Quote from mlsignups:
First, Impactopia was the site I knew of but couldn't find today. I think registration is (or was) free when I registered some time back.
Second, this has been a great journal and I don't want to get anyone sidetracked so i won't go down this path too much longer if it is not needed or desired.
I did some interesting stuff with AmiBroker today. It is not designed out of the shoot for pair trading but it is easy to compare tickers and/or display them and ratios. I was able to recreate the charts that show up with a pair in pairtradefinder. I was actually able to recreate the entry and exit signals of most of the trades (not sure if i can catch them all) but out of deference to the folks at pairtradefinder won't go into what I found.
As I mentioned in my last thread i purchased the software (pairtradefinder) and intend to use it, so this is not a knock on them but I thought AmiBroker might help me refine my thinking as follows.
Assuming I have recreated the entry and exit signals I can now back test against a large basket of stock pairs at once for P&L. I can also now add some other criteria. For instance, since we want the correlation to be going up, I could add as a criteria to my backtest that I want correlation to be > 90% and that I want correlation to be greater than an x day moving average of correlation (so I know it is on the upswing). I could also add that the correlation for the last x days is > than x. There was also a post early by another contributor who talked about RSI of one stock being above 40 and another below 60. I can now add that parameter and backtest that to see how the results (P&L) change.
So my thinking is I can take what i've learned from their system and test various permutations of add-on criteria to see if any of these criteria improve the results.
IF I figure some of this out my thinking would be I can then use pairtradefinder during the day to identify the entries, but would then have more information about which ones to select for entry.
As I get time over the next few weeks i'll post some findings unless Johnny says this is too far off track for his journal.
Mike
Hi Mike,
I will be interested in what your find. I tried to setup some pairs in Wealthlab and an old version of Tradestation to test across a portfolio, but I haven't been successful. I have PTF , but the thing with PTF is that I can go thru and find high correlation and high average trade and export to a spreadsheet then analayse the trades across a basket of pairs, but of course the report looks fantastic because I've stated off with the cream of the pairs. There are plenty of pairs with high correlation but negative average trade so they are the ones I'm interested in to see why they didn't work.
I like PTF as it's simple, but now I have no choice but to paper trade for months as I can't really backtest a portfolio (coz I'm dumb when it comes to code !
regards
Ivan
!
)... that would require knowledge of how to code in C++ plus code maintenance would be required over time too..