To some extent, this is valid. The average investor can now do things in a matter of minutes that would have required considerable exertion on the part of a prior generation of pros, and if the things the average investor chooses to do make sense, the probability of success can be high. There, however, is where things get delicate. Ease of access and easy-to-use tools lead, naturally, to easy analytic principles many of which were widely disseminated during the late-1990s and early-2000s through a variety of books, seminars and on-line content. Ultimately, though, equity investments succeed or fail based on what happens in the future, and I'd always thought it was very difficult to predict the future.
This brings us to a "good news, bad news" situation. The good news is more often than not, change comes about in an evolutionary rather than revolutionary manner, so a well-executed analysis of the past can often point investors in the right direction. The bad news is that achieving a well-executed analysis of the past may require more effort than many in the early on-line investing generation realized.
Perhaps the Grand Master of K-I-S-S is Joel Greenblatt, who best known for his book The Little Book That Beats The Market, wherein he advocates for what he refers to as a "magic formula" based on just two factors, return on capital and valuation. In other words, he tells readers to focus on good companies whose shares are reasonably valued.
Some who advocated simplicity in the early years of the internet era wound up leading their followers over a cliff. (I don't want to name names and point fingers right now, but if you were surfing the web a decade ago in search of investment ideas, I'm sure you can easily fill in the blanks.)
Fortunately for followers of Greenblatt, the strategy he advocated was sensible and my own testing suggests it really did deliver market-beating performance. (Each of us can decide for ourselves whether and to what extent this has something to do with the fact that Greenblatt was active well before the internet era began.) Figure 1 shows the result a 3/31/01 to 9/23/10 of a backtest of a screen and ranking system I built on Portfolio123 which followed Greenblatt's strategy as closely as feasible and assumed four-week rebalancing.
So that's it. Simple analysis really works. We're done.
Not so fast.
Simple On Top; Complex Under The Hood
Greenblatt defines return on capital as EBIT (earnings before interest and taxes) divided by tangible capital, the latter being defined as net working capital plus net fixed assets. Valuation is expressed in terms of earnings yield, which Greenblatt defines as EBIT divided by Enterprise Value. We can use these definitions on Portfolio123 but it would require some effort on the part of those who are new to this platform.
Enterprise Value requires a custom formula (it, and others required by the Greenblatt strategy, exist as pre-defined formulas).
We don't have tangible capital as defined by Greenblatt. Even though we're considerably more sophisticated than other platforms available for less than $60,000 per year, so we still have to build the item on our own (as I recall from having, in the past, used FactSet, you'd have to custom build it there too). My version is net plant minus working capital which, in Portfolio123 language, would be:
ItemQ(APPN,0)+( CurAstQ- CurLiabQ)
EBIT seems easier to deal with. It exists as a standard stock screening factor on Portfolio123 and elsewhere, but because of the way modern corporations format their income statements and the way these are picked up by the data vendors, the EBIT numbers we have include a lot of unusual items that violate the spirit of the Greenblatt approach (and probably the approaches used by most others who use EBIT). To address this, I use the following not-at-all-simple formula:
This translates to operating profit plus unusual expenses so long as the database doesn't return NA (Not Available) for unusuals; if unusuals are NA, we convert unusuals to zero.
Next, let's focus on Chapter 11 of Greenblatt's book. There, he recognizes difficulties in the book's use of the most recent set of numbers. After all, the stock moves in response to future developments, not based on what happened in the past. He also discusses how estimates can be dicey because even the year ahead may not be representative of the company's true capabilities. Therefore, he says, it's preferable to use numbers that reflect projections three to four years into the future. That's not done in the published version of the Magic Formula, but on page 103, Greenberg acknowledges that he and his partners engage in this admittedly harder kind of analysis when they invest.
I'm also uncomfortable doing an entire analysis based on current numbers. If I'm looking at a single stock, I can project a few years down the road as do Greenblatt and his partners. But in building a model for use by a wider audience, that's not practical. So in for the Portfolio123 version of the Greenberg model, I come at the issue of normal-versus-unusual from a different direction. I added a screening rule eliminating companies that rank below the top 35% in terms of five-year return on investment. Backtesting this version, we see it adds a bit to the more pure approach described above.
Sometimes, it's better; other times, it isn't. On the whole, both versions are OK.
This is all well and good in terms of numbers. But in terms of simplicity, the Magic Formula looks a lot less magical.
The Price Of Simplicity
Greenblatt recognizes that some (Many? Most?) won't really be able to recreate his magic formula. Those who want to do it on their own rather than just coming to his web site can simply use return on assets and P/E. But simplicity comes at a price. Figure 3 shows the performance record of this latter approach.
It still beats the market (confirming the strategy's string common-sense foundation), meaning the little book still does what the title says it does. But simplification but the excess performance is less than what it was.
Switching Gears - Adding Complexity
Given the way simplicity cut into performance, it stands to reason that increasing the complexity of the model ought to have positive results. I decided to test this idea.
