Butterfly Skepticism (Part 4)

Today I want to complete discussion of the protective put (PP) butterfly adjustment.

I might be able to come up with some workaround (as done in this second paragraph) for PP backtesting. I could look at EOD [OHLC] data and determine when the low was more than 1.6 SD below the previous day’s close. In this case, I could purchase the put at the close. This would bias the backtest against (not a bad thing) the adjustment in cases where the close was more than 1.6 SD below because the put would be more expensive.

Unfortunately, I am not sure this particular workaround would work. If the close is less than 1.6 SD below then the backtested PP would be less expensive than actual. Furthermore, if I waited until EOD then the NPD and corresponding PP(s) to purchase would be different. This would distort the study in an unknown direction. I could track error (difference) between -1.6 SD and closing market price. Positive and negative error might cancel out over time. If I had a large sample size then this might or might not be meaningful.

At best, this workaround seems like a questionable approximation of an adjustment strategy that is precisely defined.

Before dismissing the PP out of frustration, let’s step back for a moment and piece together some assumptions.

First, I believe the butterfly can be a trade with somewhat consistent profits and occasionally larger losses. Overall, I’m uncertain whether this has a positive or negative expectancy (hopefully to be determined as I begin to describe here).

Second, as butterflies are held longer, I believe profitability will be decreased. I have seen some anecdotal (methodology incompletely defined) research to suggest butterflies are more profitable when avoiding periods of greatest negative gamma.

Third, I have seen anecdotal (methodology incompletely defined) research to suggest PPs as unprofitable whether:

• purchased in high or low IV
• purchased at specified deltas
• purchased for specified amounts
• held to expiration, managed at specified profit targets, or managed at 50% loss

Fourth, this adjustment will require any butterfly to be held longer on average. The additional time will be needed to recoup the PP loss. The result will be, as described per second assumption above, decreased average profitability.

In my mind, combining the first and fourth assumptions does not bode well.

The big unknown involves the magnitude of the largest losses and in what percentage of trades those largest losses occur.

Interestingly, the trader who explained this to me said PP will lose money in most cases. What it can prevent is a massive windfall loss. Being forthright [about the obvious?] may give the teacher more credibility. Without backtesting, though, I think it leaves us with more than reasonable doubt over whether this approach tends toward profit or loss.

Butterfly Skepticism (Part 3)

On my mind this morning is skepticism regarding the protective put (PP).

I have seen the PP lauded by many traders as a lifesaving arrow to have in the quiver.

One trader described this to me with regard to a butterfly trading plan. Part of the plan provides for the following adjustment:

1. If NPD is at least 10 with market at least -1.6 SD intraday then record NPD.
2. Buy PP(s) to cut NPD by 75%.
3. On a subsequent big move, if NPD again reaches value recorded in Step 1 then repeat Step 2.
4. If market reverses to the high of purchase day, then close PP(s) from Step 2.

Upon further questioning, I got some additional information. He learned it from a guy who claims to have “mentored” many traders. The mentor (teacher) claims to have seen many lose significant money in big moves and therefore recommends this to avoid windfall losses. The teacher has shown numerous historical examples where this adjustment would keep people in the trade (not stopped out at maximum loss) and often wind up profitable. Further prodding revealed uncertainty over whether these “numerous” examples amount to more than a handful of instances. Of the several times the adjustment has been presented, he acknowledged the possibility that many could have been repetition of the same [handful of] instance[s]. He is uncertain whether anyone has presented big losing trades more than once.

Much of this casts doubt over the sample size behind this adjustment. We certainly wouldn’t want to fall prey to that described in this this second-to-last paragraph.

As described in this third paragraph, the PP is simply an overlay added later in the trade. This strategy also has its own catchy name and is marketed. In order to backtest, we could study the profitability of long puts purchased on days the market is down at least 1.6 SD (also explore the surrounding parameter space as discussed in the fourth complete paragraph here).

From a backtesting perspective, intraday is a huge wrinkle. Technically, I’d need intraday data to identify exactly when the market was down 1.6 SD in order to purchase the PP at the correct time. As mentioned in the second-to-last paragraph here, this is arguably another reason why certain trading plans cannot be backtested: data not available.

I will continue next time.

Constant Position Sizing of Spreads Revisited (Part 4)

Today I will conclude this blog digression by deciding how to define constant position size, which I believe is important for a homogenous backtest.

