The theory behind limiting exceptional stock losses is straightforward. Extreme losses are caused by extreme fundamentals. The challenge lies in real-world implementation. Here’s a 20-stock portfolio designed to limit losses through identification of extreme fundamentals and rules eliminating stocks thusly impacted.

A Change of Pace

The approach taken here differs from the traditional quant-based methods that seek to minimize or at least reduce Beta, Standard Deviation, Downside Deviation, Value at Risk, etc. Despite what’s often said about such metrics, they are not legitimate indicators of risk. They are nothing more than statistical report cards showing the results of what took place at specific times in the past. What counts is not the report card, but the underlying factors that caused the report card to appear as it did. Just as a B+ student can get Cs in the next marking period if he stops studying, so, too, can a stock with a good report card (i.e. a low Beta) turn riskier or even speculative if conditions take a turn for the worse. As with the kids’ report cards, the grade won’t tell you what to expect in the next marking period; you have to look to continuation of or slackening in that which caused the grade; study habits, etc.

There are strands of modern quantitative analysis that look to the asymmetry of investor desires (we like upside volatility but hate big downside moves) and seek stocks with a demonstrated propensity for delivering returns skewed toward the upside (in statistical jargon, the “right tail” or, avoid stocks with track records of big downside moves (“Left tail” returns). As discussed on 9/21/15, that’s the wrong way to go. The characteristics capable of inspiring big gains are equally likely to produce extreme losses if things break badly for the company.

Any investor who aims to maximize upside volatility while minimizing downside risk is engaging in a dangerous fantasy. If you have one, then you have the other no matter how much so-called research you can show that suggests otherwise. For this strategy, I’m going to keep it real. The goal is to reduce the probability of extreme losses. I won’t go out of my way to close my eyes to the upside, and I believe the model can deliver market-beating gains. But don’t look for quick doublings or ten- baggers, at least not by design. We’ll catch those only if we get lucky.

Setting Expectations re: Maximum Drawdown

Maximum Drawdown, or Max DD, is a measure of the worst peak-to-trough movement experienced by a stock or a portfolio (or any asset for that matter). Many consider a modest number here as a sign of good control of downside risk.

I wish it were true. If Max DD really could be thoughtfully analyzed and modeled, I’d be all over it. But alas, it’s not so.

It’s easy to reduce Max DD during the process of strategy development. All we need do is look at the period in which Max DD occurred in the past (for studies with long testing periods, this will turn out to be a brief span in late 2008), identify characteristics associated with the best and/or worst performers during that period, build those into our model, and voila, we can show the world a strategy that reduces Max DD. But there’s a huge problem: This is an example of what’s called data mining, or sham research. To really control Max DD in the future, which, when push comes to shove, is what we really care about, we need to be confident in the persistence of the factors that caused bad Max DD experiences in the past. And it’s hard, if not downright impossible, to even set up a decent, non-data-mined, research effort because Max DD is so sensitive to the peculiarities of one specific point in time.

Bear in mind, too, that many Max DD experiences are based heavily on something we cannot control, market behavior (that was especially so in late 2008). And a lot of the bad 2008 Draw Down stocks turned out to be great for those who didn’t have to liquidate on the spot, given the extreme recoveries (and then some in many cases) they quickly experienced when the market crisis abated. Bear in mind there’s a difference between a stock and a company. This was acknowledged in a Wes Gray – Jack Vogel study in which they envisioned “tail risk” in terms of potential bankruptcy. But I disagree with their effort to analyze the risk in terms of Max DD, which can be subject to all sorts of unpredictable things that can come and go quickly. The really bad losses come from bad decisions regarding the company, situations where a big price drop isn’t a temporary Mr. Market hissy-fit but a reaction to something real, something fundamental, and something potentially wealth killing. Substantive MaxDD research needs to be based on substantive things we can research, analyze and use as decision-factors.

