Back testing is the process of evaluating a strategy by looking at its return over a series of historical dates using real (i.e., historical) data. This lets you see how a strategy would have performed in past markets if it had actually been entered into in the past.
Although such a tool can never actually prove a theory correct (because past results do not necessarily indicate future results), it can be a useful tool of analysis and prediction. In addition, its use of actual historical market data lets you better analyze a strategy as it encountered real-world conditions in the past.
This topic looks at how back testing is carried out in SDX Interest Rates.
For more information on:
Activating back testing in SDX Interest Rates see Activating Back Testing in SDX Interest Rates.
The displayed results see What Back Testing Data Is Displayed in SDX Interest Rates?.
Comparing Strategies in the Back Testing Page
Comparing Legs in the Back Testing Page
Accessing Recently Calculated Strategies in the Back Testing Page
Using the Favorites List in the Back Testing Page
How Back Testing Is Carried Out in SDX Interest Rates
Back testing is supported in SDX Interest Rates for a defined period up to and including the current date (a period that is referred to as the back testing period).
It is supported in the Single Option and the Portfolio page, where you can back test a portfolio of up to ten instruments.
The following instruments are supported as relevant to the pricing page:
Vanilla swap
Swaption
CMS spread option
Cap/Floor
CMS Cap/Floor
Swaption strategies
These strategies are not supported in the Portfolio page.
Zero cost swaption collar
Cap/floor strategies
These strategies are not supported in the Portfolio page.
Zero cost cap/floor collar
Callable swap
Range accrual swap
Callable range accrual swap
Zero coupon swap
Callable zero coupon swap
Callable inverse floater
Callable capped floater swap
CMS spread range accrual swap
CMS range accrual swap
You can manually select any of the following calculation modes:
Payout at Maturity
In this mode by default each point shows the annualized1 total payout (from inception to maturity) of the structure that expires on that date.
For an instrument with underlying coupons/caplets/floorlets, the results are calculated for each underlying coupon/caplet/floorlet. For a swaption, it is the payout calculated on the instrument’s expiry date.
You can, however, choose to see the actual result for the defined period, rather than the annualized result, by setting the Annualize field to No.
P/L from Inception
In this mode by default each point shows the annualized P/L (from inception to the current date) of the structure that begins on that date.
You can, however, choose to see the actual result for the defined period, rather than the annualized result, by setting the Annualize field to No.
If the structure also contains options, then you can choose to calculate these results for those options with or without taking the premium into account. By default, the premium cost is included (the Subtract Premium field is set to Yes), but you can change this by setting the Subtract Premium field to No.
Daily P/L of the Structure
In this mode by default each point shows the actual daily P/L of your original structure. The date of each point is the later of the 2 dates defining the period for which the P/L is being calculated.
You can, however, choose to see the annualized result rather than the actual result for the defined period, by setting the Annualize field to Yes.
P/L after 6 Months
In this mode by default each point shows the annualized P/L of the structure that began 6 months prior to that date. The P/L is calculated for the period that ends 6 months from the structure’s start date.
You can, however, choose to see the actual result for the defined period, rather than the annualized result, by setting the Annualize field to No.
|
|
This is the default calculation mode. |
Price at Inception
In this mode by default each point shows the market value of the structure that begins on that date, i.e., it shows the cost of the structure at its inception.
|
|
For a swap if you do not define your own fixed rate the system automatically uses the breakeven rate for each strategy. Therefore the price at inception will always be zero. |
For all these modes (with the exception of the P/L from Inception mode which only uses the original structure itself) to perform back testing the system in effect simulates a series of replicate instruments, each of which expires or matures on a subsequent business date in the back testing period. Because the system keeps a constant time to maturity for each of the replicated instruments (i.e., it assumes that the number of days in each replicate instrument is the same as the number of days in the original instrument) it defines each replicated instrument’s trade date as follows:
Replicated instrument's expiry date or maturity date - no. of days in the original instrument
For each of these replicate instruments the system then calculates the payout or P/L (as indicated by the selected calculation mode). So for example, on 4 November 2010 you price a 1Y3Y USD swaption and you choose to back test its payout at maturity for a period of 3 months to the current date. The system will then calculate the payout for a 1Y3Y USD swaption vanilla that expires on 3 August 2010, for a 1Y3Y USD swaption that expires on 4 August 2010, for a 1Y3Y USD swaption that expires on 5 August 2010, etc., up to and including 3 November 2010.
When calculating the results:
The system uses the relevant market data. That is, for each replicate instrument it uses the end-of-day market data saved in the system for its trade date and expiry date and, if relevant, for each of its fixing dates.
The system takes into account any relevant events, such as fixing dates.
The payouts are displayed as a percentage. This is calculated as the cash payout / notional.
For a swap, if you do not define your own fixed rate in the original instrument, for each replicate instrument the system calculates the break even rate on the fly (using the market data saved on the replicate instrument's trade date) and uses this as the fixed rate. If you do define your own fixed rate in the original instrument, this fixed rate is also used for each replicate strategy.
For an option, the system firsts calculate the strike to use for each replicate strategy. How this is done depends on how the original strike is defined.
The system supports both of the following strike definitions:
The original strike is defined as an absolute strike.
In this case, this fixed strike is also used for each replicate strategy.
The original strike is defined using a shortcut, e.g., atm.
When the strike is defined as a shortcut, regardless of whether or not you calculate the instrument before back testing it, the strike’s shortcut is kept and is calculated on the fly for each replicate instrument according to the market data on the replicate instrument's trade date.
This is also true if in the Single Option page you enter a a zero cost cap/floor collar or a zero cost swaption collar, i.e., if in the pricing page you enter the strike of one leg and for the other leg you enter z. When you then load the strategy to the Back Testing page the strike’s z shortcut is kept. That is, for each replicate instrument in the back testing period the system will calculate the strike for that leg that will result in a zero cost strategy. Again, the z shortcut is used in the Back Testing page whether or not you first calculate the strategy in the pricing page.