The leading candidates—all mentioned in Part 3—are notional risk, leverage ratio, and contract size.

Possible means to achieve—both mentioned in Part 2—are fixed credit and fixed delta.

I thought it might be the case that fixed delta results in a fixed leverage ratio. I suggested this in the last paragraph of Part 1 where I asked whether fixed delta would lead to a constant SWUP percentage. For naked puts under Reg T margining, gross requirement is notional risk. For spreads under Reg T margining, notional risk is spread width x # contracts and while notional risk may be fixed, the SWUP percentage varies.

Speaking of, we also have Reg T versus portfolio margining (PM) to complicate things. Both focus on a fixed percentage down (e.g. -100% for Reg T vs. -12% for PM) on the underlying. However, PnL at -12% can vary significantly with underlying price movement. PnL for spreads at -100% will not change as the underlying moves around because the long strike—at which point the expiration risk curve goes horizontal to the downside—is so far above.

Implied volatility (IV) also needs to be teased out since it will affect some of these parameters but not others. Given fixed strike price, IV is directly (inversely) proportional to delta (relative moneyness). For naked puts assuming constant contracts and fixed delta, IV is inversely proportional to notional risk and to leverage ratio. IV does not relate to leverage ratio for spreads, which is net liquidation value (NLV) divided by notional risk as defined two paragraphs above in the last sentence.

After spending extensive time immersed in all this wildly theoretical stuff, I seem to keep coming back to notional risk, leverage ratio, and fixed delta. The first two vary with NLV* and with # contracts due to proportional slope of the risk graph. Number of contracts can vary to keep notional risk relatively constant as strike price changes but this applies more to naked puts and less to spreads where spread width is of equal importance.

I want to say that for naked puts, the answer is fixed notional risk (strike price x # contracts), but we also need to keep delta fixed to maintain moneyness. With fixed credit, changing the latter would affect slope and leverage ratio. This is how I described the research plan originally and we will see whether an optimal delta exists or whether results are similar across the range. In the midst of all this mental wheel spinning, I seem to have gotten this right for naked puts without realizing it.

I guess I have also lost sight of the fact that this post is not even supposed to be about nakeds (see title)!

Getting back to constant position sizing of spreads, I think we can focus on notional risk and moneyness but we should also factor in SWUP. As the underlying price increases (decreases), spread width can increase (decrease) and we will normalize notional risk by varying contract size. Short strikes at fixed delta will be implemented and compared across a delta range.

Which is what I had settled on before (for spreads)…

[To reaffirm] Which is what I had settled on before (for naked puts)…

As I unleash a gigantic SIGH, I question whether any of this extensive deliberation was ever necessary in the first place?

I think at some level, this mental wheel spinning is what I missed as a pharmacist. The complexity fires my intellectual juices and is great enough to require peer review/collaboration to sort through. Once that is done, selling the strategy is an entirely separate domain suited to different talents, perhaps.

I left a job of the people (co-workers/customers) for a job that begs for people, which I have really yet to find. Oh the irony!

* By association, this is why I stressed magnitude of drawdown as a % of initial account value (NLV) in previous posts.

Constant Position Sizing of Spreads Revisited (Part 3)

Happy New Year, everyone!

The current blog mini-series has been a tangent from the automated backtester research plan. Today I will discuss whether fixed notional risk—with regard to naked puts and spreads—is even important.

This issue is significant because it seems like fixed notional risk is the “last man standing” since I initially mentioned it in Part 1. I have reassessed the importance of so many concepts and parameters in this research plan. The fact that they get misunderstood and reinterpreted is testament to how theoretical and highly complex they are. Especially from the perspective of avoiding confirmation bias, I believe this is all debate that must be had, and a main reason why system development is best done in groups as a means to check each other.

The reason fixed notional risk may not matter is because leverage ratio can vary. I also mentioned this in the third-to-last paragraph here. Leverage ratio is notional risk divided by portfolio margin requirement (PMR). Keeping PMR under net liquidation value and meeting the concentration criterion are essential to satisfy the brokerage. Leverage ratio can be lowered by selling the same total premium NTM. This will affect the expiration curve by decreasing margin of safety as it lifts T+x lines. Analyzing this, somehow, might be worth doing if backtesting over a delta range does not provide sufficient comparison.

