Tradestation may be just fine but I think the emphasis of a new trader should be more on risk and money management rather than on automating trading strategies. Strategy automation is IMO a task for experienced traders and at that level there are better and more efficient solutions. Many new traders think automation is the key to success but that is a great fallacy. Understanding trading, price action and how markets work is the key and automation is a kind of a reward.
I program my own models in Tradestation for the last 12 years since before 2000I. But I am now testing them to Ninja Trader for a variety of reasons. I wrote my first commercial software in 1971. Many of the programming tools in that era were easier to use and test with than the software of today. In short the retail trading industry flunks programming 101 - especially Tradestation. Writing computer programs in todayâs retail trading industry is much different than the commercial world. Most retail traders have no idea how restrictive this environment is for programming: Some of the basic things are missing. Here are a few examples from a long list: Debugging Tools â Commercial programs put the trading industry to shame. In commercial programs you can step through code in each line of a program and watch the variables change in value. The âabnormal program endâ or âabendâ dump does not exist in trading software. The C# environment has many debugging tools that programmers who are traders desire. TradeStations idea of debugging is dump the variables data to print statements. Testing environment â I still get sick to my stomach when I see what testing is like in this trading industry. One optimization and we risk all our money. Poor by any standard much less ANSI Standard Program Language Conventions - Every trading language has similar statements with often totally different meanings and execution. For example run an in-sample optimization in Tradestation that produces, say 500 profitable optimization settings. Now try and move these settings to an out-of-sample test. It has to be done one setting at a time manually. Brokerage interface â Tradestation is still hung up on âBuy or Sell on the next barâ technology. Next bar technology can kill an automated trading system unless you know how to code around it. Plus the trader is locked in to TradeStationâs second rate Brokerage. Data interface â Tradestation nickel and dimes you for data cost. Hook NinjaTrader to your IB account and IB provides the data - Free Tradestation shut down its user conferences a few years back. They became tired of hearing the traderâs complaints about how backward their technology was. They have not made an effort to move forward to meet the competition in years. Just take a look how traders in ET rate Tradestation this should give you a clue. Now on the flip side for traders without an extensive background in IT, Tradestation does provide an integrated platform as you noted. Many complex strategies can be built and run there profitably. I have recommend Tradestation to many starting traders who do not have the capability to program extensively. I totally agree with intradaybill that focusing on risk and money management is far more important for starting traders than automated trading. My strategies are still executing on Tradestation every day. But I donât know for how much longer.
Thanks for the replies from intradaybill and Rabbitone. I agree with both of you that risk and money management is far more important than the strategy itself. I have played with money management(mainly position sizing) rules in my backtesting and saw how bad money management can damage a potentially good strategy. The major reason I need to automate my trading is that I am a working full-time and the strategy involves monitoring a basket of stocks for intraday signals. While I don't have extensive IT background, I do have an undergrad degree in computer engineering. So college level programming should not be a problem to me. Is my background sufficient to program something like NinjaTrader? I have been backtesting strategies using AmiBroker. Its own language allows me to wrote entry/exit signals, specify entry/exit prices, and position sizing. Since my trades usually lasts a few hours to a couple days, I don't really need and don't want to write highly customized high frequency trading system. So I already know TS would fulfill my need, but it's a little costly and I heard its brokerage is not that great. If I choose to use other automated trading platforms such as NinjaTrader or OpenQuant, what kind of "additional" programming will be needed besides just coding the trading strategies themselves? I totally don't have a problem with using a more sophisticated platform to implement the strategy itself(entry/exit, position sizing). But will I be required to do peripheral coding that's not directly related to the strategy itself? Such as writing some networking code to interface with the broker? The bottom line is, if NinjaTrader allows more sophisticated strategy/money management implementation than simply "buy on the next bar" without requiring too much peripheral programming, I think that would be suitable to me. Thanks for sharing your experiences.
