I'm finding that PTF appears to have appalling customer service. I emailed them about a few issues I thought the app had. No reply. I even tried to buy it but they use Share-It and they wouldn't accept what they called a 'free' email address used with my PayPal account. I emailed PTF about it. No reply. Fine, if that's how good their support is, I have no qualms about running another month free trial using a different IP address I have.
A couple of interesting observations. I've now got 2 PTFs running (on 2 separate IP addresses) with the same pairs loaded and they come up with quite different correlations. One may be in a trade while another is watching. Ones natural inclination would be that both instances would be showing the same figures, especially as my setup for them both is the same. Also a couple of days ago I came across another app called 'Pairs Trade' which I downloaded a trial of. Its their 'Gold Edition', so I've no idea what their 'Bronze Edition' must be like because this one has hardly any functionality at all. And the interesting thing is here is that their pair correlations bear absolutely no relationship with those of PTF. So we must be talking about a different type of correlation altogether, or a correlation taken over very different time frames (though I think its the same). Anyway, it would be interesting to hear from those with more experience of pairs trading than I.
hi DR who, with the trial version of PTF is there any difference w/ the full version? IE funtionality
Hello guys! Thank you for a good thread. Sorry I interrupt your conversation, but I just wanted to hear a comment from anyone in the biz. I trade mostly stocks, but I used to be trading forex earlier.. When I came up to this thread and saw you guys looking for correlated stocks which intend to come back to their mean I remembered currency pairs.. So what I did is the following: * Exported H1 data since 2008.01.01 from metatrader GBP/JPY and AUD/JPY * Their relation is 97 procent correlated * As GBP/JPY / Aud/JPY = GBP/AUD I can follow this pair as a ratio between gbpjpy and audjpy * I took a MA and counted the spread of ma and high/low to see its average/max/min deviation and based on that I also programmed in excel that I would enter the trade (short/long) everytime rate will goes away from MA more than 80 points and close it as soon as it hits MA results are these: winning trades 870 loosing trades 322 winning average 44 points loosing average 90 points The whole thing produced pretty good results, I think period of backtest i long enough I want your comments on this... Can I use MA on GBPAUD as you use Mean on your ratio? Whats the period of the Mean you use (if its the same)? Waht do you think all in all about it Any other inputs... thanks D.
Hi Dr Who, I run multiple instances of PTF on different PC's (actually virtual machines) as it's dbase seems a bit stressed if I run everything in one instance. I also regularly see same pairs with different correl values across the instances. So for a given pair, one instance may show "watching" while another might show "in trade". The way I work around this now is that since the release which showed +/- I now sort on this and correl value rather than sorting my watchlist by the flags such as "enter trade". So I sort using +/-, % and correl and when I enter a trade I keep an eye on the +/- and eyeball the ratio chart to see when it returns to the mean. I don't really take much notice of the flags. I have all my open positions (typically 6 pairs) in a PTF instance with nothing else, so there is no clutter whilst eyeballing. I don't trade intra day as I'm based in Australia and just place market orders for the opening of the USA markets. Not sure how others are doing it, but so far, that's my angle and so far I'm still in profit ! I also hit the refresh button on any pair that I'm looking at sorting upon as that usually always changes the numbers by quite a bit for some reason. regards Ivan
Interesting approach Ivan. I'll take a look. Meanwhile, I think I understand why my figures are different for the different instances of PTF. My main PC has PTF open all day and therefore has intraday data whilst my other one only has PTF with EOD data. So I can understand why there will be differences.
Can you reset the database of both PTF and see if they still show different results? Best would to have an exact copy. After that I will be susprised if you get different results.
It's been on my to do list for a little while now. First, we dont trade a fix number away from the mean, we trade a fix number of stdev away from the mean (well, I can only talk for myself). The difference is that it takes into account for the "current" (recent past) volatility and it can also be applied to different pairs without looking back. Some pairs may need different fix number away stdev from the mean. Also, looking back and fixing the number is curve fitting. You may get different results with a different sample period. Next, one year of data is a small sample for currencies. You may want to sample from other market conditions. The last year was very volatile and produced very good opportunities in term of arb. Maybe the previous years where not as good. I did not test and can't tell really. I would also be concerned about periods like 98 when the Yen went crazy. But, I guess in real life trading you could apply a human filter over that and decide or not to take the trade based on fundamentals. But on the other side there are many feats that make it appealing. Less overnight gaps (only week end), liquidity, easy access to leverage (enough to hang you with). With forex pairs from what I can see with your numbers, you get a trade every 6-7 hours. Which seems very frequent when you consider that I get 1-5 good trades per day with all the pairs I can make with over 1500 stocks. If it's that good because it creates a lot of opportunities and you can spread your P/L distribution over many trades and not having to take too large positions on each. It's interesting.