Whether “homogeneous backtest” should mean constant leverage ratio throughout is another highly theoretical question that is subject to debate. Keeping allocation constant, which I aim to do in the serial, non-overlapping backtests, is one thing, but leverage can vary in the face of fixed allocation. I discussed this here in the final four paragraphs. In that example, buying the long option for cheap halves Reg T risk but dramatically increases the chance of blowing up (complete loss) since the market only needs to drop to 500 rather than zero. While the chance of a drop even to 500 is infinitesimal, theoretically it could happen and on a percent of percentage basis, the chance of that happening is much greater than a drop to zero.

Portfolio margin (PM) provides leverage because the requirement is capped at T+0 loss seen 12% down on the underlying. In the previous example, 500 represents a 50% drop. Even under PM, though, leverage ratio can vary because of what I said in second-to-last sentence of paragraph #4 (above).

When talking just about naked puts, much of this question about leverage seems to relate to how far down the expiration curve extends at a market drop of 12%, 25%, or 100%. This brings contract size back into the picture because contract size is proportional to downside slope of that curve.

With verticals, though, number of contracts is less meaningful because width of the spread is also important. The downside slope will be proportional to number of contracts. The max potential loss of the vertical depends not only on the downside slope, but for how long that slope persists because the graph only slopes down between the short and long strikes.

Either way, you can see how number of contracts gets brought back into the discussion and could, itself, be mistaken as being sufficient for “constant position size.”

I certainly was not wrong with my prediction from the second paragraph of Part 1.

February 2018 Incident Report (Part 2)

Today I conclude documentation of catastrophic losses suffered earlier this year.

February 2018 was not the first time as a full-time trader that I seriously contemplated going back to find a corporate job [perhaps as a pharmacist]. I felt “too old for this”—like I might not be able to mentally survive another bout of catastrophic losses. At the same time, I realized that unless I changed my trading strategy the only thing practically guaranteed was that I would run into a similar (or worse) situation at some future date. Those are some painful working conditions, indeed.

Many articles cite a statistic like “90% of all traders fail.” Although I have questioned the validity of this claim, I can definitely believe it. Over the last 10+ years I have realized that traders rarely talk about losses. I believe many people who suffer catastrophic loss throw in the towel for good and go quietly into the night. I think I know the feeling because contemplating career change means being on the precipice myself.

After February 9, I sat on the market sidelines for nearly three weeks before slowly starting to dip my toes back in the water.

Seven months have passed and for the moment the market trades within normal limits. Bid/ask spreads are pretty much typical. VIX is trading within a range. The market is not crashing.

I did make some changes to my strategy. One big change was to decrease the likelihood that I will have to deal with so many losing positions at once. Increased granularity is better with regard to diversification but worse with regard to execution complexity. This is bad in fast moving markets because slippage increases dramatically.

A second change was to shift positions out in time. I thought this would also give more time to adjust when the market goes against me. In addition to decreased liquidity, I have since realized this approach will require closing positions due to margin concentration. While I haven’t completely resolved this, I know two things: (1) the market can trade “normally” for long periods, which allows me to think about it; (2) I better figure it out ASAP because despite (1), the market can crash at any time and thrust me right back into February conditions. “Up the river without a paddle” would be a very bad outcome.

Reliving traumatic events is a very difficult thing to do and these have been difficult posts to write. In effect, I have taken seven months to document this. It’s always easier to reflect on the painful past in times of plenty and I do think “better late than never.” The downside to delay is fading memory of events experienced in the moment but I think I have done a pretty good job remembering the salient points here.

And I certainly know what needs to be done next.

February 2018 Incident Report (Part 1)

This will serve as my incident report from February 2018.

In retrospect, warning shots were fired on the first Friday of the month. According to Yahoo! Finance, VIX increased from 13.47 to 17.31 (close-to-close) on this day: an increase of 28%.

The following Monday, February 5, is when the market went completely to hell with VIX closing at 37.32, according to Yahoo! Finance: an increase of 115%. This is largest single-day VIX percentage increase since calculations are available (1990).*

For this reason, all I did after 2:00 PM was close losing positions at [far beyond] maximum loss. Bid/ask spreads were gigantic, which forced me to lose a significant sum to transaction costs (slippage). It felt like playing whack-a-mole for the better part of two hours. I would close one position, record relevant notes in my spreadsheet, and then determine what open position was at largest loss and in need of being closed next. Around 3:15 PM, the Schwab website became unresponsive. I had to call in and speak to a representative (thank goodness I was able to do this before hold times became prohibitively long) to get the rest of my trades executed. I was also trading a smaller account, which meant I had to execute mirror trades. During all this time I was envisioning dollar bills (or $20s or $50s) flying out the window. I could detect the shakiness in my voice, and at one point the representative even gave me a couple empathy statements. Between all the waiting for orders to work and eventually fill, I felt as though my breathing had stopped in wait only to resume once the bloodbath had concluded.

