Strategy Backtesting

[equus]

Kim, when backtesting the pre-earnings calendars, is the idea to track the price of the ATM spread each day in the runup to the historical earning date, whatever the ATM strike price happens to be on a particular day? 

 

For example, FSLR calendar: the underlying might be 60.00 on Monday, 62.50 the next day and 65.00 the next day etc -- do you look at the price of the 60 calendar on Monday, the 62.5 calendar on Tuesday and the 65 calendar on Wednesday?

[Kim]

Yes, always ATM calendars. This is how Dustin has his charts as well. The reason is that we want to eliminate the impact of the stock move. This is something we cannot control in real life, but for many stocks we trade it will not make significant impact.

[equus]

Kim, I had a look at the historical ATM spread values for the recent GOOG April 15 calendar we did using ThinkBack's End of Day data.

 

Across the days I looked up running up to earnings, the average ATM value was 1.52, with a minimum of 1.43 and maximum 1.73. In other words, the granularity of the EOD data is not fine enough to show the 1.15 entry that some folks achieved on 20 April.

 

Is it necessary to be looking at the intraday spread prices to establish the range for the day or is EOD good enough?

[Kim]

I think you answered your own question 

[equus]

Got it, thanks.  

 

Are there any other metrics it would be good to record during the backtesting process, such as IV or IM of the long/short legs? 

[Kim]

No, I'm looking at prices only.

[equus]

For backtesting earnings calendars, how many quarters back is it helpful to go?

[equus]

As a pointer, I wonder what is a good minimum and maximum range for number of quarters to backtest?

[Kim]

I would say 3-6.