Pair Trading Strategy Journal

[total_keops]

Quote from Trader13:


Now, if I were to abruptly wake you up in the middle of the night and ask you to explain the lag variable, you should be able to provide a complete answer without hesitation. Until you can do that, don't trade mean-reversion strategies.

I just did not really have the time (and still dont) to jump in this conversation to help but this is also what I think.

 

[total_keops]

Quote from Kevin_in_GA:

Having both is good but I want to know that the pair has a high cointegration NOW rather than over a long time frame.

There is no now, cointegration is always backward looking and like correlation evolves over time.

 

[Kevin_in_GA]

Quote from total_keops:

There is no now, cointegration is always backward looking and like correlation evolves over time.

There is always a now when it comes to placing a trade, but let's avoid philosophical arguments for the moment.

I agree that the half-life needs to be considered when trading pairs, since it helps in managing your trades. PTF gives you and "average days in trade" number which is not the half-life.

I just completed some interesting statistical analyses of cointegration versus correlation on 6 of the 9 S&P sector constituents. Preliminary analysis is posted here.

This is based on 1 year of data (shortest timeframe for backtesting in PTF) and 1 year of trades. Settings were entry at delta > 2, exit at delta < 1, 100 day mean and SD used.

To me, this is a pretty compelling set of data for using cointegration rather than correlation.

 

[jtrader33]

Might have a typo in the financials grid. Coint <.95 and Corr < .75 has a negative sign for profit per trade...if that's correct then your profit per trade for all financial pairs is off.

Quote from Kevin_in_GA:

There is always a now when it comes to placing a trade, but let's avoid philosophical arguments for the moment.

 

[amitkumar_ny]

Quote from Kevin_in_GA:

There is always a now when it comes to placing a trade, but let's avoid philosophical arguments for the moment.

I agree that the half-life needs to be considered when trading pairs, since it helps in managing your trades. PTF gives you and "average days in trade" number which is not the half-life.

I just completed some interesting statistical analyses of cointegration versus correlation on 6 of the 9 S&P sector constituents. Preliminary analysis is posted here.

This is based on 1 year of data (shortest timeframe for backtesting in PTF) and 1 year of trades. Settings were entry at delta > 2, exit at delta < 1, 100 day mean and SD used.

To me, this is a pretty compelling set of data for using cointegration rather than correlation.

First, thank you for doing the work and sharing instead of just preaching like others. I have not had much luck trading mean reversion using PTF. Although all backtesting shows great results when I actually do it I almost never make any money. Here is what I think the gap is : The price points used to do the calculation are rarely achieved due to slippage and time lag. If there was an automated way to purchase using real time feed I can guarantee returns but as is it seems information arrives too late.

 

[Kevin_in_GA]

Quote from bs2167:

Might have a typo in the financials grid. Coint <.95 and Corr < .75 has a negative sign for profit per trade...if that's correct then your profit per trade for all financial pairs is off.

You're correct - that was a typo. Correct number is $147.33.

 

[newguy99]

my ptf takes so long to update prices for my pairs, only about 50 pairs, my internet is very fast but this software internet connection goes to red and green back and forth and takes so long for pairs that are already saved, any suggestions

 

[Kevin_in_GA]

Quote from newguy99:

my ptf takes so long to update prices for my pairs, only about 50 pairs, my internet is very fast but this software internet connection goes to red and green back and forth and takes so long for pairs that are already saved, any suggestions

System Preferences -> General Options

Change your update frequency to something faster than the default (300 seconds). Other than that I got nothing.

 

[riviera]

allow me to post this pair before market opening

long : EFX
short : NOC

 

[Kevin_in_GA]

Quote from riviera:

allow me to post this pair before market opening

long : EFX
short : NOC

Equifax versus Northrup Grumman?

1. These are not in the same sector or industry, which while not required is usually a good thing.

2. I ran a series of Johanson tests on this pair, across four different time frames since 2008, and in no case are they showing co-integration.

This trade may work out for you, but I would not say that it was well thought out.