I exited positions throughout the week until I was totally flat on February 9. When the smoke had cleared, I lost roughly half my profit from all of 2017 in six trading days with the vast majority of those losses coming in just one.

Perhaps as much as the financial, the emotional repercussions of catastrophic losses like this are why trading is so difficult. My mind was abuzz for several days. I felt like a deer in the headlights even though I was now out of the way. I felt like the proverbial “run over by a steamroller” had come to pass. As a trader, I have encountered catastrophic loss 6-7 times to date and it is always difficult. More than the painful moments of the present, in February it felt like I was simultaneously reliving trauma from catastrophic loss events of the past.

I will continue next time.


* I find it interesting to see different sources report different numbers. Business Insider, for
   example, reported VIX up 84% on the day. A closer look at that article reveals a posting
   time of 3:38 PM, however, which explains why the numbers don’t match. Either way, it
   was the largest single-day percentage increase ever.

Prerequisites for Trading as a Business

Trading as a business means generating consistent profits to cover all living expenses. Usually this is a full-time, independent pursuit. While I don’t think advanced degrees are necessary, I can think of three prerequisites for a trading start-up.

First, some degree of math proficiency is necessary for this level of money management mastery. Arithmetic is most applicable in order to calculate profit targets, stop-loss levels, performance, buy/sell points, position sizing, etc. A more advanced math background can make some derivative concepts more intuitive because option pricing models are based on differential equations. This isn’t mandatory, though; one can trade without having any knowledge of option greeks at all.

The second prerequisite for trading as a business is a strong capacity for critical thinking. No industry may be more littered with fraudulent pitches, false marketing/advertising claims, and overall chicanery than finance. I have written about this subject extensively (e.g. here, here, here, and here). In designing a business plan and sticking to it one must avoid being derailed into the quicksand. A background in statistics is a good defensive arrow to have in the quiver. Statistics provides a solid foundation from which to put questionable performance claims into reasonable perspective.

Critical thinking defends against the death knell that is being defrauded and suffering losses. Even greater than the financial losses may be the psychological damage in terms of trust, confidence, and safety: three key components to a fledgling business effort. I have spoken with a handful of people who have been violated by shady advisers. Once (if) they got out they never wanted (or couldn’t afford) to invest again (also reminiscent of catastrophic loss). I strongly suggest avoiding any black box system or advertisements of unrealistic returns. Steer clear of advisers who charge too much or smooth talkers who seem like they may have cut their teeth selling used cars. These last two sentences deserve separate blog posts of their own.

Finally, trading as a business requires money! Start-up capital is a safety cushion because like many other businesses, beginning traders often lose money (early success may actually be a curse leading to false confidence and a subsequent failure to adequately assess risk). Without this cushion, the stress of being forced to profit to stay afloat with bill payments can make long-term survival difficult. As the learning curve is scaled, magnitude of living expenses will dictate how large the trading account must be. A six-figure trading account might be necessary to cover the house, credit card, and insurance payments without taking on risk so great that Ruin is likely.

Managing Winners (Part 2)

Today I want to conclude my discussion on managing winners.

I can imagine something like the following if I were to start managing winners. Placing trades daily, I would see some days where number of open positions would decline precipitously. Over the next couple weeks I would build back toward a maximum number unless other positions had been managed in the interim. The big losers would be unaffected. Smaller losers would have a better chance at being flipped because they may at some point hit the profit target (think probability of touching). Management of winners (e.g. at 50% max potential profit) means the average win would be smaller.

Taken together, persistent big losers, higher win rate, significantly smaller average win, and equal number of trades all strikes me as a recipe for disaster.

One thing I observe in daily trading without managing winners is a mix of “in-play” and “out-of-play” positions. Recently placed trades are in-play with responsive position greeks. Older trades often carry unrealized gains approximating max potential profit. Even big market moves are unlikely to affect these older, “out-of-play” positions due to negligible, unresponsive deltas, gammas, and thetas. With the exception of MR, the total portfolio acts as if these out-of-play positions were already closed.

