Test 2: Stop Losses

What level of stop losses are effective for managing iron condors? Our baseline test effectively had a 100% stop loss on all scenarios. Will adding stop losses based on the at risk margin (See the test on Profit Limits to learn why I use this metric) improve the performance of Iron Condor trades?

My guess before looking at the data is that stop losses will be shown to drastically reduce the maximum risk, and should be able to improve the effective expected return. Because some tests will be stopped out quiet frequently (i.e. those with a 10% stop loss threshold), I expect many tests to have a higher number of trades compared to the baseline.

Test Results

The test scenario use stop losses on the margin at risk for the following values:

10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90%, 100%

The following summarizes the performance of the test run:

1 Month 2 Months 3 Months All
Improved PF 32(36%) 28(31%) 28(31%) 88(33%)
Improved MaxDD 90(100%) 90(100%) 90(100%) 270(100%)
Improved ExpRet 34(38%) 26(29%) 25(28%) 85(31%)

As expected the risk was substantially reduced in all cases. The vast majority of scenarios actually experienced lower expected returns and lower and lower profit factors. Meaning that the gains in risk were still outweighed by the loss in expected returns.

Profit Factor Improvement VS Delta

Looking at a histogram of the difference in profit factor compared to the baseline, we can see some patterns, but its not as clear cut. Very small stop losses perform poorly on average. And High Delta trades appear to perform poorly on average, although there is a cluster in the 3 month range and Delta 35 that does well. The clusters between the 30 to 60% stop losses and deltas under 22 seem to consistently outperform as well. But again, no very clearcut patterns emerge.

sl-hist-pf

 

sl-diffPF-vs-delta

Expected Returns

Looking at the expected returns, we can see that there appears to be a sweet spot between -30% and -60%, and Deltas less than 22. Similar to what was displayed in the profit factor tables.

sl-hist-expret

We can also see that regardless of the stop loss, high delta trades perform poorly. This is not surprising given the construction of the test.

sl-expRet-delta

Conclusion

Traders may often hold the Front Month trades to expiration but also use stop losses. This test shows that many of those scenarios can be improved marginally, but using a 30% to 50% stop loss on 15 to 20 delta trades seem to gain the most benefit.

The second month, low delta trades with a 30% to 50% stop loss may also be marginally improved.

The tests with Stop Losses between 10% to 20% are significantly impacted across all deltas and all months. These trades are just stopped out to frequently.

Most other tests are only marginally impacted either positive or negatively.

Test Stats

Permutations 270
Profitable Scenarios 174 64%
Total Trades 41402
Profitable Trades 17911 43%
Expired 19394 47%
StopLoss 22008 53%
Test Duration 1.52(min)

Test Files

StopLoss Excel Data

Advertisements

3 thoughts on “Test 2: Stop Losses

  1. You mentioned being open to suggestions for future tests. I’d love to see some similar tests on the impact of various delta / theta scenarios on calendar spreads, comparing use of profit targets, time-based targets (e.g., always get out with 1 week remaining), and stop-loss exits (e.g., always out at X% loss).

    Also, I’m guessing your software is not consumer-ready, but I’d love to get an idea of how you built it, what language / DB you used, key functions, etc. Or anything else you’d be willing to share.

    • I’m definitely open to suggestions. I need to work though some of the iron condor tests I already have planned, but then I can move on to Calendars. I still have not built out that functionality (calendars), but I need to so I can test using Calendars as a Vega hedge with ICs.

      The software is completely custom (except for opensource libraries), written in Java with no db. I find databases would be a performance tax on a system like this since effectively I need to read in 100% of the data during the tests. But one day I may be forced to move to one. I might have a write up on my system another day, but it piggy-backs on some other option-trading code I’ve written for my own personal trades.

      • Thanks for the reply. Sounds good, Dustin. I appreciate your choice to share what you find.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s