Put Diagonal Backtest (Part 6)

I spent time backtesting yesterday and wanted to make a couple comments before presenting additional data.

First, I did ask my brokerage about this footnote:

      > I’m thinking about selling a weekly SPX ITM put and rolling this as a
      > campaign. However, rather than close out the expiring SPX ITM put, I
      > would just let it get assigned and open the next one. I want to make
      > sure this will not incur any fees aside from debiting my account for
      > the intrinsic value of the option at expiration.
      >
      > Can you confirm this?

They responded:

      > SPX is cash settled so you would just see a cash credit or debit
      > based on the settlement value. There are no commissions/fees for
      > exercise/assignment.

This implies I may have overcompensated for slippage as discussed in Part 1. I like to err on the side of caution. Since backtesting has a tendency to artificially inflate results, decisions should be based on conservative projections.

Taking assignment on short options is beneficial. Not only does it allow me to retain maximal extrinsic value, which goes to zero at expiration, it also saves me transaction fees on the buy-to-close.

The potential downside to taking assignment is legging risk. Bid/ask spreads widen at 4:00 PM (normal close) and beyond as mentioned by OptionNet Explorer (ONE) support here. To avoid the wider spreads, I would open the new short position after 3:50 PM knowing the old will expire 25 minutes later. Legging risk refers to the possibility I may lose money due to an underlying price decline. Had I waited until 4:15 PM to open the new short, I would have avoided unrealized loss.

The backtest avoids legging risk by rolling to ensure concomitant execution of both options. “Trade like you backtest” has therefore been violated, which signifies the presence of backtesting inaccuracy.

Legging risk introduces backtesting uncertainty due to data limitations. Because ONE data stops at 3:55 PM, I cannot measure what happens through the 4:15 PM expiration. Over a large sample size, I would expect any gains and losses to average out since I don’t believe any edge exists by consistently going long or short equities from 3:55 PM to 4:15 PM.* Too bad the put diagonal backtest does not encompass a large sample size.

Another source of uncertainty is the remaining extrinsic value I pay to close short options in ONE. I used either 3:30 PM or 3:55 PM in the backtest. Shame on me for not being cognizant and consistently using the latter since time decay accelerates in the last 25 minutes: 3:30 PM includes more extrinsic value than 3:55 PM. This paid extrinsic value represents additional expense that would not exist in live trading, and uncertainty exists because I failed to measure it.

I will conclude next time.

* — This can be tested (make sure to factor in transaction costs).