Naked Call Backtest (Part 5)

As mentioned previously, increasing position size makes this profitable: not anything about the naked call strategy itself.

Do any research on Martingale betting systems and you will see they are not recommended. I wrote about this here. The smaller I start as a percentage of the total bankroll, the lower probability I will run into a string of consecutive trades long enough to go bust. Make no mistake, though: it most definitely can happen [and since Mr. Market “can remain irrational longer than you can remain solvent,” it probably will].

What are some ways a strategy like this may be viably implemented?

One way is to position size as a fraction of the entire account. I did calculations in Part 4 based on $240K risk. If this is 10% (for example) of my total account, then I can rest easy because at absolute, never-before-seen worst, I lose 10%.

Another way to trade this responsibly might be to overlay on top of a long strategy that offsets naked call (or vertical spread) losses when the market rallies. Keep in mind that calls are NTM compared to equal-delta puts due to vertical skew, which means portfolio margin requirements can grow faster. Trading fewer call than put contracts as net short premium is one way this can make sense; just remember the number of call contracts may increase at least 16-fold.

Implementation of stop-losses is another avenue for the naked call (or vertical spread) strategy. By improving the avg win:avg loss ratio, I can mitigate position size increases. Stops increase number of losses, though, because a trade cannot recover once it has been closed. Backtesting is needed to better understand whether one factor is clearly more likely to prevail.

Finally, intraday backtesting (discussed in this blog mini-series) remains undiscovered country. Mine is a once-daily backtest,* which allows up to 24 hours for things to go from bad to worse. The worst backtesting loss is on 4/6/20 when a call sold for $1.35 is closed for $45.30. As mentioned in the second-to-last paragraph here, I think ITM short puts are best rolled before they go OTM. The mirror image dictates rolling short calls before they go ITM. Win percentage would decrease, but magnitude of loss would be lower. Again, backtesting is needed to better understand whether one factor is clearly more likely to prevail.

As discussed in the last paragraph here, by rolling rather than taking on some unknown legging risk and leaving short options to expire worthless, the current backtest errs on the side of conservatism. Rolling involves buying out remaining premium and realizing excessive slippage upon exit, which would both be expected to dampen performance slightly.

* — I used market prices at 3:50 PM until Dec 2020 and at 3:45 PM daily thereafter.
       This was due only to inadvertent oversight. The catch-22 is to use data as
       close to expiration (4:15 PM ET) while not suffering widened, distorted bid/ask
       spreads often seen after normal market close (discussed in fifth paragraph here).

Naked Call Backtest (Part 4)

Do you take it or leave it? I say the latter. With a couple changes, though, this trading strategy becomes a bit more enticing.

All of this exempts previously-addressed caveats having to do with dynamic strategy guidelines:

1. Use OptionNet Explorer (ONE) between 3/23/2020 – 7/1/2021.
2. Sell 10 naked calls in nearest weekly expiration (1 – 4 DTE) for just over $1.00/contract.
3. Assess transaction fee of $16/contract for slippage and commission.
4. Monitor market and/or trade once daily at 3:50 PM ET (3:45 PM ET starting Dec 2020).
5. If ITM, roll to following weekly expiration as far OTM as possible for credit; increase size with discretion.
6. Otherwise, rinse and repeat on expiration day with 10 new contracts.

With regard to (2), the target premium is $1.00 but I aim for ~$1.30 – $1.40 to offset transaction fees. If I have to pay material premium to close (e.g. more than $0.05) then I add to target premium in order to net target premium. For example, if I BTC for $0.75, then I will look to sell the next call for ~$2.00.

With regard to (5), I aim to increase size and leave sufficient cushion in case the market continues upward. What constitutes “sufficient” is what makes this discretionary.

Here are the results:

Shocking improvement from Part 1! This equity curve climbs consistently with eight large, brief pullbacks. The average loss is about seven times the average win. While not so great, the strategy wins 93% of the time.

The maximum drawdown (MDD) occurs here:

Albeit confusing, SPX price increases 10% from April 3 – 14, 2020. I could have rolled the initial 10 to 40 instead of 20, but I would still have lost a second time and had to roll to 80 contracts. A market pullback allows for a profitable exit.

