Options Trading

The strategy in a nutshell

I’ve written about this strategy in previous posts, most recently in Part 10 and Part 11. But briefly, here is what I’m doing:

  • Keep a portfolio of productive assets (bonds, preferreds, stocks, etc.) in a taxable account at Interactive Brokers.
  • Sell short-dated, far out-of-the-money naked options on margin for additional income. I use the CBOE SPX options (100x multiplier), i.e., options on the S&P 500 index. These are cash-settled European options, and they qualify for the advantageous Section 1256 tax treatment. Even with potentially thousands of trades, there is no need to itemize your transactions on your tax return. You simply enter the net options trading profit, one single number, on IRS form 6781.
  • Very occasionally, you suffer a loss if options go “in the money” to a stop loss is triggered before then. But long streaks of making the full premium are usually enough to compensate for that. So far, I’ve made money with the strategy in every calendar year since 2011. In every market condition: bull markets and bear markets. March 2020, the height of the pandemic panic, was my most profitable options trading month ever.
  • You make money (on average) because the option implied volatility is far higher than the average realized volatility. I will provide some stats below!
  • If you trade this daily, you have about 250 independent investments per calendar year. Each individual option trade may have a highly non-normal and negatively skewed return profile. But averaging over enough independent trials, you again make returns (mostly) well-behaved. They even approach a Gaussian Normal, compliments of the Central Limit Theorem; see Part 3 for more details.
I miss the old 10 Deutsche Mark note. It had a picture of Carl Fridrich Gauss and a small figure with the Normal distribution named after him. Source: Wikimedia
From Part 3: Even a skewed distribution looks more and more Gaussian-Normal when you average over enough independent observations!

Returns since 2018

Why 2018? I’ve been trading options since 2011, but the account size was much smaller then, and I ran this strategy and a few others with much higher risk and return targets. In 2018, this IB account went to “prime time” when we sold our San Francisco condo, and a large part of the proceeds went into my Interactive Brokers account. That’s also when we started funding our retirement expenses out of this account. Average returns would look even better when including the early period (+100% in 2012!!!), but it would be comparing apples and oranges.

Here’s my options trading cumulative alpha chart; please see below. I should stress that these are the returns from the options trading part only. You trade options on margin, and the underlying portfolio, comprised of cash, preferred shares, and equities, is separate. Thus, your options revenue supplements your underlying portfolio. Even +0.1% p.a. would be a win, but I did better than that.

Cumulative excess returns Jan 2018 – Feb 2024. 12/31/2017=100. Nominal dollars.

There were a few drawdowns, but they didn’t last very long. Each down/up cycle was much faster than your average equity bear market. I got caught in the early 2018 vol spike right out of the gates. 2020 started with two down months in January and February, though the March 2020 blockbuster return made up for that again. So, quite intriguingly, the early part of the 2020 bear market caused some losses, but the really volatile month of March 2020 made up for it. As I explained in Part 10, the first half of 2022 was a bit choppy. But overall, this was a very successful strategy.

For the month-over-month excess returns, please see below. As mentioned before, since late 2022, I’ve walked down my return and risk target as a percentage of the account size. That’s really all the money I need to generate because I also have the dividend and interest income from the underlying portfolio.

Nominal Month/Month excess returns Jan 2018 – Feb 2024.

And some return stats, please see the table below. I list the annualized return and risk and the Information Ratio (IR), i.e., the excess return per unit of risk. One can think of the IR as the Sharpe Ratio equivalent for an alpha strategy because you subtract the benchmark (e.g., the underlying portfolio) instead of the risk-free return in the numerator. In fact, if you implemented this options trading strategy with a risk-free asset as the underlying asset (money market, 3m T-Bills, etc.), then your Sharpe Ratio equals the IR. Notice that a 3.0+ IR is phenomenal. Even an IR of 1.0+ is already quite impressive. The stock market Sharpe Ratio is about 0.30-0.35 (if you think of the Sharpe as an IR with a cash benchmark). 14.0+ over the last twelve months is astronomical, though it should be taken with a grain of salt due to the short horizon and the remarkably calm market environment. But it certainly makes for a good conversation starter in finance circles!

Return Stats Jan 2018 – Feb 2024. Nominal dollars.

