Pair Trading Strategy Journal

amitkumar_ny · Jun 30, 2011

Quote from Trader13:

There are really two time factors for a cointegration test. The first time frame is simply the length of historical data used for the test. For example, last n days or years. The second time factor is the lag value used to conduct the test. In theory, this lag value (aka, half-life) should represent the half-life of the mean reversion cycle. The lag is either user-defined as a single value, or more advanced software may iterate and report results across a range of lag values. <i>Interpreting the lag value along with the significance statistic is very important. </i>There are resources on the web to learn more about this.

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. [/QUOTE

Emperically, would this be the same as average time to revert to mean in the past ( for example 10 yrs)...

klurby · Jun 30, 2011

Hello would someone mind explaining to me what pair trading is? thanks.

total_keops · Jun 30, 2011

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

Good job, it's always good to have numbers backing you. A more compeling test would be to take the same pairs out of sample and see if the cointegration holds.

amitkumar_ny · Jul 1, 2011

Quote from total_keops:

Good job, it's always good to have numbers backing you. A more compeling test would be to take the same pairs out of sample and see if the cointegration holds.

What do you mean ? I think these pairs are arrived at using backtesting. Are you suggesting paper trading these pairs to see if they hold in future ?

amitkumar_ny · Jul 4, 2011

The latest version 303 is miscalculating the cointegration. I checked and double checked with catalyst corner and after reverting to the older version. beware...

total_keops · Jul 4, 2011

Quote from amitkumar_ny:

What do you mean ? I think these pairs are arrived at using backtesting. Are you suggesting paper trading these pairs to see if they hold in future ?

Could be that. What I meant was that you should have backtest you thing on say, 2009 (in sample) and then select your pairs and adjust your parameter values with that data. Then, with this, you backtest (out of sample) on 2010 data and see if it holds becaues it is that that you would have traded.
As a bonus, you could run you backtest on a non trendy or abnormal year like 2008 and see the house of cards you may have, I dont know.

Trader13 · Jul 5, 2011

Quote from optioncoach:

I also think you might find more success and better entry signals if instead of doing all of these 1:1, take the ratio to make the trade mroe even. In other words, BIDU is about 1.83 times AAPL so a spread can be, for example, 100 BIDU Long/Short 183 AAPL.

Chart that spread and you will see entry and exit patterns. That is how I have been doing spread trades with some success.

+1

Your weighting of the stocks in the pair should match how you chart it when conducting your analysis. Typically, weighting schemes are by shares, dollars, or volatility. Your attached charts appear to be ratio charts which align with equal dollar-weighting, just as optioncoach advises.

miketrent · Jul 5, 2011

Its unbelievable how a stock split destroys all the data on a pair in pairtradefinder. :mad:

Kevin_in_GA · Jul 5, 2011

Quote from miketrent:

Its unbelievable how a stock split destroys all the data on a pair in pairtradefinder.

I asked Jared this exact question, as I noticed the same thing. Here is his response:

Kevin,

This depends on the datafeed and the stocks. Using the yahoo feed with large stocks is correct 99.9% of the time, they adjust the prices and are reflected in our historical data, with IQfeed it's always right, so it doesn't affect backtest results. It's easy to see with a low liquid stock in yahoo whether's it been affected by large spikes.

Regards,
Jared.

He really did not answer the question about whether or not the backtest data and Deltas are affected by this (I can't see how they wouldn't be). He talked about spikes, not splits.

miketrent · Jul 6, 2011

I don't know why he mentions spikes and low volume. Thats not the issue. When there is a stock split all the data is destroyed and the software generates completely false signals. For example, TMK did a 3-for-2 yesterday and every pair with that stock went haywire in the software. This is a really bad bug that has to be fixed.