What works in investing: the research behind my approach

In 1992 a small New York firm, Tweedy, Browne, founded in 1920, published a slim booklet with an unglamorous title: What Has Worked in Investing. It collected more than fifty academic and practitioner studies, covering decades of data and many countries, and asked a simple question: which characteristics of shares have actually been associated with higher returns?

The answer was not “the hottest story”, “the most exciting technology” or “what the experts on television are buying”. It was dull, measurable things: shares priced low relative to their earnings and assets, bought with discipline and held with patience. The booklet was updated in 2009 and is still passed between professional investors today.

This guide is my version of that booklet, brought up to date. It explains the approach I teach, why each part of it is there, and the research behind it: from Nobel-prize-winning behavioural economists to the quantitative teams at some of the world’s largest investment banks and asset managers. It is long, because a process you are going to trust with your own money deserves more than a slogan.

Wherever you are in the world. The research in this guide covers the US, UK, continental Europe, Japan and many other markets, and in some cases more than a century of data. The principles apply wherever you invest. Examples are in pounds because most of my readers are British.

Currency guide: £1 is about US$1.34 and about €1.19 (20 October 2026). The £750,000 portfolio used in examples is roughly $1,003,000 or €893,000.

“I don’t ask what a share is going to do. I ask what kinds of shares, bought in what way, have done well for decades, and why.”Alpesh B Patel OBE

The principle: factors, not forecasts

Most people think investing means forecasting: deciding which company will win, which country will boom, which theme will take off. Forecasting is hard, and the evidence on professional forecasters is humbling.

Factor investing asks a different question. Instead of “what will happen?”, it asks: which measurable characteristics of companies and share prices have been associated with better returns, across many markets and many decades, and is there a sensible reason why? Those characteristics are called factors. The best-documented include:

• Value: shares that are cheap relative to earnings, cash flow or assets;
• Quality and profitable growth: companies that earn high returns and grow sales and profits;
• Momentum: shares whose prices have been rising relative to others;
• Lower downside risk: shares and portfolios that fall less in bad times.

No single factor works all the time. The approach I teach combines several, because, as the research below shows, they tend to work at different times. Then it adds the two things that decide whether any of it works for a real person: risk management and behaviour.

Pillar 1: Valuation, because price matters

The oldest and most widely studied factor is value. The idea is simple: if you pay less for each pound of earnings, cash flow or assets, you need less to go right to make money.

What the research found

• Tweedy, Browne’s collection brought together studies showing that shares with low prices relative to book value and to earnings outperformed more expensive shares over long periods, in the US and abroad. One study it cites, by Roger Ibbotson, ranked New York Stock Exchange shares by price-to-book value each year from 1967 to 1984 and found that the cheapest groups had significantly better returns than the most expensive.
• Werner De Bondt and Richard Thaler, the latter a future Nobel prize winner, found in 1985 that portfolios of past “losers” went on to outperform past “winners” over the following years. Their explanation was behavioural: investors overreact, becoming too pessimistic about companies with bad news and too optimistic about companies with good news.
• Eugene Fama and Kenneth French, in one of the most cited papers in finance (1992), found that the ratio of book value to market value was among the most powerful characteristics for explaining differences in average share returns in the US.
• Josef Lakonishok, Andrei Shleifer and Robert Vishny (1994) found that value strategies outperformed “glamour” strategies, and argued that this was because investors extrapolate past growth too far into the future, not because value shares were riskier.

How I use it

The screen I teach looks at a company’s price-to-earnings ratio in two ways: in absolute terms, and against its own history. A share can look cheap against the market but expensive against its own past, or the other way round. Requiring both helps avoid paying too much for popularity.

The honest part

Value does not work every year, or even every decade. Fama and French themselves found that the value premium was much lower in the second half of the period from 1963 to 2019 than in the first half. Anyone who bought only cheap shares through the 2010s had a hard time. That is precisely why value is one pillar, not the whole building.

Pillar 2: CROCI, a sharper measure of real value and real returns

Ordinary accounting figures are not designed for comparing companies. Different countries, industries and accounting choices treat leases, research and development, brands, pension deficits and depreciation differently. Two companies can report the same profit and be very different businesses.

