The Tale of Two Backtests: Why Aren't Opposite Strategies Absolute Returns Equal?

[Matt_ORATS]

You might think a strategies like short call spreads and long call spreads should have equal and opposite returns given the same trading rules. That would be true except that ORATS considers commission and slippage in our Backtester.

Take these VXX examples: Long call spread and short call spread with the exact same days to expiration and strike selection deltas. Returns are -9.23% for the long and 3.51% for the short. Combining these two using the function in the backtester we get an annual return of -5.72%. Why is that?

Download the trades file for both strategies and see that the average slippage per trade (the absolute value of the trade price of the long - short) is $0.10 and commission is $0.02 for the two leg spread held to expiration. Doing some math: the average days to expiration is 30 so there are 12 trades per year; the average price of the underlying stock is $26 and =(0.12*12)/26 = 5.5% ~ the difference in returns of return of -5.72%.

That is a big difference, something to consider when trading: Try to minimize your slippage and commission.

More reading: https://blog.orats.com/backtest-downloading-return-results-to-excel

See our documentation on how the backtester commissions, slippage, calculating daily returns, calculating annual returns, and graphing an equity line.

 

[GRULSTMRNN]

Your post is absolutely confusing and causing mental headaches. You seem to be mixing up multiple concepts. Your thread title suggests you are talking about strategy returns where on one hand trades are entered one direction and an otherwise equal strategy where trades are entered in the opposite direction. So far, all good. However, then you suddenly go into different slippage and cost of execution. Unless your own backtest methodology makes the explicit assumption that trades entered one direction cost a different amount to execute than trades entered another direction, both cost of execution should be equal. It might even be prudent to assume different costs of execution (such as going long stock and shorting a stock might incur a different cost), however, nowhere do you explain your rationale why in your example this should be the case. Mixing up cost of execution with the non-linearity in payoffs is probably equally confuse a lot of other readers on this site.

An example where one enters longs always on the first bar of the day and exits on the 10th bar each day, with otherwise identical strategies, and trading futures (a safe assumption that identical cost of execution can be applied) one must necessarily arrive at exactly opposing returns. The reason is the linearity of payoffs in futures markets in general. The same does not necessarily hold with options due to the non-linearity property of options contracts (though, it might, depending on the exact options contract specifications and the concept of Put-Call parity)

You might think a strategies like short call spreads and long call spreads should have equal and opposite returns given the same trading rules. That would be true except that ORATS considers commission and slippage in our Backtester.

Take these VXX examples: Long call spread and short call spread with the exact same days to expiration and strike selection deltas. Returns are -9.23% for the long and 3.51% for the short. Combining these two using the function in the backtester we get an annual return of -5.72%. Why is that?

Download the trades file for both strategies and see that the average slippage per trade (the absolute value of the trade price of the long - short) is $0.10 and commission is $0.02 for the two leg spread held to expiration. Doing some math: the average days to expiration is 30 so there are 12 trades per year; the average price of the underlying stock is $26 and =(0.12*12)/26 = 5.5% ~ the difference in returns of return of -5.72%.

That is a big difference, something to consider when trading: Try to minimize your slippage and commission.

More reading: https://blog.orats.com/backtest-downloading-return-results-to-excel

See our documentation on how the backtester commissions, slippage, calculating daily returns, calculating annual returns, and graphing an equity line.

 

[Matt_ORATS]

Your post is absolutely confusing and causing mental headaches. You seem to be mixing up multiple concepts. Your thread title suggests you are talking about strategy returns where on one hand trades are entered one direction and an otherwise equal strategy where trades are entered in the opposite direction. So far, all good. However, then you suddenly go into different slippage and cost of execution. Unless your own backtest methodology makes the explicit assumption that trades entered one direction cost a different amount to execute than trades entered another direction, both cost of execution should be equal. It might even be prudent to assume different costs of execution (such as going long stock and shorting a stock might incur a different cost), however, nowhere do you explain your rationale why in your example this should be the case. Mixing up cost of execution with the non-linearity in payoffs is probably equally confuse a lot of other readers on this site.

