Everything you ever wanted to know about developing systems but was afraid to ask

[Murray Ruggiero]

Quote from Murray Ruggiero:

Let's start talking about data.
Understanding how data effects testing and developing trading systems is very important.

One example is how futures contracts are rolled over. There are many ways to do that. First rolling on volume , open interest and then volume and open interest. We select the most active contract. Another way is to roll on fixed dates used by many commercial traders, for example 30 year treasury futures roll on the 26th day of the month before expiration. How you roll makes a big difference on a market by market basis as much as 35%. If we trade a large portfolio with the same system these results balance out to maybe a 10% difference.

Another issue is how the birth of the electronic markets effects long term backtests. Some data vendors have electronic only data and combined electronic /Pit history. This combined history is not good because it include old history prior to 2008 , when many electronic markets were not liquid.

In order to create long backtests we have created, Pit/Pure electronic data series rolled on fixed dates. We create a roll from pit to electronic when the volume of the electronic markets became greater than the pits. Pinnacle data create these back adjusted contracts with my help and make them available.

A few clients have told me that CSI has now create the same type of options for the data series , pit/electronic. I need to discuss with them to verify.

 

[Murray Ruggiero]

http://www.futuresmag.com/2011/06/01/how-to-backtest-trading-systems-and-avoid-curve-fi

http://www.futuresmag.com/2011/08/01/trading-with-rolling-contract-data

Here two articles which explain this data issue in more detail.

 

[JackDogII]

Thanks Murray.

A simple and useful explanation of a meaningful topic.