there are some readers. This next week we will start featuring actual trading strategies and code and by the end of October we should start having some backtesting results.
Thanks for the comments, more coming soon.
Chabah
So I got TradeStation up and running, and it looks pretty decent. I deposited $7k for now - again I won't actually be trading this account until December (manually) and January (automated). They started me off with $28k intra-day buying power, which was a surprise since my account is not pattern day trading yet, but that's fine.
Note that I am running the system on an Intel Mac with a Core Duo chip running at 2 Ghz with 1 gig of RAM allocated to Parallels. One slight annoyance I found is that I can't use all my screen real estate - my screen is just about 1600x1024 pixes. Windows XP Home supports 1268x1024 or 1600x1200. The latter cuts off the menu bar and start bar, so while it is a ton of space it is unusable. The 1268x1024 looks pretty good but does not use about 6 inches of horizontal space.
My buddy came over to test his system. He normally runs on a Dell laptop with a 1.8 Ghz Code Duo and this particular test runs for about 13 minutes. On my system it also ran in 13 minutes, although it *reported* running in half that. I don't know why the clock was messed up, but we were timing with a real watch so we know it took about 13 minutes. Therefore, running in Parallels on a 2 Ghz chip is about the same as running on a Windows straight on a 1.8 Ghz - definitely fast enough for normal use.
The challenge now, aside from learning the interface (the online tutorials are pretty decent), is to figure out how to implement my system concept within TS's framework. I'll go ahead and post my updated system plan this week so you have an idea of the adaptations I am applying. If you have any suggestions, fire away!
Note that I am running the system on an Intel Mac with a Core Duo chip running at 2 Ghz with 1 gig of RAM allocated to Parallels. One slight annoyance I found is that I can't use all my screen real estate - my screen is just about 1600x1024 pixes. Windows XP Home supports 1268x1024 or 1600x1200. The latter cuts off the menu bar and start bar, so while it is a ton of space it is unusable. The 1268x1024 looks pretty good but does not use about 6 inches of horizontal space.
My buddy came over to test his system. He normally runs on a Dell laptop with a 1.8 Ghz Code Duo and this particular test runs for about 13 minutes. On my system it also ran in 13 minutes, although it *reported* running in half that. I don't know why the clock was messed up, but we were timing with a real watch so we know it took about 13 minutes. Therefore, running in Parallels on a 2 Ghz chip is about the same as running on a Windows straight on a 1.8 Ghz - definitely fast enough for normal use.
The challenge now, aside from learning the interface (the online tutorials are pretty decent), is to figure out how to implement my system concept within TS's framework. I'll go ahead and post my updated system plan this week so you have an idea of the adaptations I am applying. If you have any suggestions, fire away!
System Overview (Updated)
The system consists of the tools and rules used to govern trading, providing a framework for specific trading strategies. The primary element of the system is the Model.
A Model is a self-contained trading system combining internal logic with external Research, Entry, Unwinding, Checklist, and Filtering strategies. Each Model has its own trading funds and time frame, and can be considered an independent account to some extent. A given Model may represent a logical market view or strategy, but the goal of the system is to create the most profitable combination of Models, regardless of their logic or internal components. Models have a set of common functions and components, including:
Position Table: Table containing all current positions and related information
Orders: Table containing all current orders
Signal Table: Table containing active signals and related information
Research Strategies: List of currently available Research strategies
Entry Strategies: List of entry strategies, primarily used for position sizing
Unwinding Strategies: List of Unwinding strategies
Checklist: List of rules governing trades
Filters: List of filter mechanisms used to block or modify trades
The system runs multiple models, both through back testing and live trading, to determine the success as measured by the profit factor. Multiple models are then combined to achieve the desired mix of risk and return (much more research to be done in this area).
_________________
For those of you running systems, my Model is equivalent to a large TS Strategy, capable of buying, selling short, covering and closing positions based on a number of conditions. The Research Strategies are similar the the simple entry/exit TS strategies.
Again it will be some work to fit all this within TS while maintaining a coherent design.
Enjoy,
Chabah
The system consists of the tools and rules used to govern trading, providing a framework for specific trading strategies. The primary element of the system is the Model.
A Model is a self-contained trading system combining internal logic with external Research, Entry, Unwinding, Checklist, and Filtering strategies. Each Model has its own trading funds and time frame, and can be considered an independent account to some extent. A given Model may represent a logical market view or strategy, but the goal of the system is to create the most profitable combination of Models, regardless of their logic or internal components. Models have a set of common functions and components, including:
Position Table: Table containing all current positions and related information
Orders: Table containing all current orders
Signal Table: Table containing active signals and related information
Research Strategies: List of currently available Research strategies
Entry Strategies: List of entry strategies, primarily used for position sizing
Unwinding Strategies: List of Unwinding strategies
Checklist: List of rules governing trades
Filters: List of filter mechanisms used to block or modify trades
The system runs multiple models, both through back testing and live trading, to determine the success as measured by the profit factor. Multiple models are then combined to achieve the desired mix of risk and return (much more research to be done in this area).
_________________
For those of you running systems, my Model is equivalent to a large TS Strategy, capable of buying, selling short, covering and closing positions based on a number of conditions. The Research Strategies are similar the the simple entry/exit TS strategies.
Again it will be some work to fit all this within TS while maintaining a coherent design.
Enjoy,
Chabah
Quote from uncleTom:
What are going to do about the screen issue?
Can you what kind of test takes 13 minutes?
Thanks
The screen isn't really an issue, it is more of a shame. Eventually I may get a bigger screen (something that supports 1600x1200), but that is certainly not a priority. 1268x1024 looks good (the iMac screen is very high quality). Also note that since I am focusing on automated trading, I shouldn't lots of windows open at a time (too many open windows is actually a pet peeve of mine anyway).
The test was an optimization of my buddy's trading strategy across 3 inputs for a single stock, about 4,000 tests. I recognize it wasn't very scientific, but I was just looking to rate my system performance versus his, and "about the same" is convincing enough to go with Parallels rather than buy a new Windows box just for TS.
Thanks,
Chabah
The "test" was an optimization. That function runs a strategy using a range of inputs. We optimized for 3 inputs, which created a total of just over 4,000 individual tests.
The point is that the processing power on my system was about the same as the processing on his system, so running through Parallels on an Intel Mac does not significantly degrade performance when compared to a PC. That's what we expected, but I had to verify the result.
Thanks,
Chabah
The point is that the processing power on my system was about the same as the processing on his system, so running through Parallels on an Intel Mac does not significantly degrade performance when compared to a PC. That's what we expected, but I had to verify the result.
Thanks,
Chabah