Machine designed strategies. Do they work?
- Thread starter v75z52
- Start date
Nope. Marketing is commercial activity to support selling a product or service.
Uploading some program for free download on your website, and talking about it has nothing to do with marketing as long as you don't go commercial or charge money for it. Besides, I was here only answering the question if I use such a program.
Uploading some program for free download on your website, and talking about it has nothing to do with marketing as long as you don't go commercial or charge money for it. Besides, I was here only answering the question if I use such a program.
Like I said earlier, I personally have no issues with marketing (although this forum has rules against it). What I worry about is the conscious or unconscious skewing of objectivity due to competing interests. This is why every scientific publication must disclose the author's funding sources. It doesn't matter if the product is free. If you generate huge publicity, you will be making a lot of money regardless. In fact promoting "free" things is the dominant model for marketing internet or software related businesses.
Unfortunately, scientific publications usually don't disclose the author's funding sources. And most people are not objective in a publication or discussion anyway - even without funding. They are just usually in favor of their own ideas. I am certainly no exception. Still, discussions can be useful and ideas can be exchanged. And I can assure that no one funds my posts here.
Quote from jcl:
Unfortunately, scientific publications usually don't disclose the author's funding sources. And most people are not objective in a publication or discussion. This is just natural and has nothing to do with funding. They are in favor of their own ideas. I am certainly no exception. Still, discussions can be useful and ideas can be exchanged. And I can at least assure that no one funds my posts here.
Most reputable journals require you to disclose potential conflicts of interest, such as being funded by an oil company when writing about lack of evidence of global warming. Yes people are not objective and they all advocate for their own ideas. However, when you attach financial interests, then it becomes a whole different ball game. Your unconscious willingness to skew the facts become much much larger. You may not be funded directly by your current activities, but I believe you expect at some point in the future to have a return on your investment (even subconsciously). You may say it's a hobby now, but I am sure you will not refuse several million here and there if your software generated enough buzz in the future.
Fair enough I guess, so long as you "don't go commercial or charge money for it".
I like your filtering ideas.
It's not often you see a fisher transform in an off the shelf package.
I have used several and by and large the results are due to a curve-fit. Systems may stay profitable for a short period of time but this is random. Eventually they break down big way. The only program that may avoid curve fitting depending on how you use it is Price Action Lab. This one uses a different philosophy based on finding OHLC patterns. The nice thing about it is that it doesn't serve a finished system but allows you to see the code and either use it to build strategies or analyze the patterns within the context of prior market action. This is a different philosophy with which I agree because I do not believe an automated system is the answer.
The Phds work on pricing models for synthetics and on arbitrage. They rarely get involved with system trading. Some hedge funds use Phds for system development, like for models involving cointegration and related stuff. Big firms move size, they do not need trading systems. Those that try to predict what the big firms will do need the systems. I hope it is clear now.
How does Price Action Lab avoid curve fitting? From their program description, it seems to be the textbook example of curve fitting - it apparently verifies the price patterns using the very same data in which those patterns were found. This way of course you can find any pattern, even in random data. Or does it use some OOS test?