Machine designed strategies. Do they work?

Quote from logic_man:

If a hedge fund offered me a few million bucks, I would explain my entire strategy to them and then we would agree to share the market in the assets where it worked. Since the strategy has an objective answer of when to trade, how to manage the trade, etc., I would easily be able to determine if someone else was using the exact same strategy.

You simply won't because of exchange proliferation unless it is something that is traded only on a single exchange. In forex you cannot do it at all since there are thousands of little exchanges or pools. In stocks there are so many dark pools too.

No offense intended but you are talking about millions and hedge funds without knowledge of the market microstructure. Disclosing your strategy to anyone is a huge mistake. If it is a good strategy with good timing they will trade it from multiple accounts and you will get paid only from one. You won't be able to prove anything even if you have statements in your hands. Litigation will cost you more than your profits. I have been in such situations several times. When it comes to money even best friends can become the most vicious enemies.

Keep this principle: In trading you cannot prove anything.
 
Quote from milewski05:

IMO i would never trade a system that's already 30% OOS but just to prove a point here. ill do 60% OOS 40% IS.

If you are still around I asked this before:

Quote from alexandermerwe:

Question: do you provide TSL with the combined IS+OOS to start with? Or you provide first the IS and when the system is ready you use the OOS just to validate it?

I am asking you because based on experience of 15 years with trading system development, it appears there is possibility of data-snooping with your results. But that cannot be confirmed unless you either provide the code of the system or the company provides a demo.
 
Quote from alexandermerwe:

You simply won't because of exchange proliferation unless it is something that is traded only on a single exchange. In forex you cannot do it at all since there are thousands of little exchanges or pools. In stocks there are so many dark pools too.

No offense intended but you are talking about millions and hedge funds without knowledge of the market microstructure. Disclosing your strategy to anyone is a huge mistake. If it is a good strategy with good timing they will trade it from multiple accounts and you will get paid only from one. You won't be able to prove anything even if you have statements in your hands. Litigation will cost you more than your profits. I have been in such situations several times. When it comes to money even best friends can become the most vicious enemies.

Keep this principle: In trading you cannot prove anything.

Certainly I'm sure your experience is valid, but if you look at what I said about setting up a trust to trade the strategy, you can get around some of these issues. Rather than charge a "price" for the software, you agree with the buyer to set up a fund, with some preset minimum amount of funding, and any trades executed through that fund count toward the profit-sharing. Then, even if the buyer is trading it elsewhere, you are still getting your payment. I don't necessarily need to get paid from every account the buyer trades it from, just the one which will be pre-funded with an amount that I think will ultimately provide me with the compensation I think is fair.

But, yes, overall I am just thinking out loud a bit here, not speaking from experience.
 
Quote from alexandermerwe:

If you are still around I asked this before:
milewski05 probably won't come back here. His visit was very quick and had specific purpose. A trader who can afford to spend USD 60,000 to buy software without testing it before is too large to fit in this place. :)
 
Quote from goodgoing:

milewski05 probably won't come back here. His visit was very quick and had specific purpose. A trader who can afford to spend USD 60,000 to buy software without testing it before is too large to fit in this place. :)

hmm... I hate to admit that you may be right...
 
I've started using the PAL demo with ML downloader data. I understand how to import data from MLdownloader to PAL but how do you upload part of the data in the Out of Sample data , so the in sample tests stop before that date ?

I noticed Tim changing the OSS data apparently easily... any one care to explain how does it work ? I haven't found the explanation in the manual either.
 
Quote from luisHK:

I've started using the PAL demo with ML downloader data. I understand how to import data from MLdownloader to PAL but how do you upload part of the data in the Out of Sample data , so the in sample tests stop before that date ?

I noticed Tim changing the OSS data apparently easily... any one care to explain how does it work ? I haven't found the explanation in the manual either.

http://www.priceactionlab.com/Literature/pal30manual/pal30manual.html#process32

Basically you must have two files, the IS and the OOS, stored at different locations. When the search using IS is done like in the above manual link then you click on Test Pattern and you select the location of the OOS. If the program finds the file the start and end date are displayed in the dialog box. Make sure the two files have the same name.

Edit: If you meant how to create the two different files, then just use copy & paste to duplicate the file and save it at a different location and then your editor to open the files and remove the part of data you do not want in there.
 
Quote from v75z52:

http://www.priceactionlab.com/Literature/pal30manual/pal30manual.html#process32

Basically you must have two files, the IS and the OOS, stored at different locations. When the search using IS is done like in the above manual link then you click on Test Pattern and you select the location of the OOS. If the program finds the file the start and end date are displayed in the dialog box. Make sure the two files have the same name.

Edit: If you meant how to create the two different files, then just use copy & paste to duplicate the file and save it at a different location and then your editor to open the files and remove the part of data you do not want in there.


Thanks ! I seem to have managed it following your instructions.
 
Quote from luisHK:

I've started using the PAL demo with ML downloader data.

How is MLdownloader working for you? I looked the other day at YLoader and it seems to be just fine for the job.
 
I dowloaded MLdownloader the same day as PAL so it's all very new but I found ML very easy to use with PAL. You can select and download whole index members in one click in ML and it takes a couple of click to transfer it to PAL . I had a quick look today into YL and it looks a little less convenient (but I may have missed the way to do... )
 
Back
Top