It seems obvious that it would be worthwhile to impose some sort of rule or rules relating to trading volume, value traded (number of shares traded multiplied by price per share) or market capitalization in order to enhance the likelihood that stocks identified in a particular model should be tradable in the real world, on reasonable terms. Actually, though, this can be quite a delicate matter.
We Need To Reach Far Afield
These all stars did not invest the way many modern institutional investors do. In other words, they didn't use the S&P 500 roster and weightings as a starting point and then try to outperform by overweighting or underweighting (or shorting) individual sectors or positions based on whatever fundamental, technical or other quantitative criteria they think will help them generate superior performance. It's not easy, even for geniuses, to do such a thing, to get a consistent edge when investing in a collection of behemoths many of which are among the most widely owned and closely watched companies in the world. And to the extent the all-stars articulate fundamental rules, these are usually designed to uncover extraordinary opportunities, as distinct from the more general ordinariness that tends to characterize the largest most visible firms.
If you spend enough time examining works by and about legendary investors and trying to discover commonalities that help distinguish them from the average money manager, one important thing you'll notice is a willingness to stray far from the beaten track when seeking ideas.
Some, like Peter Lynch, present elaborate arguments in support of a quest for obscurity which often means smaller companies that don't get much if any attention. Ben Graham thinks in print about a minimum-company-size requirement but expressly rejects it in favor of a belief that any firm that can meet his fundamental standards should be worth considering. Others simply speak with their actions.
And yes, yes, yes, I know all about Warren Buffet and Burlington Northern and other household names like Coca Cola, American Express, etc. But those who read all the details of the Berkshire Hathaway filings and pay attention to everything he actually does, as opposed to just what the Buffett-watchers talk about, know full well that much of his activity is devoted to areas that wouldn't have a prayer of getting mentioned in the Wall Street Journal or on CNBC absent Buffett's stepping in.
As suggested, realistically, it's going to be hard to find obscure stocks unless one goes small, often very small, given the way information is so widespread. This goes to why I'm so willing to use stock screens, even though the all-stars themselves probably didn't do it. Screening is great for bringing to your attention interesting ideas you would not otherwise have known were worth a look.
A Careful Approach To Liquidity Filtering
The need for a far-ranging search is what makes it so tricky to impose liquidity constraints. Some of the most common such rules, usually a requirement that market capitalization exceed a certain level, operate to automatically defeat an important aspect of what the all-star investors might find if they were using screens today.
That said, it's hard to simply sit back and let the liquidity chips fall where they may. No idea, no matter how fundamentally intriguing the company may be, is worthwhile if we can't act on it. So we have to do something.
Obviously, I could create rules relating to market capitalization, trading volume and/or value traded. That's what I did in the past. Untimely, though, it's hard to make a strong case for one set of numeric thresholds over another. (Should minimum market cap be $500 million,$300 million or $100 million? The answer can have a dramatic impact on the performance of a model.)
Considering how important it is to be open to small obscure situations, if we are to exclude something at the outset, we really should have an objective basis for doing so, rather than a seat-of-the-pants standard. In other words, if we're going to sacrifice great-on-paper opportunities because of tradability, it's very important that we get the liquidity rule right. Overkill is not a virtue.
So rather than try to impose my own hunches as to what it takes to make a stock tradable, I decided to, in effect, defer to the traders and compliance officers of the on-line brokerage firm Folio Investing (often referred to under its original name, FolioFN). If a stock qualifies for Tier One trading as part of their "window trading" system, I'll accept it. There is an implementation challenge here, but before discussing that, I'll explain for the benefit of those who are unfamiliar with Folio Investing what this is all about.
FolioInvesting.com As A Liquidity Threshold
For a flat subscription fee, Folio Investing customers can engage in unlimited trading of "folios;" baskets of stocks that, actually, can look a lot like the results we get from our screens and simulations (that is what I do at Folio Investing; I buy in bulk all the stocks in my result list). The cost-advantaged trading applies if you are willing to have your orders executed through one of two of the firm's trading windows, one at 11 AM (Eastern time) and the other at 2 PM.