An example where one enters longs always on the first bar of the day and exits on the 10th bar each day, with otherwise identical strategies, and trading futures (a safe assumption that identical cost of execution can be applied) one must necessarily arrive at exactly opposing returns. The reason is the linearity of payoffs in futures markets in general. The same does not necessarily hold with options due to the non-linearity property of options contracts (though, it might, depending on the exact options contract specifications and the concept of Put-Call parity)

Thanks for the feedback.

The point of the post is how costly slippage and commissions can be. With none of these assumptions, you might expect the returns from a long call spread and short call spread to be equal in absolute value, like the long call spread returning -6.37% annualized and the short makes 6.37%. However, due to commissions and slippage, each strategy does worse by about 2.86%, with the long call spread losing -9.23% and the short call spread making only 3.51%.
We do have fixed slippage as sited in the post: https://docs.orats.io/backtest-api-guide/backtester-methodology.html#slippage

For a 2 leg strategy as here we assume a slippage of 66%, or traveling 66% of the bid-ask spread to trade. This amount is consistent whether buying or selling the call spread.

Our entry and exit methodology uses a snapshot of the entire options market 14-minutes before the close each trading day back to 2007. We use actual bid-ask quotes to simulate the trades.

I hope this helps.

 

[GRULSTMRNN]

the cost of execution should always be assessed as a function of trade frequency and avg gross percent return per trade. Even if I multiplied your cost assumptions by 10 then on a 10 million notional position that I held for half a year the cost would be absolutely irrelevant. Cost of execution become relatively speaking larger the more frequent one trades and hence the lower the avg gross percent return per trade becomes. You presented one example but ignored the entire larger concept and scope. If it helps someone, good for them, I found your post confusing at best.

Thanks for the feedback.

The point of the post is how costly slippage and commissions can be. With none of these assumptions, you might expect the returns from a long call spread and short call spread to be equal in absolute value, like the long call spread returning -6.37% annualized and the short makes 6.37%. However, due to commissions and slippage, each strategy does worse by about 2.86%, with the long call spread losing -9.23% and the short call spread making only 3.51%.
We do have fixed slippage as sited in the post: https://docs.orats.io/backtest-api-guide/backtester-methodology.html#slippage

For a 2 leg strategy as here we assume a slippage of 66%, or traveling 66% of the bid-ask spread to trade. This amount is consistent whether buying or selling the call spread.

Our entry and exit methodology uses a snapshot of the entire options market 14-minutes before the close each trading day back to 2007. We use actual bid-ask quotes to simulate the trades.

I hope this helps.

 

[Matt_ORATS]

the cost of execution should always be assessed as a function of trade frequency and avg gross percent return per trade. Even if I multiplied your cost assumptions by 10 then on a 10 million notional position that I held for half a year the cost would be absolutely irrelevant. Cost of execution become relatively speaking larger the more frequent one trades and hence the lower the avg gross percent return per trade becomes. You presented one example but ignored the entire larger concept and scope. If it helps someone, good for them, I found your post confusing at best.

My experience (12 years as a market maker and 30 years in the options business) is that the cost of execution is best estimated by a percentage of the width of the bid-ask spread, especially for an infrequent trading strategy. I stated a trade frequency of 12 times per year for the backtest. I agree that more trades would affect the returns more negatively. However, even trading this strategy relatively infrequently, the width of the bid-ask spread is punitive on returns.

 

[GRULSTMRNN]

Well, then it's a poor strategy. Simple as that. One can get away with even 25% execution related cost as percentage of gross returns if one can turn over the strategy very frequently and generate many intraday trades. Same with very infrequent trades that generate much larger returns. Obviously, what is acceptable to each trader is his/her decision but there are well and poorly performing strategies. Poorly performing strategies are those which trade very infrequently and generate so small returns that the cost of execution greatly affects overall performance.

If that was your point then I guess we agree overall. And if you wanted to inform beginners that cost of execution cause opposing trading strategies to not yield opposing matching returns then we also agree.

My experience (12 years as a market maker and 30 years in the options business) is that the cost of execution is best estimated by a percentage of the width of the bid-ask spread, especially for an infrequent trading strategy. I stated a trade frequency of 12 times per year for the backtest. I agree that more trades would affect the returns more negatively. However, even trading this strategy relatively infrequently, the width of the bid-ask spread is punitive on returns.