MFI Diary abandons the formula

MFI diary, a blog that tracks a portfolio of stocks using Greenblatt's Magic Formula, has decided to abandon the approach. His rationale is explained in this post:

My MFI Index... is down 5.5% since I started it in February of 2006. The Russell 3000 is down ... 7% since February 2006 (recall, my measurements are weighted by when I put money in, so it is not a point movement from February 2006).

Some of my own thoughts on MFI:

1. Nobody really knows what the Magic Formula actually is! Everyone has their own interpretation of it, but Greenblatt hasn't spelled out the exact details in a blow-by-blow analysis. In my opinion, he would do well to chop out a lot of the book, and replace it with example balance sheets and P&L statements to show how the numbers are calculated. BTW, he makes the same omission in his latest book, The Big Secret for the Small Investor. He never actually tells you how to construct a value-weighted portfolio.
2. The ROCs that are generated are generally too high to be credible. My understanding is that Greenblatt is attempting to determine an "incremental return on capital"; he wants to know what each new dollar will earn. I think this is very difficult to use accounts to determine that. ROC, and most return on capital calculations, exclude intangible assets in the denominator. I am very sceptical of the exclusion for the following reason: the company did actually expend money on intangibles (probably goodwill) in order to generate its current return, so calculations of future returns must take that into account. After all, if a company didn't "need" to pay goodwill, then why did it? If you argue that a company overpaid for an acquisition, then surely you must conclude that the company is a poor allocator of capital, and any calculation you produce for a return on capital will be overstated. Contrariwise, if you argue that the company consistently underpays for acquisitions (a rather less likely scenario), then this is in itself a game-changer as to how you assess the company. Finally, almost no companies are able to generate returns suggested by most return on capital calculations. Sharelock Holmes reports ROCE for BATS (British American Tobacco) of 74%. Digital Look reports a figure of 65%. Whilst undoubtedly BATS enjoys excellent returns on capital, it seems unlikely that it can generate returns that high. Maybe a company like Microsoft can generate eye-popping returns, but truly excellent returns on incremental capital are available to only a very few. To borrow from Bruce Greenwald, you can't get good information from bad data.
3. It seems implausible that it is possible to generate returns of over 30% pa (page 56) using any mechanical formula - beating market averages by a monster 18.5% pa. If only. What's more curious, is that in Greenblatt's latest book (the "Secret"), on page 122 he posts returns from a value-weighted index of 13.9% pa, compared with 7.6% pa for the S&P500. In that case, the outperformance is 6.3% - a far more plausible figure, and one in which I actually think is possible as a mechanical strategy over the long term. The puzzling thing is, though, is that Greenblatt on page 121 of Secret, he refers to constructing a value-weighted based on good and cheap, and specifically cites his "Magic" (The Little Book That Beats the Market) book. So what's going on here? Is he saying that the magic formula doesn't really return 30% pa, but that it's more like 13.9%?