More evidence Greenblatt's formula works in the UK, and a suggestion to improve it

The briefest of recaps: The Magic Formula is a simple quantitative strategy developed by US hedge fund manager and value investor Joel Greenblatt that, according to his tests and other researchers, beats the market handsomely. You can read all my notes on the Magic Formula here, preferably starting with the last post. Using and testing the strategy in the UK is tricky though, because we don't have a proprietary database like Greenblatt's and testing by individual investors, mine particularly, probably lacks the rigour of the best in academia or the City. That hasn't stopped us trying though…

Reader Tom Norman has back-tested the of Magic Formula since 2001 in the UK that corroborates my results. Like me, he used Return on Assets (ROA) instead of Return on Capital (ROC), see this post for why, and he used the same data source (Sharelockholmes.com), so I'm relieved, but not altogether surprised, he got a similar result.

Rather wonderfully, he also tested the combination of the Magic Formula and the F_Score, which I use to run the Nifty Thrifty portfolio for Money Observer magazine, an idea we both got from a research note written by Morgan Stanley's Edmund Ng. The F_Score, which discriminates between companies getting into trouble and companies getting out of it, should weed out some potentially big losers and reduce the risk inherent in the Magic Formula.

The Nifty Thrifty has not performed well so far, so these results, clipped from Tom's email, come at a good time:

Average return greater than FTSE 100 for MF after 1 year (using ROTA rather than Greenblatt’s method): 9.62%

Average return greater than FTSE 100 for MF after 1 year (as above) but F_score >= 7 : 8.27%

BUT

Standard deviation of MF only: 13.18
Standard deviation of MF/F_Score: 8.74

So filtering for F_Score should considerably reduce the volatility of one’s portfolio and the 8.27% surplus return is much more statistically reliable than the 9.62% from MF only portfolios.

Also of interest was what happens if we look at surplus return / standard deviation further than one year out:

MF only holding for:

2yr 15.68% / 17.29
3yr 16.94% / 19.93
4yr 19.90% / 23.73

MF/F_Score

2yr 16.77% / 10.73
3yr 21.05% / 15.07
4yr 24.27% / 16.93

My conclusions are:

1. Contrary to popular opinion the F_Score provides useful information about large companies – particularly their ability to provide superior risk adjusted returns over an extended period of time.
2. With even small transaction costs it’s worth holding on to shares for two or even three years rather than the one year advocated by the MF.

Sharp-eyed readers will notice Tom's average returns for the Magic Formula are a few percent below mine, but we think that's because he started his tests a year later, missing an exceptional year for the Magic Formula.