CROCI, Cash Return On Capital Invested, was developed in 1996 in Deutsche Bank’s equity research department, and is now run by DWS, the asset manager that grew out of Deutsche Bank. It adjusts company accounts to put them on an economic, cash basis. According to DWS, the method:

• counts hidden liabilities, such as operating leases, pension deficits and warranties, as debt-like;
• capitalises cash-generating assets that never appear on the balance sheet, such as research and development and brands;
• values assets at their replacement cost, adjusted for inflation;
• depreciates similar assets in a similar way, whatever the accounting rules.

The result is a cash return on the real capital a business uses, and an economic P/E that can be compared fairly across sectors and countries. In plain English: CROCI tries to answer “how much real cash does this business generate on the money tied up in it, and how much am I paying for that?”

How I use it

CROCI is one of the factors in the screen. It acts as a quality check on value: a cheap share with poor cash returns may be cheap for a reason. A business with strong and improving cash returns, at a sensible price, is what I am looking for.

Pillar 3: Growth, but the right kind of growth

If value were the whole story, investors would simply buy the cheapest shares and wait. But some of the best long-term investments are companies that keep growing sales, profits and cash flow for years. The research helps separate good growth from expensive hope.

What the research found

• Robert Novy-Marx (2013) found that profitability, measured by gross profits relative to assets, predicted differences in returns about as powerfully as value did. Profitable companies earned significantly higher returns than unprofitable ones, even though they tended to look more expensive. Combining profitability with value improved the results of value strategies, especially among the largest, most liquid shares.
• Fama and French later added profitability to their standard model of share returns, a recognition that quality of earnings matters, not just price.
• Research on earnings announcements has long found that share prices tend to keep drifting in the direction of an earnings surprise for weeks after the announcement, a pattern known as post-earnings-announcement drift. Companies that beat expectations often keep doing so for a while.

The warning in the same research is just as important: “glamour” growth shares, bought at very high prices on the hope that past growth will continue for ever, have historically disappointed. Lakonishok, Shleifer and Vishny’s work is, in part, a study of that disappointment.

How I use it

The screen requires growth in both sales and earnings. Sales growth alone can be bought; earnings growth alone can be engineered for a while. Both together, alongside sensible valuation and strong cash returns, point to a business that is genuinely getting better. My four-step way of thinking about a great company is: growth in profits, sales and cash flow; beating what the market expects, often through innovation; and then sustaining it.

A positive dividend yield is also part of the screen. A company that pays a dividend has to produce real cash and tends to be more disciplined with it. It is a modest signal, used as one vote among several, not a reason to buy on its own.

Pillar 4: Trends and momentum, because prices carry information

Value says “buy what is cheap”. Momentum says “buy what is going up”. They sound contradictory. The research says they are complementary, and that is one of the most useful findings in modern finance.

What the research found

• Narasimhan Jegadeesh and Sheridan Titman (1993) found that US shares that had performed best over the previous 3 to 12 months tended to keep outperforming over the next 3 to 12 months, with significant profits in their 1965 to 1989 sample. Part of the extra return faded in the following two years, which is why momentum needs regular review.
• At Goldman Sachs Asset Management in the mid-1990s, Clifford Asness was asked to set up a quantitative research group. He and his colleagues built models combining Fama and French’s value insight with Asness’s own research on momentum: the cheapest shares, but only those that seemed to have started moving up. Within two years the group was running billions of dollars, before Asness and colleagues left in 1998 to found AQR Capital Management.
• Asness, Tobias Moskowitz and Lasse Pedersen (2013), in “Value and Momentum Everywhere”, found value and momentum premiums in eight different markets and asset classes: shares in the US, UK, continental Europe and Japan, plus country stock indices, government bonds, currencies and commodities.
• Brian Hurst, Yao Hua Ooi and Lasse Pedersen studied trend-following across global markets back to 1880. They found positive average returns in every decade, low correlation with traditional assets, and good performance in 8 of the 10 largest crisis periods for a traditional 60/40 portfolio over that time.