The walk-down of average returns is also noticeable in the all/5Y/3Y/1Y return stats. It’s not because the strategy performed worse (in fact, the IR increased over time); it’s simply because I’ve treaded more cautiously over time.

Does Options Trading Generate “Alpha?”

I recently discussed options trading on Twitter, and a fellow personal finance influencer scoffed at my claim that I generate “alpha” with my options trading strategy. Well, the proof is in the pudding. The charts above prove that I received additional returns over and on top of the underlying portfolio, so by one definition, that’s undoubtedly an alpha strategy. But I like to use a narrower definition:

Alpha = an excess return not attributable to the market and/or style factors.

So, we can think of this as an estimate for the average excess return, but accounting for exposures a market factor or benchmark and sometimes other style factors, e.g. the Fama-French SMB and HML factors and others. In other words, it’s the intercept in a univariate or potentially even multi-variate factor regression model. The advantage of this approach is that we can separate returns into simply capturing risk premia – the part modeled by the beta factor loadings – and the left-over intercept, i.e., excess returns that look like alpha: skill, market timing, stock picking, arbitrage, market making, etc.

I took my Put Writing returns and regressed them on both S&P 500 and the CBOE Put-Writing index, see here for more info and return data. So here are the regression results; see the table below:

  • Model 1: There is slight exposure to the S&P 500, though the weight is minimal. While the 3% beta is statistically significant, it’s economically insignificant. The alpha is 7.12%, which is highly significant with a t-stat of almost 7. The R^2 is minuscule, too.
  • Model 2: The ERN strategy also has a tiny correlation and exposure to the CBOE put selling index. However, the beta is both statistically and economically insignificant. The R^2 is even lower than in Model 1. The alpha estimate is now 7.39%, the t-stat at 7.19 indicates a strong statistical significance.
  • Model 3: By adding both market betas to the mix, we certainly increase the R^2, though it’s still below 0.10. The SPX beta is a bit higher, but intriguingly, the PUT index beta is now negative (likely the effect of multicollinearity). The alpha estimate is still at 6.92% and still highly significant.
Factor Model Regression Results: ERN put selling vs. equity and options-selling betas. 1/2018-2/2024.

How about a kitchen sink model? Because I have the seven return series handy in my Safe Withdrawal Rate toolkit (See Part 28 of the series for more details), I can run a regression model with those seven factors. Please see the table below. Why would I use equity-style factors like SMB and HML on an equity index options strategy? No idea! I just throw everything at the wall and see if anything sticks! In any case, we still maintain an alpha of almost 7%. The factors are all in single digits and mostly offsetting each other (e.g., +5.52% SPX but -1.59% international stocks, or +7.87% 10Y bonds and -3.93% 30Y bonds). So, nothing captures my return series. It’s mostly alpha!

Factor Model Regression Results: ERN put selling vs. equity, bond, gold, and Fama-French style factors. 1/2018-2/2024.

It doesn’t necessarily mean the intercept is genuinely due to skill, though. That’s because of (at least) two reasons: 1) the alpha may be beta in disguise, i.e., we might have forgotten to include all relevant market betas and styles. 2) the alpha could still be due to luck. Item 2 is easy to address; I ran statistical tests to confirm that the intercept was highly significant. Item 1 is a bit more challenging; we can include a bunch of regressors as in Model 4. But who knows what other factors I might have missed. Please let me know if anyone wants to run their regressions on their factors, and I can provide my return series.

How Options Trading fits into my portfolio

I must stress that nothing I’ve posted here so far means that we should all abandon our existing portfolios and go all-in with trading options. Quite the opposite, I always saw my options strategy as supplementing my existing portfolio, which is perfectly aligned with the standard passive index fund philosophy. Think of me as a Boglehead with a sense of adventure.

I’d also never recommend using excess leverage. For example, if the strategy did so well with 7.7% return and 2.5% risk, why not run this with an additional 10x leverage and make 77% returns with 25% risk? What can possibly go wrong? Check my post on the “optionsellers” debacle again!

