One thing about being in Switzerland for me is that the weekends can be a bit of a drag, especially when it is pea soup foggy. Maybe not coincidentally, some of my best thoughts about magic formula investing have happened while I am in Switzerland. I just start noodling with my database.
At first I asked myself, would you rather buy an MFI stock that is near it's 52 week high? Near it's 52 week low? Or should we be agnostic. I started down the road. But frankly it was more effort than I wanted to put forth.
The Problem With MFI
Magic Formula Investing does have a couple of "problems". Greenblatt even alludes to them as beneficial as they will deter people that have shorter time horizons.
- MFI is extremely streaky. It will work well for portfolios over 12 to 18 months and then it will not work so well for the next two years. I will embed a graph illustrating that fact. I'd suggest that it has been more streaky to the downside that Greenblatt found in his back testing. He used 17 years, I have ten so far. I do think he had an unusual stretch (there was an article about this in Barron's today) in that during the internet bubble of 1997-00, value stocks were largely ignored. So there were some very, very good deals to be had in the stretch which a MFI would be very adept at recognizing. (you could argue we have been going through a similar streak with Netflix, FB, Tesla, Amazon etc - so perhaps value is about to make a second comeback).
- MFI stocks even within a portfolio seem to me to be very volatile. Some stocks do very well, but there are some stocks that end up being absolute stinkers. Wouldn't it be great to be able to find something that helps predict which bucket you're in? I have bucketed the stocks, let us see if beer-infused rambling sees anything noteworthy.
Sorry, I cannot talking streaking without a shout out to Frank the Tank. Sure, a lot of Will Farrell movies are flat out stupid, but every now and then he just makes me laugh out loud.
I have put together a very simple chart that graphs the difference returns of every one of my monthly MFI tracking portfolios and the return of the benchmark Russell 3000 for the same year. What you will notice is that it isn not random. It is not like some months MFI does better and some months R3K does better. Rather R3K does better for twenty straight months and thenMFI does better for 12 straight.
Overall I have 132 tracking portfolio of 50 stocks each. They are started in one month intervals and run for a year. Of the 132 (12 of which are open), MFI has won 56, which is 42%. JG in his book suggested MFI had won 5/8 in his back testing. Obviously if you just win 42% of the time, it is tough to win in aggregate. Here is the promised chart:
So note how it does just randomly go above and below the X-Axis. Rather you get many consecutive portfolios under and then a spurt above. Obviously, if you knew where you were in this time series, you could increase or decrease your investment. So we have clearly spurted back above (the drop to 0 on last tranche is based on a tranche less than a month old). You can also see that when MFI is good, it can be really good. The 42 months it won, it won by average of 12.3% (sweet). The months it lost, it lost by 6.7%. That is kind of an interesting dynamic. It does show the power of MFI. When it wins, it can REALLY win.
So I suppose you could put in a rule that when two straight of my tranches close a winner, you are in. When two straight close as losers, you're out. What would that look like (recall there is a lag, when one ends up a loser, I have 11 in motion)?
I won't bore you with the work up. But basically the lag is too great. This year is a great example. We've lost every portfolio that has closed this year until 11/27, which won by a hair. So the two straight rule will not be triggered until year end. You have missed the next ten portfolios that are looking like winners (by a mere average of 6 points right now) - but winning streaks only seem to last 12 to 14 months, so you'd possibly be starting again just as the tide might be turning.
Seems like there could be something there, but I don't know what.
Individual Stock Vol
Over all the years, the average MFI stock goes up 9% in a year. But the standard deviation is a whopping 49%! That is insane. A very quick and dirty look at 50 stocks from S&P 500 shows a standard deviation of 22% (most of the hits have been pharmas, like AXLN or AGN). Looking at the range of 50 S&P stocks from 12/31/2015 - the best is up 76% and the worst is down 38%. For the 50 MFI stocks from 12/31/15, the best is up 126% and the worst is down 69%.
I could do more - but I think you get my drift. Either by standard deviation or range from worst to best, there is significantly more volatility in MFI.
Here is a chart showing number of stocks by return band (recall overall average around 8.2%).
So being able to just cut off those left-most 2 would take you from 8.2% to 19.3%. Whoa. Now that is serious juice. Figure out the 14% of stink bombs (7 of the 50) and we can all retire at Millionaire Acres (love the Game of Life).
Of course, to do this with certainty is impossible. But let us see what stinkers look like and noodle about what they might have in common a priori (little statistics there). That will be part two... tomorrow.