It is not my intent here to tell you to trade through Folio Investing. What I am trying to drive home is the significance, from the vantage point of liquidity, of a stock's being included on their Tier One window-eligible list. If it is so included, it stands as a statement that Folio Investing is confident in its ability to execute orders for the stock on any given trading day. It's like the old "Good Housekeeping Seal of Approval." If you are a Folio customer, you know you can execute trades there. If you are not a Folio customer, there is a very high probability that liquidity in these stocks will be sufficient for you to similarly execute at the firm of your choice.
This does not mean the spread on a $0.50 stock will be as narrow percentage-wise as the spread on Microsoft or IBM. And even for a particular $0.50 stock, spreads on some days will be wider than on others. But based upon my own experience with the firm, it does seem to be the case that spreads even on $0.50 stocks can be considered consistent with the reasonable expectations of one who invests in that segment of the market.
I assume, here, I'm addressing an audience of individual investors and investment advisors; Folio serves both groups. Larger institutional investors would have difficulty with this liquidity threshold and any readers who trade for such accounts would have to experiment with other rules that produce lists tradable under their circumstances. But even this would likely be do-able assuming an element of reasonableness. Realistically, institutions are going to be unable to participate in certain segments of the market, this, in fact, being an important reason why some all-stars like such segments.
Accordingly, the best liquidity rule I can come up with for our use of the all-star models would be based on the Folio investing Tier One window list as captured weekly by Portfolio123. Specifically, I invest real money using screens containing the following liquidity rule: Universe(FolioFN).
Here's a catch - sort of. Use of Universe(Foliofn) makes for an unreliable back-test because the names of the list are static as of the day of our last update. Companies that, in the past, were on the list but later vanished will not be counted in the backtests. (This situation is known as survivorship bias).
A few years ago, that would have been a big problem. But fortunately for us now, Folio Investing has been expanding its list over the years and is now very close to 100 percent inclusion of all non-pink-sheet stocks.
That means Universe(NOOTC), which will produce valid backtests, is almost identical to Universe(FolioFN). As a result, I have set all the community versions of the all-star screens to use Universe(NOOTC) in lieu of the Folio universe. This is done so users can properly backtest the posted models. I suggest that users, after experimenting with, testing and if they wish, editing any of these models chooses to use them in the real world, a second version be saved with Universe(foliofn) in place of Universe(nootc). In other words, test using Universe(nootc) but trade using Universe(foliofn).
Interestingly, this gives us better backtests than we'd get if we were to plug in all the old historical FolioFN lists. Going forward, we're going to be trading under conditions that will, for all practical purposes, make the Universe(NOOTC) stocks fully tradable so we should backtest on that basis. We would not be helping ourselves if we were to allow the backtests to be impacted by the fact that in the past, FolioFN traders may have been less proficient trading the smaller end of the market than is the case now.
This is not a 100 percent perfect solution. But backtesting can never be perfect since we can never be sure the future will match the past in all respects. The past-versus-future discrepancy being introduced here is actually quite modest compared to many others that must be considered when interpreting test results.
Going forward, we will continue to monitor the Folio universe. If (more likely, when) there is a complete match between the Folio list and the Universe(nootc), we would expect to eliminate the FolioFN universe and stick with Universe(nootc) for backtesting and trading.
MISCELLANEOUS FINANCIAL SERVICES
In all cases, added a rule omitting companies classified in the Miscellaneous Financial Services industry. In the Thomson Reuters database, which we use, most of the companies there are closed-end mutual funds, and most of the rest are investment companies that are structured similarly to a fund.
These firms use the corporate entity and, hence, have financial statements that look like those we see for most other companies. Realistically, though, we do not consider these securities on the basis of their financial fundamentals. This of any mutual fund or ETF you own, or even a brokerage money market fund where you park cash. They, too, have revenues, margins, earnings per share, debt, return on capital and so forth. Do you care about any of that?