I don’t even assume that the 3+ IR will last forever. But even assuming a rather mundane expected return of 2% and risk of 2%, thus, (thus IR=1.0) will generate impressive results. Let’s look at the following numerical example. Imagine we have stocks, bonds, and short-term fixed-income assets with the following expected returns and standard deviations:

  • Stocks: Expected return/risk = 8.5%/16.0%
  • Bonds: Expected return/risk = 4.5%/6.0%
  • Cash/risk-free expected return = 3.25% (currently much higher, I know, but we need to factor in that the Fed will lower interest rates soon, so a 10-year average cash return is likely lower than today’s 5%+)
  • Options-trading alpha: Expected return/risk = 2.0%/2.0% (For example, assume a 5% return and 5% risk in the taxable account. But the taxable account is only 40% of the total portfolio; thus, the options add only 2% alpha to the overall portfolio.)

Also, assume that the stocks-bond correlation is +0.1, the stock-options correlation is 0.5 (higher than my actual correlation, but I want to be on the cautious/conservative side), and the bonds-options correlation is 0:

Efficient Frontier Return Assumptions

Before adding options to the picture, let me plot the efficient frontier of S/B portfolios. See the chart below. Being a math stickler, I insist on drawing the efficient frontier only up to the min-vol portfolio. I don’t draw the parabola all the way to 100% bonds because the backward-bending parabola with less return and more risk is no longer efficient:

Efficient Frontier: S/B only.

Now, let’s add the 2% extra expected return from the options trading strategy. That’s a substantial move in the efficient frontier!

Efficient Frontier: S/B plus options.

Do I get 2% extra expected returns for free? Not exactly. Due to the correlation between the options trading and your stock portfolio, going from, say, an 80/20 portfolio to 80/20 plus 100% options will give you 2% extra expected return but also more expected risk, hence the move to the Northeast direction (more like NNE, actually); see the efficient frontier plot below. If I like to keep the same expected risk, I’d then move along the red efficient frontier back to about 72.5% equities and 27.5% bonds. I’d have the same risk but only about 1.7% extra expected return at that point. But a 1.7% extra return is nothing to scoff at. Not even a 100% equity portfolio would have accomplished that on the Baseline Efficient Frontier.

Efficient Frontier: S/B plus options. Start with an 80/20 portfolio, add options, and de-risk to 72.5/27.5/100.

Because this issue came up in last month’s post, with a little bit of financial engineering, we can even push the efficient frontiers a bit higher if we take the Max-Sharpe-Ratio portfolio and lever that up; hat-tip to Dr. Cliff Asness at AQR Capital Management. So, I also include those efficient frontiers for the math and finance wizards. The leverage-based frontiers do a bit better, but the more significant boost in the return/risk tradeoff still comes from the options trading alpha!

Efficient frontier: Max-Sharpe portfolios plus leverage. Note: for any leverage level greater than 1x, I assume that there is a marginal 30bps (0.30%) drain from employing futures. In other words, the blue line has a slope slightly lower beyond the green dot to account for the leverage costs.

If you don’t like my return assumptions and correlations, here’s a link to a Google Sheet you can use. As always, you must create your own copy of the sheet before editing anything!

Why trading options is an excellent FIRE tool

What kind of an impact would a 1.7% extra return have on retirement planning? Imagine you plan to withdraw 4% under the baseline, as recommended by the naive 4% Rule of Thumb. You might have to do a more detailed, personalized analysis – see Part 28 of my SWR Series for a free Google Sheet retirement simulation tool. But for simplicity, let’s run with the dumb 4% rule. If you can raise your safe withdrawal rate to 4%+1.7%=5.7%, that’s a 42.5% increase in your retirement budget. Not a bad retirement boost. Instead of 25x annual expenses, you target only about 17.5x to retire.

How much of a difference would 1.7% make during accumulation? Assume we have a FIRE enthusiast planning to save $3,000 a month for the next 15 years. With a 1.7% extra return, how much faster to accumulate during those 15 years? An assumed 5% annualized return in the baseline would accumulate to just under $800k after 180 months. With a 6.7% average return, you’d expect just above $900k, or about 14.8% more than in the baseline. And the combined effect of 14.8% more accumulation and 42.5% more withdrawals yields a (compound) 63.59% increase in your retirement budget. Sweet!

FIRE sample calculations. 15 years of accumulation, 4% Rule SWR after that. All returns are real (CPI-adjusted).