Is this a viable strategy for live-trading? Follow the logic:

• Contract size increases a maximum of eight-fold (significant improvement from 64-fold seen in Part 1).
• Err conservatively by position sizing for a 16-fold increase (one additional consecutive loss).
• Err conservatively by utilizing a 150-point call credit spread that costs an additional dime.
• Account for spread expense by subtracting ($0.10 + $0.16*) x 192 trades from $22,504 backtesting net profit.
• Max margin $15K/contract * 16 contracts = $240K.
• Net profit ~$17,500 in 16 months (rounding up) is ~5.6% annualized return (rounding down).

Given this may supplement a long portfolio, I am somewhat encouraged especially in comparison to the monthly approach.

Unfortunately, increasing position size is what makes this strategy work—not the strategy itself. When I apply constant position size throughout and add the additional cost to spread off risk, I get a profit factor of 1.03: dismal at best.

But, but… I’ll take any bonus return since I’m trading this sufficiently small to support a 16x increase in position size!

Is it conceivable that we could have more than four consecutive big up periods maybe lasting two weeks or more?

Could I go bust as a result?

Yes. Absolutely.

I will continue next time.

* — This is a very conservative estimate; with the option being $0.10, slippage would actually be half that.

Naked Call Backtest (Part 3)

Before proceeding with more backtesting data, I want to clarify some previous points and make a casual observation.

Position sizing based on max drawdown (DD) says “the largest DD I ever saw in backtesting was X. If I never want more than a Y% DD on my account, then minimum account size is X / (Y / 100).” As an additional margin of safety, suppose the largest future DD will exceed max DD by a factor of Z. The minimum account size becomes (X * Z) / (Y / 100).

Given the $407K MDD (one contract initial size), preference not to see my account down more than 20%, and anticipation of future max DD 2x worse than that seen in the past, minimum account size should be ($407K * 2) / (20 / 100) = $4.07M.

Even if I can position size this way based on max DD, portfolio margin requirements (PMR) could still be a limitation. As account equity decreases and margin limits approach, I may not be able to continue holding the position. Recall from Part 1 that depending on index value, PMR is $1.7M – $3.3M when starting with one contract. A $4.09M account suffering a 40% DD becomes $2.45M, which can no longer support a $3.3M PMR.

DD and position size both determine whether an account can support a position. Position size is proportional to PMR. DD affects buying power. A margin call will be issued by the brokerage as buying power approaches [and certainly slips below] PMR forcing clients to deposit additional funds or to close positions (else the brokerage will do so for them).

Vertical spreads cap PMR, which is still proportional to index price. We probably need to determine the larger of max PM or max DD (along with associated fudge factors discussed in second paragraph) and base position sizing off that.

Large variance across IV and index price level demands more dynamic guidelines for normalization. This backtest implements a static premium in the face of wide-ranging underlying price and volatility. Implementing a static delta, instead, would allow for a normalized strategy that proportionally self-adjusts over time.* If I were doing spreads, then width could be better defined [dynamically] in terms of index price since a static 200 points is 10% of 2000 and only 5% of 4000.

This can all be confusing, which is why I wrote an entire mini-series on it.

As a final note, this backtest closes all short options no later than 3:50 PM on expiration Thursday. In live trading, the short option can often be left to expire thereby saving premium and transaction fee. A cursory review of current results reveals a difference up to $15K: sizeable, to be sure, but amounting to little more than a rounding error all things considered.

*—Added to my to-do list.

Naked Call Backtest (Part 2)

I left off analyzing naked calls with respect to portfolio margin (PM) and maximum drawdown (MDD).

The strategy could be implemented based on maximum drawdown (MDD), but margin is another clear limitation. Even one contract could require up to $3.3M in PM for ~$108K profit. This paltry 0.34% total return gets even smaller as the index continues to climb. Not worth it!

One thing I could do to decrease PM would be to sell the vertical spread instead of naked call. Going 200 points wide would risk up to $20K/contract. Ten contracts would be up to $200K in PM, but I certainly could not increase 64-fold from there.

