Pair Trading Strategy Journal

Quote from total_keops:

No backtest, none.
I trade mostly discretionarily so a backtest does not really make sense as it includes many trades I would not have taken. It is good if you trade fully automated.
I guess many highly profitable trades occured in the last year and could skew your average win in $ to the upside.
More...

You already do your analysis in MATLAB, why not automate your system and trade in MATLAB? I guess everyone has their own style.
 
Quote from academic:

I think you're right. I'm looking at a standard deviation that's calculated over a moving window of about 2 years. When the spread pushes past 2 stddev, it doesn't catch up so much as it does for you guys.

Obviously, I'm looking at pairs with a long term cointegration (years) and I'm looking to hold onto trades for days up to months.
More...

Yes, I don't want to knock any of the good will going here, but using moving averages or bollinger bands to compute any of this stuff, completely defeats the idea behind looking for a cointegrated series. The idea is that you are assuming the cointegration will continue into the future. Using MAs as it is being applied is akin to looking for entries and exits using 'TA' stochastic oscillators (which is not good and also why you see the signal 'hugging' or railing against 2 sigma even though it should have told you the relationship has broken down and exceeded 3 sigma already).

If you use the approach I suggested, you will know exactly when your cointegrated series stationarity has broken down (quantitatively not qualitatively). Which is how it's meant to be applied.
 
Quote from academic:

You already do your analysis in MATLAB, why not automate your system and trade in MATLAB? I guess everyone has their own style.
More...
On any given day I see about 20-30 pairs poping in my scanner and end up trading between 0-3. A purely mechanical system is very hard to build and there are many things to look at.
Look at jonny (or PTF users), they use correlation which is not as good as cointegration and they still can make money by making a judgment call. I would says that a purely mechanical system is more a science and trading non systematically is an art.
 
Quote from dtrader98:

Yes, I don't want to knock any of the good will going here, but using moving averages or bollinger bands to compute any of this stuff, completely defeats the idea behind looking for a cointegrated series. The idea is that you are assuming the cointegration will continue into the future. Using MAs as it is being applied is akin to looking for entries and exits using 'TA' stochastic oscillators (which is not good and also why you see the signal 'hugging' or railing against 2 sigma even though it should have told you the relationship has broken down and exceeded 3 sigma already).

If you use the approach I suggested, you will know exactly when your cointegrated series stationarity has broken down (quantitatively not qualitatively). Which is how it's meant to be applied.
More...

If they're making money using MAs, that's great, but I agree with you, I thought it was a bit strange to be applying them as these guys seem to be doing. Specifically, I don't know why they would be looking at an MA and std deviation that is calculated on such a short window. It's more logical to be calculating standard deviation over the same period that you test the pair to be cointegrated.
 
Out of CSX/CNI.

Still suffering with NVLS/KLAC...but the time stop will kick in next week.

With CSCO's earnings out of the way, playing some long ORCL against it again.
 
Quote from Don87109:

I use the combo order everyday with good success.

Best way to create a combo is to drag each stock from the main screen to the combo window.

You can create ratio orders although you must use whole numbers. You cannot say buy 1.5 of stock A and sell 1 of stock B. You must say buy 3 stock A and 2 stock B.

And you can create limit orders which will execute when your predefined price is reached. You can even set alarms on a combo order, but there is a TWS bug that causes you to re-enter the alarms every time TWS is started.

In fast moving markets, like at the open, the combo order can be slow in executing even if your price has been reached. At most other times it works pretty well. BTW, I have submitted conventional Market orders at the open that have taken minutes to execute so the problem is not exclusive to Combos.

Don
More...
So I'm going to give these combo orders another try. The problems I mentioned did in fact happen right at the open, so perhaps I was being a bit harsh. My reason for change is simply to get in and out at a predefined spread limit price, rather than accepting whatever happens at the open or the close.

Intraday spread variation on these pairs seem to be relatively volatile at times, which can allow for nice exits and entries.

Thanks for casting doubt on my opinion Don. :-)

Adrian
 
Not to sound like a broken record, but today PTF has a buy in the stock ETE at the price of 25.47, which was at best a fluke trade prior to the regular market opening. These kind of "trade recommendations" seriously distort the perceived performance of their system. When you are investigating the the prior performance of any pair, it is very difficult to go back 6-12 months and filter out these kind of fantasy trades.
 
PTF uses whatever trade data Yahoo posts. Yahoo does correct errors as the day goes along so that past performance runs generally show the correct data and can be used for relative comparison purposes.
 
You must log in or register to reply here.
Back
Top