Why did we not boost the retirement nest egg by (1.067/1.05) 15-1=27.2 %? The answer is simple: this is not a buy-and-hold investing calculation. Because we regularly contribute to the retirement portfolio, only the first monthly contribution would grow to 27.2% more, but subsequent contributions have less time to enjoy higher returns. Hence, there is a nontrivial but still slightly underwhelming impact on the retirement accumulation part. You’d get better results over 40 years, i.e., the traditional retirement planning horizon.

My takeaway: for accumulation, the alpha boost from options trading is not as useful for early retirees. Sure, 1.7% compounded over 15 years amounts to an additional 14.8%. But the real impact comes in retirement when you can raise your withdrawal rate by about 1.7 percentage points. For example, I did not start trading options until 2011, seven years before retirement. And I did it on a small scale only.

So, if you’re not yet retired and have a relatively small nest egg, maybe don’t worry about options trading yet. You’d also need a minimum account size of $110,000 to qualify for portfolio margin. But options trading is certainly a powerful tool once you are close to or in retirement! There is no reliable way to get around Sequence Risk in retirement. Maybe a glidepath can alleviate a small portion of the risk. The only viable solution to retirement headaches is to raise the expected return. Everything else, like bucket strategies, etc., is wishful thinking and window dressing.

How to deal with objections

1: Options should have a zero expected return.

This is an issue I frequently encounter, for example, years ago in my appearance on the White Coat Investor Podcast (in the recording at about the 50:10 mark). Let’s go back to finance fundamentals to prove that options cannot all have a zero expected return. For example, let’s use the well-known Put-Call-Parity equation. If we buy a call and sell a put option with the same strike and hold the notional capital in a risk-free asset, e.g., T-bills, we have generated a synthetic version of the underlying. To squeeze out any ill-gotten profits, the following non-arbitrage condition must hold:

Call – Put + Risk-Free Asset = Underlying

For the options trading pros, we can even write this without the risk-free asset return if we’re trading futures, i.e., we can generate a synthetic futures contract with a long futures call option and a short futures put option.

Actually, the put-call parity equation is even an identity, i.e., no matter how the market evolves, the synthetic stock will consistently track the underlying, so we could even replace the “=” sign with an identity sign (“≡”). Because of that identity, we can also write the put-call parity in expected return terms as:

E(Long Call) + E(Short Put) = E(Equity Premium)

If we believe there is a positive excess return of equities over risk-free assets like T-Bills, the sum of the two options trading flavors, long calls plus short puts, should also have that same positive return. Thus, options can’t all have zero expected returns. You must be compensated with positive expected returns for exposure to risky equities, whether you hold risky stocks or options.

2: The market is efficient.

Related to the issue in part 1, people often point out that efficient markets negate the attractiveness of options vol sellers. I beg to differ. Returning to the Equity Premium composition, i.e., the equity premium is the sum of the downside risk premium plus the upside risk premium. Which side of the equity premium is better compensated, the downside or the upside? Do I need compensation for a call option payoff profile where I participate in all the equity upside but none of the downside? Likely not. In fact, quite the opposite, this type of positive skewness, lottery-like payoff will offer low, no, or even negative compensation, not despite but because of market efficiency. I.e., you normally pay a premium to participate in a lottery. See Dr. Antti Ilmanen’s FAJ paper for a great discussion.

If I rearrange the put-call parity equation as

E(Short Put) = E(Equity Premium) – E(Long Call)

… and the Long-Call is costly on average, then I get an expected return E(Short Put) > E(Equity Premium). So, selling insurance must be a profitable business, likely more profitable than equities, especially as a multiple of the standard deviation.

3: Black Swan events.

I agree that option selling, if done wrong, will lead to ruin during a “Black Swan” event, i.e., an unexpected and significant economic/financial shock like the pandemic or the Global Financial Crisis. For that exact reason, I’ve showcased in numerous posts over the years how not to run a short-vol strategy:

What all these accidents have in common is that shorting long-dated options (or VIX futures) can go awry during black swan events. With a market move and a vol spike large enough, that short option can lose a ton of money compliments of the options Greeks, especially Delta (change in the option price per unit of underlying change), Gamma (change in the Delta per unit of underlying change), and Vega (change in the option price per unit of implied vol increase)! And I know, Vega is not even a Greek letter! But that’s much less of a headache for 0DTE and 1DTE options. You don’t go from a 2006-style stock market to the 2008 Lehman Brothers failure literally overnight. You don’t go from a 2019 market to a March 2020 pandemic market overnight. The volatility usually builds over time. So, with 1DTE and 0DTE contracts, you successively sell strikes farther out of the money. When March 2020 came along, all the calm weather options you sold in 2019 had already expired, and the options you sold the day prior were so far out of the money that even the 12% drop on 3/16/2020 didn’t get close to my put strikes. 0DTE and 1DTE options work beautifully during those Black Swan events, while long-dated Short-vol strategies get clobbered. If the UBS strategy hadn’t sunk in 2019, it would have failed even more spectacularly in 2020.