The spread concept brings with it some other interesting revelations. The backtest starts in high IV:

Observations:

• 200 points wide, which is $20K/contract risk
• Net credit ~$3.75 (~$4.40 going one strike NTM)
• Well into second standard deviation (SD) OTM
• Short delta ~5 (~5.6)

The backtest ends in low IV:

Observations:

• 200 points wide, which is $20K/contract risk
• Net credit just over $5.00 ($7.25 going one strike NTM)
• Just barely into second SD OTM (last strike in first SD OTM)
• Short delta ~8.6 (~11.4)

High IV is lower risk in terms of delta, but generates less credit because the long option is relatively expensive compared to the short. Low IV collects more credit but does so at a relatively higher delta, which is more risky.*

These examples illustrate two major differences: IV and index price. With regard to the latter, selling a fixed dollar amount is relatively NTM (higher delta) for index at lower value compared to relatively OTM (higher delta) for index at higher value.

To see what SPX around 2300 would look like in low IV, we can go back to 12/23/16:

Observations:

• 200 points wide, which is $20K/contract risk
• Net credit $5.50 (~$2.90 going one strike OTM)
• Last strike within first SD OTM (just into second SD)
• Short delta 14.5 (8.7)

The first two examples suggest higher IV may hurt credit, but now we see that lower IV also results in lower credit if sold at a comparable delta value.

Also note that I can cut risk 50% by purchasing the long leg 100 points NTM for only $0.45 more.

I will continue next time.

* — I have discussed two different kinds of risk here. Spread width defines risk as the maximum
       loss for any given trade. Delta represents risk by approximating probability of expiring OTM
       (full profit) as opposed to ITM (likelihood of loss).

Naked Call Backtest (Part 1)

For the last several years, I have not been a proponent of selling naked calls. However, “necessity is the mother of invention.” Volatility does not explode to the upside; how bad can they actually be?

Here are my backtesting guidelines:

• Use OptionNet Explorer (ONE) between 3/23/2020 – 7/1/2021.
• Sell 10 naked calls in monthly expiration closest to 30 DTE for ~$7/contract.
• Use 25-point strikes only.
• Assess transaction fee of $16/contract (slippage and commission).
• Monitor market once daily at 3:50 PM ET.
• If ITM, roll to following month as far OTM as possible for credit and double size.
• Otherwise, rinse and repeat on expiration day.

This is the backtesting period:

To understand how bad naked calls can possibly be, this seems to be a pretty good place to start. SPX 2254 to 4316 is an increase of ~91% in just over 15 months: a staggering [recovery and] ascent! I expect to see losses under these conditions.

Here are the overall results:

This is a strange looking equity curve. It certainly does not match my ideal, which would be upward sloping at 45 degrees. From inception through Sep 2020, this strategy does a wonderful job losing money. From Nov 2020 on, it looks flat. In between the two periods, it makes a ton of money.

Trade statistics indicate this strategy loses more often than it wins and that the average winner is bigger than the average loser. Overall, this shakes out to a profit factor of 1.3.

Take it or leave it?

Let’s look closer to the individual trade results:

The first seven trades are all losses and position size doubles each time. No wonder the numbers on the equity curve get huge. A win finally recoups all losses and then some. Position size remains comparatively low for the remainder of the backtest (10-20 contracts compared to the high of 640), which is why the curve is flat on the right side.

One trade out of 16 saves our bacon.

Actually, what saves our bacon is not the particular trade but rather the position size. Any of the winning trades sized at 640 contracts would generate enough profit to overcome the losses.

We must not forget about risk. Portfolio margin (PM) for naked calls is 12%. With SPX at 2254, this amounts to 2254 * 100 * 0.12 = $27,048/contract. With SPX at 4316, this amounts to 4316 * 100 * 0.12 = $51,792/contract. The initial position size carries a PM requirement between ~$270K – $518K. 640 contracts carries a PM requirement between ~$17M – $33M where
M = million. This backtest includes a 64-fold increase in position size!

At this point, I have to say leave it. Even with one contract, PM requirements are too large and the MDD ($406K) is too big.

I will continue next time.