I’ll warn you right now that the model I present will show traditionally computed pot-luck market-influenced Max DD numbers typical of most other strategies (i.e. rotten, as was pretty much everything in 2008 that doesn’t come from after-the-fact simulations). But I will also show different more substantive data points that indicate the strategy really is increasing the probability of avoiding extreme losses,

Building the Strategy

My 9/21/15 post outlined a pretty clear-cut approach to minimizing the probability of big losses; pick stocks with more balanced and less extreme scores on relevant investment factors. But it was structured in an academic-like manner, a best 100, a mid-ranked 100 and a worst 100. When we adapt ideas developed that way for use in actionable, investable portfolios (in this case, one that holds 20 socks and has trading costs), we have to tread a finer line. So, although I’ll continue to avoid fundamental extremes, I will give the strategy a more generally upward flavor in terms of rankings.

In one sense, it would seem as if we could very easily eliminate large losses simply through use of stop-loss orders. It’s imperfect since sometimes, adverse events cause stocks to plunge right past the limits causing us to realize bigger losses than we had targeted. As annoying as that is, however, it’s the least of our problems. The main drawback to stop-losses is the propensity such requirements have to kick us out of stocks we really should have held due to the fact that the losses represented transitory and ultimately meaningless hiccups on the part of Mr. Market. So I don’t usually use stop-loss orders, finding, in most cases, that they actually reduce performance.

I’m going to change pace here, however, and use a stop loss defined as a 15% from the entry price. The reason I do so now is because I believe the rest of the strategy will cut down on the erroneous stopped-out sales.

I’m also going to model against S&P 500 constitutes only. Besides the being the default equity exposure for many (whether we approve or disapprove of the reasons, it’s hard to deny this is so), there are fundamental reasons to expect bigger companies to produce less volatile earnings streams. They are better than smaller firms in covering fixed costs. Also internal diversification (in many ways, when you look at the businesses in detail, it’s often reasonable to see a large-cap company, even those that seem to be single-industry firms, as a portfolio of wholly-owned businesses many of which if they were independent, would look a lot like garden variety small- or micro-cap companies) can be expected to produce less volatile profit streams. We see indications of this in Table 1, which shows S&P 500 and Russell 2000 averages for:

• Fixed-cost burdens (because GAAP accounting does not include entries for fixed and variable costs, as is the case for internal accounting, I’m making the very rough assumptions that interest and SG&A are fixed and that cost of sales is variable. Therefore, the averages shown in Table 1 represent SG&A plus interest expense as percentages of gross profit. The numbers are five-year averages.)
• Sales Variability (the ratio of five-year standard deviation of annual sales to the five-year average)
• Operating Pretax Profit Variability (the ratio of five-year standard deviation of pretax profit adjusted to exclude special items to the five-year average)

Table 1

Within the more fundamentally stable S&P 500 group, I screen for stocks ranked above 50 and below 90 (with zero being worst and 100 being best) according to a Quality factor developed on Portfolio123 by UCLA Anderson School of Management MFE Candidate Yong Kim based on the following:

• Return on Assets
• Change in Asset Turnover
• Accruals
• Debt Leverage
• Asset Growth
• EPS Growth

The passing stocks were then ranked under a Value ranking system based on Enterprise Value to Sales and Price to Free Cash Flow. The 20 best-ranked stocks (the 20 with the most attractive valuation metrics) were selected for the portfolio.

The model is refreshed weekly and stocks are sold if the Quality rating rises above 90, if the Quality rating falls below 50, or if the Value rating falls below 90. (And, as noted, there is also the 15% entry-based stop loss.) I don’t often create models for weekly trading. But considering what we want to do here, limit large losses, I think it’s reasonable to refresh the model and look for sell signals more frequently than I usually do. Note, though, that given the model’s emphasis on stability, many of these weekly updates produce no sell signals. Turnover is actually in line with many of my other models.

Table 2 shows the current list of passing stocks.

Table 2

If you register (for free) on Portfolio123, you can follow the model on the Ready-to-Go platform, where it available at no cost to the user.

In my next post, I’ll evaluate the strategy’s performance, focusing on its potential for limiting the risk of large losses.