So, to sum up, my live trading survived the bear market in 2022, the black swan in 2020, the volatility spike in 2018, and several other crazy market moves before then, like the Brexit vote in 2016, the 2015 Chinese devaluation, the 2011 US debt downgrade, and a few more. I ran some backtests, and I would have done great during 2008 as well. Trading options can succeed even in Black Swan scenarios!

4: Negative Skewness

A valid concern is that the standard deviation may be an incomplete measure of the risk of many options strategies in light of negative skewness. I agree. That’s precisely the reason why some of the longer-dated option-selling strategies fail. You can suffer left-tail losses that are severe enough never to recover, see the XIV ETF debacle I mention above. A sequence of several bad days will sink your longer-dated short-vol strategy. But with 0DTE and 1DTE contracts, you reset the strike constantly.

For example, my daily return skewness is -2.4, but my monthly and annual skewness are each almost back to zero; -0.1, to be precise. That’s less skewness than even the S&P 500. Thus, resetting each gamble each day (or even twice a day with 0DTE and 1DTE contracts), you start getting the benefits of the Central Limit Theorem, and your longer-range returns become more Gaussian-Normal. Not so for your long-dated short puts because your daily returns are no longer uncorrelated thanks to the options Greeks.

So, my response to the skewness concerns: I share those concerns, but if you can keep day-to-day returns “mostly” independent of each other by keeping your DTE as short as possible, skewness washes out over longer horizons. You’re barking at the wrong tree. Your and my equity index funds likely have the same or worse skewness stats over longer horizons!

5: But, but, but… I read somewhere that negative skewness is terrible!

I might be beating a dead horse now, but I would like to bring up another issue and misunderstanding of the “negative skewness apostles” out there. In fact, let’s assume that even when averaging my 0DTE/1DTE option trading profits over months, quarters, and years, I still maintain negative skewness. Even more negative than equities. Some mathematically illiterate folks will tell you that that’s bad and you should thus avoid options trading. But that’s a fallacy, and I would like to demonstrate it with a simple numerical example. Imagine someone offering me the following gamble: With a 99% chance, I make a +1% return today on my total net worth. And with a 1% chance, I make only +0.9%. Would I take that gamble? Absolutely! It’s an almost guaranteed return of 1% in one day, and even in the worst case, I still make +0.9%. What a fantastic deal!

But, of course, reality often creates more complicated tradeoffs. How about if the 1%-chance outlier gives you a 0% return? Still attractive! Or -10%? Still acceptable! At what point would I say, “No, Thank You”? At -20%? Or -50%? Or -100%? To help with that decision, let me display the return stats of these various gambles in the table below: mean return, standard deviation, and the skewness of the return distribution. Unsurprisingly, the worse we make the worst-case 1%-chance outcome, the lower the expected return and the higher the standard deviation. But did you notice what didn’t change? The skewness is the same for all. Skewness alone cannot guide us in determining what’s too risky.

Return Stats for 0.99 vs. 0.01 probability gamble with different worst-case outcomes.

So, what causes this quirky result? Skewness is a unitless measure of how lopsided the distribution’s tails are. Unitless because in the skewness formula, you calculate the third central moment of the distribution in the numerator but then divide again by the standard deviation-cubed.

Because you normalize by the standard deviation, all six gambles must have the same skewness. If someone tells me I should ignore standard deviations and look only at the skewness, I have to roll my eyes. The answer should be more nuanced. I want to look at the skewness and standard deviation in concert. For example, I would probably pass on the gamble with the minus 20% downside risk. Sure, it has a standard deviation of only 2.1% (about twice the long-term average daily stock market volatility) and around 30x(!) the average daily stock return. But why would I risk 20% of my net worth for a measly +1% on the upside? So, I agree that the mean and standard deviation alone are not very useful when you can have a ten-sigma downside. I would certainly agree if the gamble involved an expected return of +0.79% with a 2.1% standard deviation and zero skewness or equity-like skewness of around -0.50.

