Test 7: IV/HV

I thought it would be fun today to test whether it makes sense to use implied volatility as an entry criteria to our tests. Specifically, I am looking at whether the ratio of IV to HV (historic vol) will be beneficial.

I’ve heard that generally speaking it is a bad practice to sell premium when the HV is larger than the IV. We like to sell in high IV environments with less underlying movement (actually that is ideal). When IV is lower than HV, can we get enough premium to justify the potential large swings in the underlying? We are actually under that situation right now with the VIX at 13.49.

For this test, I construct rules that only allow trades to be entered when the IV/HV ratio is greater than (1.0, 1.1, 1.25, 1.5, 1.75, and 2.0). No other entry or exit parameters are used. Note that in this test I am only using the implied volatility of the front-month option chain calculated based on a weighted average of the ATM call and put IVs.

Results

1 Month 2 Months 3 Months All
Improved PF 9(17%) 38(70%) 38(70%) 85(52%)
Improved MaxDD 54(100%) 54(100%) 54(100%) 162(100%)
Improved ExpRet 9(17%) 38(70%) 38(70%) 85(52%)

The initial pass is interesting; the back months show a much higher benefit that the front month. The heatmap shows this in even more detail:

iv-heatmap-diffpf

The obvious conclusion is that if you find yourself with a 2x IV over HV, make the trade, but the data shows these are extremely rare (i.e. 6 trades in 9 years). Discounting these trades, we can see that generally speaking in the other cases using IV/HV as a marginal impact.

There are several trades that really stand out to me, all in the 2nd month. The 10-Delta, 15-Delta, and 18-Delta trades with IV/HV of 1.25 and 1.5 all saw at least 25 trades during this period and had substantial improvements. The 10-Delta 1.5 IV saw only winners. What is so special about this cluster of trades? Digging into the specific trades they show that they only have a few losses, but their neighbors have about the same number of trades and just 1 or 2 more losses. Those extra losses were enough to dramatically skew the results considering this is an expiration test and the losses were all max losses.

Now, what happens if we also include the early exit trigger from the previous test?

iv-style-heatmap

In this case we start to see pretty consistent and significant improvements in the back month trades. These are also fairly significant improvements over the early exit tests themselves. And note the cluster in the second month trades is no longer as significant.

Conclusion

On its own, the IV/HV ratio is not a great indicator to help enter trades. While there are scenarios where the ratio seems to help a great deal, these come at the cost of significantly less trades. However, it seems that paired with other triggers it may be more beneficial than not.

The higher the IV/HV ratio, the less number of trades that are going to occur. Is it worth it to wait until the IV is greater than the HV, in my opinion… yes if you are looking at back months. I am actually quite surprised that there was not a more dramatic effect for the front month since this test was using the front-month IV in its calculation. In the last 9 years there are only a few cases where the IV/HV ratio was less than one (i.e. HV greater than IV), and only one period where the HV was higher for sustained period of time.

Overall, I’m not convinced this tool is critical unless one is already in a low vol environment. If the HV is high, but the IV is also high, we would expect to be paid a decent premium and be able to get wider strikes, possibly nullifying the effect.

It is also interesting to note that the HV is a lagging indicator and the IV is a current sentiment indicator based on the market’s perception. Realized volatility in the past may not carry though to the time when HV is greater than IV. This would result in potentially missing trades good trades, and possibly allowing some bad ones.

Test Stats

Permutations 162
Profitable Scenarios 121 75%
Total Trades 7283
Profitable Trades 5212 72%
Expired 2442 34%
MaxDays 4841 66%
Test Duration 1.38(min)

Test Files

Excel Data