I have developed some code that runs in TradeStation for trading stocks intraday. My challenge is, because the way I am selecting my stock data set, I am only able to do live walk forward testing day by day, which I don't feel is very practical.
I would like to do a better job back testing with intraday data, and to figure out what may be a better process that I have now, to develop what may be a valid tradable strategy with positive expectancy to then potentially automate in real time using AMIBroker to accomplish this task with Interactive Brokers.
Does anyone have any experience or success in doing what I am looking to accomplish that they would be willing to please share? ie, suggested intraday data source to use for back testing, etc..
@algobuilder
I use data from "Lenz und Partner" in Germany for precisely this purpose.
They offer a .NET component. This .NET component is thread-safe, well-documented, and very comprehensive. For example, you can retrieve pre-compressed data. I use 15-minute compression, but any other compression works as well, as does tick data. Bid/ask prices, including volume, also work well.
You get real-time data for the USA and (almost) all of Europe. The rest of the world is delayed by 15 minutes. The data is (almost) all split-adjusted. Historical data goes back to 2002. However, the years 2002 to 2008 are not well split-adjusted. The .NET component is very fast.
I do it like this: I use a custom C# program to read the data with the .NET component. I store the data on SQL Server 2022 and then export CSV files. I then import these CSV files into AB via a scheduled COM interface.
Well... I've been actively trading this way for over 18 months.
I have about 2200 titles. Importing and saving them to SQL Server takes about 7 seconds with compression, and exporting to AB also takes about 7 seconds.
@algobuilder
But beware... "Lenz und Partner" also offers an AB plugin. This was developed many years ago for the 32-bit version of AB. It doesn't work at all.
Stay away from it. Spend your time on something else. It's better.
The .NET component, on the other hand, works perfectly.
Yes... it really works. I tested it down to a one-minute compression. The computer is an AMD Ryzen with 16 cores and 192 GB of RAM.
In my C# program, I limited the threads to 25. It always retrieves and overwrites the last (as yet unclosed) bar. Then it saves the current bar. Similarly, I import the data into AB via CSV. It speaks to AB's enormous performance that it's just as fast as the data retrieval (which now runs in 25 threads).
Although, here in Switzerland, I have a very, very fast internet connection: 10 GB down / 10 GB up.