ezbentley In addition to the software characteristics you might also want to consider customer support. Iâm also investigating platforms with a goal similar to yours and conveyed my requirements to customer support at both Ninja and Trade Station. The response was radically different. One difference between our goals is that I need an external interface to a custom application which generates my trading decision using non-linear models. I only need the platform to make and manage the trade (set and adjust stops and implement a simple mechanical strategy). To do this the platform has to pass current prices to my application and then listen for a signal to trigger a predefined strategy implemented in the platform but selected by the models. The folks at Ninja understood this perfectly and gave a prompt response including the following: ---------------------- Please see this section of our Help Guide, which explains our Automated Trading Interface and what you can do with it: http://www.ninjatrader-support.com/HelpGuideV6/Overview.html You can also use the External Data Feed: http://www.ninjatrader-support.com/HelpGuideV6/ExternalDataFeedConnection.html -You can use the MarketData() function on any ATI interface to send market data to another source. ---------------------------- The folks at Trade Station kept asking why I wanted to do this when I could use Easy Language to code my system. Well, I didnât spend over $30,000 on state of the art analytics software to reinvent it in TradeStation! It took several emails and a flowchart for their technical lead before they could understand the concept of an external driver for a decision to trade. In the end they said I probably could write some custom DLLs if I wanted to but they very werenât certain it would work easily. Jerry
Can you be more specific Jerry about this? Do you think there are security issues involved when the system is coded in the broker platform?
Bill, I'm not precisely sure what you are asking. To clarify my prior comments: I develop a decision to trade by predicting what the market will do in general terms a few bars into the future. This involves non-linear modeling of current market conditions using neural networks and related technology. This gives a direction and a general read on a scope of near term market behavior. So the decision to trade comes from a "data mining" application that has to be external to the trading platform like Trade Station. A sophisticated corporate level application like SAS that might have taken 20 or 30 person years and many tens of thousands of lines of code to develop is not something that could be converted to a script language like Easy Language in Trade Station. Software vendors of course donât give out source code but even if one had it, it would take years or dozens of people to convert it to another language. So I have to get the decision to trade external to Trade Station or any trade execution and management platform. The folks at Trade Station couldnât get their heads around why this would be a requirement. The attitude seemed to be that their product can do all things that need to be done and Iâm kind of off base in getting a decision to trade in any other way.. My concern had nothing to do with the security of Trade Station code. I donât know about that and donât really care as the trade decision would be generated by the external data mining application. All Trade Station would do is make the trade based on a simple mechanical strategy for the particular kind of trade determined by the external applications (adjust stops, manage profit target and exit after a predefined number of minutes if nether the stop or the target are hit). Simple stuff one can define in 20 lines of pseudo code. Jerry
I see what you mean now. I guess you are simply saying that traders like you with sophisticated non-linear algorithms developed in third-party software have little use for Easylanguage. It makes sense. Can I ask you which data mining application is that you are using?
I used to work in data mining in the banking industry (credit card fraud detection, loan default prediction, etc). I can't identify the specific software as I "forgot" to return or destroy all copies in my possession when I was downsized out of the company (in violation of my severance agreement). I don't need the legal problems if that got back to the software vendor or my former employer as it retails for something greater than the cost of a luxury SUV, and at that level things work differently than having an unauthorized copy of a $100 consumer software application where legal action isnât worth taking. However, the last time you used your credit card (depending on the brand) the transactions was likely evaluated for fraud with this application. One of the things we discovered with the neural networks was that one of the first things a credit card thief does with a hot card is buy gas from a gas station close to a big box retailer. If the transaction is approved he immediately goes to the big box store and buys some consumer electronics which can be quickly fenced. If the gas station declines the card then he drives off very quickly and tosses the credit card in the trash. So if you personally have a perfectly good credit card and you buy gas and then too quickly buy an identified category of consumer electronics and they check your ID, you now know why.
Interesting stuff Jerry, thanks for the info. Data mining is a fascinating field. Is the program you are using based on NN? Do you feed it with prices and values for several indicators or are you just looking for price action patterns? Intraday or daily data? How long of a price history do you input to the algorithm?
Yes, it has NNs as well as standard statistics for pre and post processing analytics. Many technical indicators have little predictive value, such as RSI (too much lag). A handful capture some characteristic of the market which can predict something. A lot of what I do is based on Landmark and similar pattern transformation methods originally developed to automate medical diagnostics. There are some striking similarities in the transitions in state dynamics in an EEG and a financial instrument, beyond the fact that they are both wavy lines! Intraday 15 to 60 minute bars in Futures and Forex.