Fellow bloggers are debating the F_Score, a financial statistic I use to help screen the market for potential investments and as a factor in the mechanical Nifty Thrifty portfolio. So far, little of the commentary has been positive.

If you like a knock-about discussion you can see it starting on McTurra’s blog and culminating in this post on the red corner’s. Although I’m an F_Score user, I agree with much of what they say.

The F_Score cannot be relied upon to tell you whether a particular company will do well, or badly in the stockmarket. Piotroski, its inventor, found that when applied to undervalued companies, the F_Score successfully predicted market beating performance for about 50% of them.

Although that’s no better than chance, it’s better than the performance of smaller undervalued companies in general. They beat the market less than 44% of the time.

The result seems counterintuitive. After all, everybody knows large buckets of undervalued shares beat the market, if you give them long enough. So how can most of them be losers? Piotroski found the gains from a minority of winners more than make up for the losers.

By weeding out some of the losers using the F_Score, he improved the performance of large portfolios of undervalued stocks.

But the F_Score does look horrible to a fundamental analyst, you can see the disgust dripping off the letters in Red’s blog post. Piotroski didn’t cobble it together from the best ratios to determine profitability and indebtedness (we’d all disagree on those anyway), he used the easiest ones to calculate. Then he weighted them equally.

The result is a generalisation. It omits important factors, like leases and pensions, and some of the factors it includes are more predictive and reliable than others. In some situations, the factors might give a false signal.

For example, if the number of shares in issue rises during the year, the company’s F_Score goes down by one (higher F_Scores are good, the maximum is nine). In some circumstances that’s the right result: a rising share count is a sign of financial weakness when a company is forced to sell new shares at a deep discount in a rights issue. In other circumstances, for example if a company has issued shares to reward staff, or fund new investment, it may not be. Likewise increasing leverage in a distressed company is a sign of weakness, but it could be be a harbinger of improved performance, if the money is invested wisely by a company with good prospects.

Piotroski recognised the F_Score’s a blunt instrument, best used to identify financially distressed companies whose circumstances are improving. He calculated the predictive power of each of the variables, the two most powerful simply indicate the company is profitable in cash and accounting terms. The number of shares in issue (EQ_OFFER), for example, looks particularly ineffective:

[For an explanation of each factor see this post on the F_Score, or the original paper.]

Mechanical investors must use the available statistics and skip around their limitations by buying large buckets of stocks and relying on, as Piotroski described, the successes outweighing the failures. That’s the way the Nifty Thrifty, which is a portfolio of 30 stocks, works.

Stockpickers should be careful. Indiscriminate use of the F_Score will lead them astray. Here are some rough guidelines I keep in mind when selecting companies to research:

1. Low F_Scores are often significant, Piotroski found that low F_Score firms are five times more likely to delist for performance related reasons than high F_Score firms. Low F_Score firms are unlikely to make good investments over the next few years.
2. High F_Scores are significant for companies that are cyclical or recovering from self inflicted problems. They may be early stage turnarounds and  make good investments. 
3. Middling F_Scores should not put investors off stalwarts, companies that have long and consistent records of profitability, or fast growing companies. Hiccups may well be temporary, and Piotroski’s factors are most likely to give false indications for these companies.
4. The F_Score should not be applied to speculative companies, companies that have yet to make any money like young mineral exploration companies.

If you scroll the Annual Results Report to the right, you’ll see three additional grey columns. The first contains a category, turnaround, cyclical, stalwart, growth or speculative, assigned to each company. These categories help me interpret the F_Score, and, along with reading the financial statements, enable me to decide whether or not to research a particular company any further.

In subsequent research you won't find me mentioning the F_Score. By then I've gone from the general to the specific analysis of a particular company in a particular situation using the best fundamental ratios and statistics I can muster.