Those interested in these outfits would be better served creating models designed specifically to select from among them. Limit consideration to this "industry" (industry=fsmisc) and then create the same sort of ranking systems and/or screen you might use for ETFs. That would produce far more reasonable outcomes than would be the case if you were to co-mingle this industry with others, especially in the context of all-star screens.
THE LYNCH MODEL
Of all the all-star models, the one based on the ideas of Peter Lynch has been by far the most challenging. Relative to the other all-star models, this one has posted lackluster five-year performance, and the comparative weakness became more pronounced when I added the new liquidity filter. I also noticed similar performance issues in Lynch models created by others.
To be fair to the Lynch model, it did look quite strong during certain time periods so it may be a simple matter of no model working all the time and the five-year test period may have been one that captured a bigger portion of this model's cold periods. There is definitely an element of truth to that. But considering the sort of outstanding year in and year out performance that has become so firmly associated with the Lynch brand, I felt compelled to dig harder before settling on "sometimes you're hot and sometimes you're not."
Upon further review, I did make some changes. But the revisions are more in the nature of fine tuning than a dramatic overhaul. The more I studied One Up On Wall Street, the more apparent it became that a properly-constructed Lynch-based model will often find itself performance challenged relative to some other all stars.
A Quest For Obscurity
We have to remember that it's been more than 20 years since Peter Lynch managed money in the public spotlight, so we have no idea how he'd have done in recent years (we did see some big-time star investors falter badly). His reputation stems from the work he did mainly in the 1970s and early 1980s, a time when the market was much less institutional than it is today. That means the investor who actively seeks unknown and/or unpopular stocks and unpopular industries now, consistent with what Lynch did, would face a much steeper uphill climb than was the case when he was active.
Make no mistake about this: the quest for stocks tinged by widespread disdain, or at least apathy, is the core of the strategy presented by Lynch in his flagship book. The popular image of Lynch (invest in businesses you discover through the course of your day to day life) was actually a minor element and one that had a more opportunistic flavor, as opposed to something meant to be at the core of an ongoing investment program.
Lynch And Fundamentals
There's another popular image of Lynch as a value investor. This is factually accurate, but the value principles articulated in One Up On Wall Street do not seem distinctive relative to those expressed by many others, including other all stars, who address the topic.
Ditto most of Lynch's other comments relating to fundamental issues except for one, a discomfort with very high and probably unsustainable growth rates. Lynch worries that these will be associated with overheated stock prices and the high risk that comes along with the inevitable deceleration.
The Lynch attitude toward hyper-growth is definitely poignant. But in this era, an investor who applies that approach will have to be prepared to be very patient. Wall Street likes hyper. Those who thought the financial crisis may have cooled investor appetite for aggression would, obviously, need to re-think that position in light of the vigor of the 2009 snap-back.
Lynch Model Revisions
So it must be recognized that the lackluster five-year performance of the Lynch model is, indeed, a case of sometimes you're hot and sometimes you're not. Actually, the not-hot period covered the first four years of that interval; in 2009, the strategy was quite strong. It was also pretty good in 2001-04. If we do experience a bona fide and sustainable economic recovery, there is reason to believe the Lynch approach may be among the stronger ones going forward.
That said and as mentioned above, some fine tuning did still seem warranted. So in addition to a rule eliminating OTC stocks, I made the following adjustments:
- The value rule has been modified from a simple requirement that the trailing 12 month P/E be less than the industry average to a more stringent threshold (that the stock's metric rank in the lowest 25%) that may be satisfied in any one of four ways: the TTM PE evaluated relative to industry peers, the TTM PE evaluated relative to all stocks, the TTM Price/Sales ratio evaluated relative to industry peers, or the TTM Price/Sales ratio evaluated relative to all stocks. PEG was Lynch's preferred metric but it's hard to get a good number for many smaller companies (that lack analyst growth forecasts); Lynch presumably used his own growth forecasts for his PEG computations. PEG is conspicuous in the Lynch ranking system for companies that have the data. But I'd overly cripple the model if I tried to force PEG in the context of a rigid pass-or-fail screening rule.