We are so Skewed!

In other words, negative skewness is only a problem if the standard deviation is large enough that you wipe out your portfolio beyond repair. If you have skewed returns that occasionally give you a minus 10-sigma event, but that minus 10-sigma event still leaves your portfolio largely intact with a potential to recover in 3-6 months, then I’m completely fine. And that’s why I sleep peacefully with negative skewness! Scaling your bets is crucial. With an appropriate risk model and risk controls, you can and should accept negatively skewed returns.

6: I’m a glorified mutual funds salesman (a.k.a. financial planner) and don’t want my clients to know about options trading!

Yeah, I met those folks on Twitter, too. You’re beyond saving. I feel sorry for your clients. I showed a way to generate alpha with a very impressive IR. If you don’t find that IR intriguing, it says more about your skills than mine. And let’s not forget, you also generate 1%+ alpha. Unfortunately, it’s minus 1%+ alpha in the form of an AUM fee.

7: Aren’t you afraid of a repeat of October 1987?

My response is that historically, significant S&P 500 drops don’t occur out of the blue. A significant drop normally (not always) occurs when implied volatility is already elevated. Qualitatively, that’s a good answer. As suggested by a longtime reader, Figuy, it would be nice to quantitatively understand how likely a significant one-day move is, conditional on different VIX regimes. For example, what were the worst historical one-day drops conditional on the VIX index hovering around just under 15 (as in the March 2024 environment when Figuy asked that question)?

I looked at the daily S&P 500 returns since January 1987 and the VIX level on the previous(!) day. So, I pair each daily return between T and T+1 with the VIX at the close of date T. We don’t want to pair T to T+1 returns with the T+1 VIX level because a deep dive in date T+1 would also raise the VIX. Instead, we want to know how much we can glean from today’s VIX level about the prospect of a significant S&P 500 drop tomorrow.

Notice that the VIX index started in 1990. Before that date, I used the alternative, differently constructed but highly correlated VXO index to backfill the first four years of data to capture the all-important 1987. I bucket the returns and the VIX into different intervals. For returns, I am only interested in the downside. So, out of 9000+ observations, I calculate the following matrix. The return bucket (in %) are the rows and the VIX/VXO buckets (in points) are the columns:

Counting the occurrences of S&P 500 returns (rows) vs. VIX levels the prior trading day (columns). Pre-1990, I used the VXO Index.

The worst return (more than 20% down) occurred on October 19, 1987. There was no drop between 15 and 20%. The second-worst drop occurred in 2020 during the height of the pandemic bear market. We have all the usual suspects in the category of 7-10% drops, i.e., another drop the week after the 10/19/1987 fall, four drops during the Global Financial Crisis, and two during the pandemic. All of them with high VIX levels the day prior. You would not have sold put options only 3-4% out of the money during that time (if the 1DTE options had been available then).

The worst S&P 500 daily returns since 1987 vs. the implied volatility the prior day. All significant drops have in common that the implied volatility measure was elevated the day prior.

That said, if we look at the still-very-painful S&P drops in the 5-7% range, there was one day in 1989 when the VXO was below 20. So, there have been some out-of-the-blue drops in the index, but they are rare. Again, many significant market drops during the Global Financial Crisis happened when Vol was already very high.

In the 5-7% S&P drop bucket, there has been only one occurrence with an implied vol level below 20.

Finally, I also want to display the conditional empirical probabilities. So, conditional on being in a certain VIX regime, what were the empirical probabilities of large S&P 500 drops:

Empirical probabilities in % of falling into the different return buckets, conditional on the past VIX bucket. S&P 500 returns (rows) vs. VIX levels the prior trading day (columns). 1/2/1987-3/20/2024.

After this long-winded analysis, Can we experience a major drop that will knock out my current (March 20, 2024) puts? Yes, for sure. But historically, the largest index drops occurred when the VIX was already far more elevated than the current level (13-14 in March 2024).

All options trading posts so far

Other Related Posts

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.