Musings on Put Credit Spreads (Part 4)

I begin by concluding some analysis of a Tasty Trade (TT) segment from February 9, 2016. I hope to take bits and pieces from this analysis and apply them to my backtesting methodology for put credit spreads.

Aside from mixing different strategies, a second problem I have with the TT research involves duration. All capital should be recycled in roughly 45 days. The tested trades were rolled for an average 35-day increase in duration. I therefore have to position size assuming 80 days in trade. This will significantly dilute returns.

Given a binary choice between diluted returns and win/loss then intuitively, I would choose the former. In addition to increased comfort, lower drawdown (DD) could mean increased position size to somewhat compensate for diluted returns. In backtesting this, I would have to be mindful of maximum DD since these positions are of the “unlimited risk” variety (the TT presentation did not include any DD data).

Continuing with the previous post’s clarification of Part 2, I also wrote this mouthful:

> This is partly why I believe the appeal to complex trade
> methodologies including multi-legged positions… is
> misguided. Large degrees of freedom significantly
> complicates backtesting: curve-fitting or small sample
> sizes muddle the interpretation.

This is called the “curse of dimensionality.” More parameters mean more categories for comparison. With a finite amount of data, this means fewer points in each category. These sample sizes may become insufficiently small and any identifiable conclusions may be curve-fit or fluke rather than representative of a large population.

> People think they are objectively backtesting while future
> leaks and free-range confirmation bias inconspicuously
> oppose them.

A future leak is when knowledge of the future contaminates calculation in the present. Realistically, this can never happen when trading live. I often hear about option traders backtesting trading strategies over recent months. They know exactly where the market is going so the guidelines are less likely to work when the market environment changes.

I see confirmation bias in much of the arrogant, ego-driven talk that I often write about. People present trading strategies that they are determined to defend as if their very lives depended on it. Why?

That’s a whole other blog post.

Musings on Put Credit Spreads (Part 3)

Congratulations are in order if you followed the entirety of my last post. In going back to reread later, I got so confused that I decided to spend some time clarifying.

The straightforward understanding advanced traders have of multi-legged spreads like iron condors and butterflies may be illusory. Not only does design of a spread backtesting approach reveal multiple parameters to consider, trying to execute these beasts in real-time crash conditions can prove detrimental to financial health.

For this reason I prefer backtesting of elementary rather than combinatorial trading systems.

In the last post I wrote:

> Adjustments present like a fly in the ointment. If these are actually new trades
> then why backtest them together as opposed to closing one trade, starting another
> one, and combining the results?

I have reservations about a February 9, 2016, Tasty Trade presentation where a naked put trade is compared to a favorable adjustment strategy rolling the tested put out in time. Two strategies are being combined in a somewhat-arbitrary fashion: an OTM naked put and an ATM [rolled] naked put. Only a fraction of the occurrences required rolling and the positive performance of those ATM puts might be a fluke.

Viewed another way, the rolled put will be profitable if volatility trades sideways or lower. Volatility always mean-reverts but live-trading requires capital allocation around a volatility stop beyond which I will have no capital left for additional trades [rolls]. “Trading small” marginalizes this practical concern because it implies I always have plenty of capital in reserve but I question whether I could ever run a viable trading business and pay my monthly living expenses this way unless I were a multi-millionaire to begin with.

The Tasty Trade team did allow for only one roll but they did not give a maximum drawdown analysis, which would help to better assess the viability of rolling.

In addition, I certainly believe the OTM and ATM naked put trades should be backtested independently in large sample size to get a clearer understanding of performance.

Musings on Put Credit Spreads (Part 2)

Last time I summarized my progress analyzing naked puts and addressed my biggest concern. One possible antidote to undefined risk is an option spread but backtesting spreads presents an array of challenges.

The simplest spreads are vertical or horizontal. Beyond that, we have:

• Diagonals
• Butterflies
• Iron Condors
• Double Diagonals (i.e. “Straddle-Strangle Swap”)
• Ratio Backspreads

A complex spread is a combination of simple spreads and/or a single option(s). For backtesting purposes, I wonder whether the separate legs must be combined as they were historically or whether this is too deterministic. In case of the latter, I should randomize the combinations and study a simulation of many more possibilities in true Monte Carlo style.

I’m talking a violent disruption to the conventional order, here…

Thankfully, I believe price interaction is a decent argument for keeping historical spreads intact. For example, consider a liquid put vertical spread that is one strike wide. With 30 days left to expiration, I would never see a three-fold increase in short put premium with a concomitant 90% decrease in the long put. For this reason it makes sense to keep them together and this should apply no matter how many legs in the spread.

I still remain leery about complex trade methodologies. These are trades with multiple legs and various [discretionary] adjustment guidelines. Target option deltas can vary as can margin requirements, PnL targets, etc. Adjustments present like a fly in the ointment. If these are actually new trades then why backtest them together as opposed to closing one trade, starting another one, and combining the results?

This is partly why I believe the appeal to complex trade methodologies including multi-legged positions with eye-popping marketing names is misguided. Large degrees of freedom significantly complicate backtesting; curve-fitting or small sample sizes muddle the interpretation. People think they are objectively backtesting while future leaks and free-range confirmation bias inconspicuously oppose them. Dream of the Holy Grail remains alive and demand to purchase the education in hopes of profitable trading is directly proportional to marketing efficiency.

Musings on Put Credit Spreads (Part 1)

The biggest problem I have with naked puts is the inability to defend them in a market crash. Studying the last 15 years has given me great perspective in terms of drawdowns (DD). I must always expect a worse DD in the future, though, and it does seem like we’re seeing historic “worsts” on a semi-regular basis. The latest included two IV-popping 10+% corrections in the span of five months (Sep 2015, Jan 2016), which is a first for the 15-year time interval.

My concern about naked puts is their undefined risk nature. This means the theoretical worst could be a whole lot worse than anything ever seen. For this reason, I would definitely limit naked puts to a percentage of the total portfolio. I also find it promising to cut position size based on some sort of bull/bear regime change.

Further study is needed to determine whether a regime change would actually help. Anecdotally, significant profits seem to accrue when the market is below an X-day moving average (MA). I suspect the sharpest losses also occur below an X-day MA, though, so it may or may not make sense to limit exposure based on regime. Given my druthers, I would sacrifice profits for DD improvement and trade a smoother equity curve in larger size. This may result in a greater overall net profit.

The real kicker is that I believe position size should be held constant in trading system development. Varying position size complicates DD analysis of the equity curve because some will be apples and others will be oranges.

My study of naked puts is not complete as of the time of this writing. My last backtest fixed short delta rather than premium collected and I need to analyze those data. I am waiting to include February expiration, which has losses from the most recent market correction.

Also holding me up are the concerns described above. To that end, trading put verticals is very tempting to take the “undefined” out of undefined risk. I will continue this deliberation next time.