72 per cent return could be even better - Nifty Thrifty five-year review

Nifty Thrifty, FTSE 350 stocks

It's a day of reckoning for the Nifty Thrifty portfolio. Conceived five years ago as a real-time test of a well-founded and thoroughly modern mechanical investing strategy, it was supposed to beat the market handsomely over any five-year period.

The strategy is based on the work of Joel Greenblatt, a US hedge fund manager, who invented the Magic Formula.

This, the subject of a bestselling book, marries two investment ratios - the earnings yield, a measure of value; and return on capital, a measure of profitability. Purportedly, the Magic Formula selects good companies at cheap prices.

To view the Nifty Thrifty's holdings and trading chronology, click here.


Analysts at investment bank Morgan Stanley thought they could improve on the formula by adding a safety factor, Piotroski's F_Score. This is a credit score bringing together nine statistics that describe a company's financial strength and whether it has improved or deteriorated over the preceding year.

We dubbed our version of the formula the Nifty Thrifty, and aimed it at the UK's largest firms, those in the FTSE 350 index.

Despite the fees and charges of an active trading strategy, the Nifty Thrifty has beaten the market, as represented by an investment in a FTSE All-Share index-tracking exchange traded fund, but not by as much as expected.

Possible reasons abound for the gap between expectation and reality - bad luck, unsuitable data, human meddling and unrealistic expectations among them.

In fact, the algorithm turned an imaginary £30,000 drip-fed into the portfolio in its first year into £51,674. The same imaginary stake drip-fed into the tracker is now worth £46,457.

The Nifty Thrifty is worth 72 per cent more, compared to 55 per cent for the tracker, a difference of 17 per cent over the original £30,000. An investor who'd backed the algorithm would have ended up with 11 per cent more money than one who'd plumped for the tracker.

Although the algorithm does not target dividends, they contributed £5,726, or 26 per cent of the portfolio's total return.

The most obvious costs, in lieu of broker fees and stamp duty, are deducted from its valuation and returns, revealing the cost of actively trading up to 30 shares a year.

The Nifty Thrifty paid £2,150 to its imaginary broker, £10 for every trade, and £774 to its imaginary tax collector in lieu of 0.5 per cent stamp duty on every new addition. The much less demanding index-tracking strategy resulted in fees of £100, and no stamp duty (stamp duty is not levied on ETF transactions).


The Nifty Thrifty's annualised return was 11.5 per cent, compared to 9 per cent for the tracker fund.

Tests on historical data performed by Morgan Stanley in European stock markets had promised a much bigger differential of about 7.5 per cent, though, and the 2.5 per cent margin of success would be thinner still if we had factored in three costs that are not tracked in the portfolio (or generally in model portfolios or backtests) because of the difficulty of doing so consistently.

The first is the spread, the difference between the bid and the offer price at the time of a trade, which is not recorded in the data. This should be very low because of the portfolio's focus on the largest and generally most easily traded shares.

The second is the cost of the data used to power the algorithm, which depends on the source. We use Stockopedia, which costs upwards of £14 a month, depending on the subscription option.

The third is time. The algorithm may come up with recommendations automatically, but the process is not completely manual; the shares must be traded and the trades recorded. It takes about three hours every quarter, the cost of which would depend on how an individual values his or her time.

Accounting for those costs, the Nifty Thrifty would still be ahead. But it doesn't feel like a handsome victory. More like a grinding one, after years of trench warfare. During its first two years the Nifty Thrifty took a battering in the trenches.

Performance reached its nadir in summer 2012, when the portfolio had not only lost money over the previous year but had actually lost more than the benchmark index tracker.


No system is perfect. A year or two of poor performance is to be expected from time to time, but concerns about the suitability of the data for mechanical trading led me to switch to Stockopedia, a supplier that not only calculates the ratios more faithfully but updates a company's statistics using quarterly and half-yearly results as well as at the end of the year. It thereby ensures the ratios are as up to date as possible.

The switch also enabled me to undo human interventions intended to mitigate my data concerns. I had tweaked the algorithm so it required an even higher F_Score, and skewed it so the Nifty Thrifty added more shares in quarters when more companies were reporting full-year results.

