Today I am going to test early exits. To date, my tests have held trades through to expiration unless a target or stop loss was reached. But many traders prefer to exit trades well before the expiration to minimize gamma risk. Of course, one would miss out on the largest theta decay, but theta and gamma often are opposing greeks.
The general sense I get is that those entering front month trades attempt to hold though to expiration, possibly exiting with 1 week left. Those entering the back months tend to be out of the trade with at least 3 weeks left to go. This run will test exiting 1 through 4 weeks prior to expiration, with no other entry or exit parameters. With this test, I am simply hoping to find a best practice depending on the expiration month.
|1 Month||2 Months||3 Months||All|
The table shows that where were substantial improvements in the Profit Factor for the back month. In all cases the risk was reduced, but in the majority of cases in the back months the expected returns were also improved.
Looking at a heatmap for the improvement in profit factor over the baselines:
We can wee that front month actually saw an improvement when the trades were exited 7 days prior to expiration. And very high delta trades may also see an improvement at all exit periods. However most other trades saw marginal or negative improvements as we might expect from these trades.
The back month trades are another matter. Clearly exiting early has a benefit on nearly all these trades. Exiting 3 weeks prior seems to be a general sweet spot for both back month trades, however, the 3-month 10-Delta trade saw the largest improvement when held to the last week.
So there is something to exiting early. Lets see if we can nail down some better rules. I reran the test, but this time included all days between 3 and 28 prior to expiration.
With this heatmap I removed all the weekend days and then color coded each row individually. Green reflects the most improvement and red the worst by delta. Red does not necessarily mean bad as any value over zero is an improvement over the baseline. The aim is to show the best exit point for each expiration month and delta. For example, if you were to enter a trade with the front month and a 30 delta, an exit with 21 days remaining may actually be in your interest.
What is interesting about this map is that several patterns emerge.
- Exiting 22 days for the 3 month trade is generally optimal except for the delta-10 trades where holding to 4 days before seems to be beneficial. This is very anomalous.
- Exiting between 17 and 22 days for the 2 month trades has the highest benefit.
- Exiting early in the front month generally has marginal to negative improvement. Low delta trades may benefit from exiting just prior to expiration. Otherwise they are harmed. Higher delta trades show a marginal benefit to exiting early at most points with the largest gain being 17-22 days prior (note these would be very short term trades).
- Exiting 2 weeks prior on the Friday seems to perform much worse than if you were to exit a few days before or after that point. Why two weeks before? This pattern is not seen at 3 or 4 weeks on fridays.
Now, why is there an abnormality around the 3-Month, 10-Delta, Day-4 exit trade? It performs significantly better compared to its neighboring trades. The neighboring trades on the same delta have identical entry points and the same number of trades. However, when exited one day earlier (day 5) the scenario wound up with 4 more losses. Similarly, the neighbor in the 15-delta row had a few more total trades, but also experienced 4 more losses. Given that many of the trades were identical and that just exiting one day vs another made such a difference in 5% of the trades, I suspect this is just a statistical anomaly that would be smoothed out over time with more trades/data.
Clearly there is a benefit to exit back-month trades earlier. It substantially decreases the risk of the trade, and in many cases leads to higher expected returns probably by missing out on larger losses. Exiting between 18 and 22 days appears to lead to the largest benefit.
For most of my tests going forward unless noted otherwise, I exit trades for back month trades based on this data.