- The institutional ownership rule has been modified from a simple requirement that it be below the industry average to a more flexible filter requiring the percentage to rank in the bottom half relative to either the industry or relative to all stocks.
- The EPS growth rule, originally confining the annualized five-year rate to a range of 15 percent through 50 percent, has been changed to eliminate the 50 percent ceiling. This is a departure from pure-Peter Lynch. But as explained in the discussion of how we approach all-star models in general, we are not attempting to create laundry-lists including each and every thing a particular all-star said but, rather, our own stand-on-their-own strategies inspired by the ideas of the all-stars. This approach is important since for the most part, the all-stars were not using screens and cannot be assumed to have rigidly applied each and every concept about which they spoke to each and every stock they examined. While one certainly could create a Lynch model using a growth ceiling, we believe the model we created, emphasizing his core preference for disdain-or-apathy already flies enough in the face of popular contemporary Wall Street norms and that the addition of the growth ceiling here was piling too much onto a single screen-based model were, unlike what Mr. Lynch could do on his own, we have to apply every rule and every rank factor to every stock.
THE O'NEIL (CANSLIM) MODEL
This was another challenging model to create.
One factor is that CANSLIM demands so many things (some of which are, necessarily, qualitative). As we observed for Lynch, those who do not use screens have the flexibility to refrain from requiring each stock to precisely comply with each rule. With screening, the more we demand, the more likely we are to drive our result set to zero. Generally, though, the O'Neil model as originally offered seemed to reasonably cope with those challenges.
S Versus I
More perplexing were aspects of CANSLIM that arguably contradict one another, specifically, S (limited Supply of shares) and I (Institutional sponsorship). When focused on S, we're pointed to situations where institutional presence is as close as feasible to zero. When focused on I, we want strong institutional interest. First, O'Neil did not create CANSLIM specifically for stock screeners so there was some flexibility for him and his initial readers to lean one way or the other from stock to stock on a case by case basis. Moreover, when discussing this topic, O'Neil makes strong references to a framework that is not compatible with screening; noting not just the number of institutional owners but the quality, who they are and how one evaluates their proficiencies.
The initial version of our O'Neil model already recognized the need to step back from the details and apply the general flavor of C and I, rather than attempt to replicate a rigid laundry list. The revision fine times that effort.
O'Neil Model Revisions
In addition to a rule barring OTC stocks, I made the following adjustments to the O'Neil strategy:
- The institutional ownership rule was originally established as company ownership percent below industry average and one of two additional tests: percentage ownership greater than five, or at least 20 institutions owning shares. The rule is now simpler: In terms of percent institutional ownership, the company's percentile (relative to all stocks) must rank at least 10 but below 50.
- The original model required percentile ranks of at least 70 in terms of most-recent quarter and trailing 12 month EPS growth and six-month share price performance. Given a tightening of the liquidity filter through the new FolioFN rule, I decided to ease these percentile requirements a bit to allow for companies to be equal to or greater than 65.
THE GREENBLATT MODEL
The no-OTC stocks liquidity filter was added.
Greenblatt's book was ambivalent about how small a market capitalization should be accepted for his model. On page 134, he observes that "[f]or most individuals companies with market capitalizations above $50 million or $100 million should be of sufficient size." The minimum-market-cap rule in our model was established at $50 million, based on that text and on that being the minimum used on Greenblatt's web site, magicformulainvesting.com.
THE SHARPE MODEL
Previously, this model had a set of alternative liquidity filters involving price, volume and market capitalization. That was removed in favor of the simple OTC ban.
Below is the current collection of Portfolio123 collection of All-Star-strategists. Remember, if you want to access the models, join the Portfolio123 All-Star Models Group. From that area, you can copy all the screens, ranking systems, and custom variables into your account. (Don't forget to save the custom formulas in your account! Several of the models will not work unless you have these.)