The idea was to increase the safety factor by dialing up the financial strength of the portfolio and prioritising recent information but, since I had no statistical evidence these instinctive changes would improve performance, I reverted to them once the algorithm was using more robust data.

Since then, the Nifty Thrifty has been on the offensive. It's made up the lost ground and overhauled the benchmark; and now, on the fifth anniversary, I've undone another human intervention - the limit on the number of companies in the portfolio from any single sector (four).

There are two reasons for this volte face. Like my other tweaks, there is no statistical evidence that it improves results: I introduced it in the hope that enforcing diversification would strengthen the portfolio when I suspected the data wasn't up to the job.

To put the newer data properly to the test, I must have faith and stop tweaking. This particular tweak also complicated the selection process horribly, because it had to be done manually, so I'm glad to see the back of it.


My vow for the next five years is not to meddle with the algorithm. It's supposed to do the work, and it will probably do it better if it's left alone.

I'm not going to change the ambition either, to beat the market handsomely, but from next quarter the Nifty Thrifty will have a new, potentially tougher benchmark: accumulation units in the Royal London FTSE 350 Index Tracker fund. This is a better benchmark because it's an even lower cost and lower effort investment.

Dividends from the ETF must be reinvested every year, while dividends in the Royal London fund accumulate within it, saving dealing fees and time. It also selects shares from exactly the same pool as the Nifty Thrifty, so it's a better comparator.

In practice the performance of the two benchmarks has been similar over the last five years, with the Royal London doing marginally better. A lump sum of £30,000 invested in the Royal London fund five years ago would be worth £47,853 today - £1,396 more than in the FTSE All-Share ETF.

The algorithm has performed well in the last quarter, widening its lead over its old benchmark (and its new one), and thanks to the strong finish, it's beaten the market. It's increased 8.5 per cent in value, compared to 6 per cent for the benchmark. Nevertheless, an uninspiring year has followed two good ones.

Every quarter, the computer ranks the companies in the FTSE 350 index - the London-listed companies with the highest market capitalisations - that also have an F_Score of 5 out of 9 or more.


Companies with the highest combined return on capital and earnings yield, the proxies chosen by Greenblatt for quality and value in his Magic Formula, are ranked at the top of the list of candidate shares. In theory, these are the best of the best: profitable but cheap companies selected from a list of the biggest and strongest listed in London.

Roughly one quarter of the shares in the portfolio, those the algorithm selected a year previously, are rejected and replaced with the highest-ranked companies in the list that are not already in the portfolio.

This quarter, seven shares were scheduled for eviction. Publishing, events and training company Euromoney Institutional Investor remains because of its high ranking, so six shares actually departed (see the Nifty Thrifty ejections table above, click to enlarge).

Sales of the six and dividends received since the last update released about £2,014 to invest in each of six new shares, less £10 in lieu of broker fees and just over £10 in lieu of stamp duty.

The six additions are highlighted in the table below. They are giant drug company GlaxoSmithKline; Workspace, which rents out office space; insurance broker and financial adviser Jardine Lloyd Thompson; housebuilder Berkeley; defence and technology consultancy Qinetiq; and environmental and energy consultancy RPS.

The next portfolio reshuffle is in September, when eight companies are scheduled for replacement. The results will be published immediately on MoneyObserver.com and in the October edition of the magazine.

It will be the first update in the portfolio's next chapter of five years, but in some ways we're back to where we started. The algorithm has reverted to its original form. Hopefully, with better data, and without human tinkering, the machine will produce better results.

Subscribe to Money Observer magazine




Richard, In your very interesting article, you wrote: "A lump sum of £30,000 invested in the Royal London fund five years ago would be worth £4,853 today". I would have thought it would be worth a lot more than that - was it a typo?


Hi Peter. Thanks for pointing out the typo. The actual value is £47,853. As you say, considerably more :-)

All the best,


Post new comment

The content of this field is kept private and will not be shown publicly.
By submitting this form, you accept the Mollom privacy policy.