Trading stock pairs

[avisarqi]

Posting new thread related to trading stock pairs.

I love pair trading strategy, it is holding my profits over last three years .

Some theoretical introduction:
The Stock Pair Trading is a market neutral strategy capitalizing on short-time market inefficiencies. The strategy is based on the fact that the ratio of prices of two suitably chosen (correlating) shares is long-term stable and oscillates around the average value. If (for whatever reason) ratio comes outside the normal area, you can speculate with high success rate that ratio will return to the mean value. Stock pair position is than formed by entering long and short position in two stocks of the same sector. This creates a powerful hedge against the sector and the overall market fluctuations.

Strategy performance is very impressive :fisheye::
70,5% stock pairs is profitable at 2 years period
Average stock pair net revenue 110% p.a. (including loosing pairs!)
Average trade 94,6$ (allocated capital 10.000$; trade length 12 days)
Average win% is 65.3%, RRR = 0.824
Tested 90.000+ stock pairs from all sectors of US economy (with no optimization!)

Do you have experience with this kind of trading? Appreciate any ideas, tips ... or questions... ;-)

Petr

[AvianeN]

Pair trading strategy is market neutral, hence it is robust and stable.

Carefully selected pairs gives you profits for years...

See real portfolio performance since 01/01/2009 to 12/31/2012:
Initial capital: 12.495 USD
Portfolio: 10 stock pairs
Maximum positions: 5
Average net revenue: 130.2% p.a.

[avidinuq]

Some theory about correlation:
Selection of suitable pairs is based on mutual correlation (dependency) of shares in the pair. The greater the correlation of stock prices, the more stable is the ratio of prices. Perfectly correlated stocks can usually be found within one sector (or subsector) economy (eg. mining and processing of gold, Life Insurance, Telecommunications, etc.). Selection of stocks in one sector also eliminates the risks associated with different economic development of various sectors.Correlation of the shares in a pair is examined at different periods, most commonly 30 to 400. Correlation of prices fluctuate over time (the more, the shorter the period of calculation of correlation). Therefore average correlation is used for quantification. Stable stock pairs perform an average correlation of 70% to 80%.


Petr

[Avrelyolert]

You need just EOD Close price & Volume for backtesting. You can download these data anywhere. Free sources: Google Finance, Yahoo; Paid service: EODData.com.
According to my investigation, all the sources are very similar. EOD Close is "basic" price, nothing to spoil... ;-)

Important:
- Historical data should NOT be "dividend adjusted". You need real (raw) historical data. In fact, you can use even "dividend adjusted" data, but your result will be affected.
- Historical data MUST be "split adjusted" - remove price gaps resulting from splits. Splits in unadjusted data generate fake signals / trades in backtest... Pair statistics are not useless then.

Stock Pair Trader can be connected to InterActiveBrokers data feed or IQfeed (unlimited history).
Once you fill your database with historical data, program automatically download updates from InteractiveBrokers.

Petr

[avociyefaqipu]

I attach Excel spreadsheet with backtest results of 90.000+ stock pairs.
Backtest was performed using screening functionality of Stock Pair Builder. (within 30 minutes... ;-) )

Pair parameters are default, no optimization.
See average values in table header - more than 65% pairs is profitable! From 10 randomly selected pairs 7 pairs is going to be profitable!

Model Ratio: http://stockpairtrader.com/EN/download/Ratio.xlsx
Model Residual: http://stockpairtrader.com/EN/download/Residual.xlsx
Model LogResidual: http://stockpairtrader.com/EN/download/LogResidual.xlsx

Pair trading performance is based on statistical approach. There is HUGE trading edge in this strategy.

Visit my project website: http://stockpairtrader.com/EN/