Reflecting on these observations makes me thankful the market hasn’t made any large, adverse moves as of late. Such moves tend to increase the number of positions in play up to 100% (managing losers—perhaps with a SL—can decrease total number). I am grateful for the trading experience that allows me to make and analyze these observations because at some point, volatility will return in a big way marking the return of stressful nights [hopefully not for very long]. My goal is to factor all this into the trading system development.

Reflecting on these observations also makes me think that carrying some out-of-play trades is part of the whole design (i.e. perpetual scaling and time diversification). I could close out-of-play positions to cut risk, though. Indeed, I strongly suggest other option traders “do the right thing” and adhere to such a guideline. In effect, closing out-of-play trades is managing winners albeit much less aggressively (e.g. 80%-90% max potential profit as opposed to 50%). The impact on performance metrics is therefore going to be different than that profiled in many of the TT studies.

I will wrap this up with an empirical question for the future. As a more conservative way to manage winners, exiting out-of-play positions may also be conceptualized as buying back-end insurance. How does this compare, I wonder, to buying the insurance up-front in the form of being net long [DOTM] contracts or buying a small number of NTM hedges?

Managing Winners (Part 1)

One direction I may or may not choose to explore at some point is that of managing winners. I have trouble wrapping my brain around this concept and today I will explain why.

Managing winners is a frequently discussed concept by the Tasty Trade (TT) folks but it certainly is not a new idea. In the past this has been described as a “yield grab” or “profit target.” Just like I might manage losers with a stop-loss order, I can manage winners based on a variety of criteria.

The TT philosophy is to manage winners, adjust (or not) losing trades if tested, and trade small. If a trade goes to max loss then I don’t get hurt because my position size is small. Some trades get down significant percentages but then recover to win, which is why they don’t advocate managing losses.

One of the problems I have with this approach is doubt over whether I can “trade small” and still make enough to pay the monthly bills. I could certainly do this with a multi-million dollar trading account. Few retail traders have this luxury, though. I see this as a big problem for anyone looking to trade full-time as a business but it probably doesn’t affect TT viewers very much if my lack of success in finding other full-time traders is any indication. “Trading small” suits part-time/hobby traders who don’t rely on trading profits to cover living expenses.

According to TT “research” (I use that term loosely for reasons discussed here), managing winners increases win rate, number of trades, and profit per day while decreasing standard deviation of returns. They generally conduct studies by comparing trades held to expiration with trades closed at a profit target. In both cases, a new trade is placed after the old trade is closed.

I have tailored my backtesting methods to avoid the pitfall of insufficient sample size by entering new trades every trading day. At some point I realized that as a perpetual scaling technique, this might be a viable approach to live trading. As long as I keep my daily position size small, the total position size may remain reasonable. This has been at the core of my NP research.

I believe it is through this looking glass that I get confused when contemplating management of winners. Managing winners allows for more frequent trade placement but I am already placing trades every day as part of my scaling effort. My number of positions is, in other words, already maxed out. Opening multiple new trades on a favorable market day that sees a number of winners managed (e.g. a big up day for NPs) would compromise my time diversification, which is not something I will pursue.

I will continue this discussion next time.

Professional Performance (Part 2)

Last time I discussed management of other people’s money. Today I want to focus on my recent trading performance.

On September 21, 2015, I began trading my personal account very similar to the way I would professionally manage money for others. I have traded every single day while adhering to a defined set of guidelines for opening trades. The little flexibility I have maintained with regard to position sizing and closing trades would be omitted as a professional money manager. Rather than using discretion, if someone wanted to squeeze out more return then I would discuss the possibility of a larger portfolio allocation to my services.

Here is a graph of net ROI from 9/21/15 through the end of 2016:

Over 15+ months, I have outperformed the index 29.2% to 16.8%.

Here is a graph of maximum drawdown (DD):

In addition to a larger total return, my max DD was smaller than the index: -10.0% vs. -20.7%. February 2016 offered a moderate market pullback and my ability to keep DD in check resulted in a 3.6x better risk-adjusted return (ROI divided by max DD): 2.92 vs. 0.81.

Because ROI and max DD are both a function of position size, I graphed daily portfolio margin requirement (PMR) as a percentage of account value:

PMR ranged from 8.27% to 62.2% with an average (mean) of 28.6% (standard deviation 11.9%).