I created my own turbocharged version of a Greenblatt ranking system. Consistent with his overarching philosophy, I stayed with just the two equally-weighted criteria he put forth: (1) return on capital, and (2) valuation. But instead of sticking with just his formulas, I added additional variations of each concept.
Figure 4 shows the factors in my original Greenblatt ranking system, the one that attempts to match the book. Figure 5 shows the 3/31/01 to present performance record assuming four-week rebalancing and a NO OTC universe.
Figure 6 shows the factors used in my enhanced version of the Greenblatt approach and Figure 7 shows its performance record.
It's a close call as to which version performed better insofar as lining up the buckets. But on the whole, I think mine is a bit better.
But when we're looking to derive an investable portfolio rather than naively assuming we'll buy all stocks in the bucket, we need to go further and test the rank in the context of a screen or simulation. Figure 8 shows the backtest results for my Greenblatt screen (including the 5-year ROI filter that I added) narrowed to a top 15 using the Gerstein-Greenblatt ranking system.
That is better by far than what was obtained from the book-version of the Greenblatt model (Figure 2).
Why Complexity Worked
Don't take this as encouragement to build mega-models that include every possible factor you can dream up. There was a method to my madness, a definite reason why I expected the more complex version to outperform the simple approaches.
It's important to start by reiterating that I stayed largely true to Greenblatt's strategy: I focused on two basic attributes: company quality and stock valuation. Like Greenblatt, I was looking for a good company whose stock could be purchased at a reasonable price. Notice, too, that in the Gerstein-Greenblatt ranking system, the original Greenblatt formulas were each weighted heavily, 60%.
Now look at the additional rank factors I added. All relate to value and return on capital. What I did was add a lot more ways to define return on capital and more ways to define value.
I wasn't really trying to add complexity just for the sake of complexity. My goal was to add flexibility as to how a stock could show merit under the return-on-capital and valuation criteria. In other words, I'm trying to avoid pre-judging the exact way in which a company demonstrates that it belongs in the portfolio. One stock may be superb under the Greenblatt formula. Another may be so-so under Greenblatt but terrific based on return on equity. Another stock may be decent but unspectacular using trailing 12 month numbers, but top of the heap when we look at five-year averages. And so on and so forth. The same thing happens with value, where I provide alternatives to Enterprise Value divided by EBIT.
My version is a more pure implementation of Greenblatt's ideas then he provides in his own book. The latter claims to advocate for good companies whose stock is available at good prices, but that's not what it really offers. Instead, it advocates for two specific formulas. A company with a very high return on equity (five-years and trailing 12 months) whose stock is cheap based on a forward PEG ratio would seem to qualify as a good company and a reasonable price, but Greenblatt's little book does not care. It looks specifically at EBIT divided by tangible capital as he defines it rather than good companies in general, and it looks at Enterprise Value divided by EBIT, specifically, rather than good stock valuations.
Complexity is not introduced here simply to fill up the kitchen sink, or make an impression. It stems from a deliberate effort to identify situations that would be accepted as worthwhile by as wide a range of good-company/good-stock investors as possible, including but definitely not limited to Joel Greenblatt. I like the way Greenblatt thinks. But the 60% weighting I give to his formulas is quite ample as an expression of this. As Figure 8 shows, my use of the other 40% to open the doors to other definitions was highly productive.
I'd be happy to stick with a single factor or formula for each idea I had if I were confident that it was, indeed, THE perfect answer - and if I were confident that I had reached such a conclusion without data mining. Thus far, though, I have yet to find any such silver bullet.
The Grand Irony - Ease Of Execution
The irony here is the ease with which I built my more complex ranking system. Notice from Figure 6 that everything I added is available as a pre-packaged factor in Portfolio123. All the extra complexity, and the several hundred extra basis points of backtested performance, came as a result of about thirty seconds worth of pointing and clicking. Another irony comes from that the most complex aspect of the Greenblatt-Gerstein ranking system is the part I copied from Greenblatt's Magic Formula.
Don't assume sophisticated ideas are hard to express using the tools that we have. It won't always be as easy as it was here. But once you learn to use the tools that are available to you, you'll find little difference between the effort required to create a simple model versus a complex model.
Fourteen syllable haikus can be every bit as artistically powerful as longer poems. The same holds true of investing strategy. So if you have a nice, clean, simple idea and it works to your satisfaction, by all means, use it!
The problem is that in the cyber-investing era, simplicity is sometimes glamorized to the point where many feel this is the way it should be. If you have a voluminous model that makes sense (we always want to avoid data mining) and works, that's fine. Don't force yourself to edit it down. And if you discover a simple model and feel you can make it better by adding some things, that's fine too.
Either way, avoid getting hung up on pre-determined notions of simplicity versus complexity. Just do whatever works for you.
As has been the case for many models (including the standard Portfolio123 Greenblatt model), the performance of the Gerstein-Greenblatt strategy has not been impressive. But consistent with its longer term tendency to show its strongest performance during good markets, it's been pretty good since the market broke out of its summer doldrums. The model has been set for Community visibility and can be seen here.