How I use it

The screen includes six-month upward price momentum. It means a company has to be cheap and good, and the market has to have started to agree. That helps avoid “value traps”: shares that are cheap and stay cheap. I also use a disciplined reading of the MACD indicator across monthly, weekly and daily charts to judge whether a market’s trend is strengthening or weakening.

The honest part

Momentum can reverse violently, especially when markets rebound sharply after a long fall. Kent Daniel and Tobias Moskowitz documented these “momentum crashes”. In the rebounds of 1932 and of March to May 2009, past losers shot up far faster than past winners; in 2009 a strategy that bought past winners and sold past losers short lost more than 70% in about three months. Momentum is powerful, but it has to sit inside a broader process with risk controls.

Pillar 5: Correlation, why combining factors matters

Here is the key insight that ties the pillars together. Asness, Moskowitz and Pedersen found that value and momentum were negatively correlated with each other: when one struggled, the other often did well. A simple combination of the two had much better risk-adjusted returns than either on its own.

Novy-Marx found something similar for value and profitability: they lean in opposite directions, one towards cheap and often troubled companies, the other towards strong companies that usually cost more, so combining them smooths the ride.

This is why the screen requires a company to pass most of several different tests, rather than excel at one. It is diversification by reason for owning, not just by number of holdings.

The trap of “diversification” that disappears

Many investors believe they are diversified because they own many funds or many countries. François Longin and Bruno Solnik studied the five largest stock markets (the US, UK, Japan, Germany and France) and found that correlations between markets increase in bear markets, but not in bull markets. In other words, the diversification you think you have can shrink just when you need it most.

The lesson is not to give up on diversification, but to understand it properly: own things that are held for different reasons, and manage the risk of the whole portfolio, not just each holding.

Pillar 6: Sortino, measuring the risk that actually hurts

The most common measure of risk-adjusted return is the Sharpe ratio, which treats all volatility as risk, including sharp rises. But no investor ever complained that a share went up too fast. The risk that hurts is the downside.

In 1994 Frank Sortino and Lee Price set out a way of measuring performance that counts only downside deviation: returns below a target. The Sortino ratio compares the return above that target with the downside volatility.

Two investments can have the same average return and the same overall volatility, and be completely different to live with. One rises in jumps and falls gently; the other rises gently and falls in jumps. The Sortino ratio tells them apart.

How I use it

The screen includes the Sortino ratio, alongside alpha: the return a share has produced beyond what would be expected from the market’s movements alone. Together they favour shares that have rewarded investors well for the downside risk they carried.

Putting it together: the screen

The Great Investments Programme teaches a screen built on the research above. A company is scored on eight tests:

Test	What it looks for	Research behind it
1. Sales growth	A business that is genuinely growing	Profitable growth: Novy-Marx; Fama and French
2. Earnings growth	Growth that reaches the bottom line	Earnings surprise and drift research
3. Valuation (P/E, absolute and against its own history)	Not overpaying	Tweedy, Browne; De Bondt and Thaler; Fama and French; Lakonishok, Shleifer and Vishny
4. Positive dividend yield	Real cash and capital discipline	A supporting signal of cash discipline, not a reason to buy on its own
5. CROCI	Strong cash return on the real capital invested	Deutsche Bank / DWS CROCI methodology
6. Sortino ratio	Good reward for downside risk	Sortino and Price
7. Alpha	Returns beyond what the market alone explains	Standard performance measurement
8. Six-month upward momentum	The market has started to agree	Jegadeesh and Titman; Asness, Moskowitz and Pedersen; Hurst, Ooi and Pedersen

A company must pass at least six of the eight to be considered. No single number decides. That is deliberate: it combines factors that tend to work at different times, and it makes it much harder for one exciting story to override the evidence.

Passing the screen is the start, not the end. It produces a shortlist to research, not a list of instructions.

Risk management: drawdowns and recovery times

The most important number in investing is not your best year. It is your worst fall, and how long it takes to recover. The arithmetic is unforgiving:

Fall	Gain needed to get back	Years to recover at 7% a year	On a £750,000 portfolio
10%	11.1%	about 1.6	£75,000 lost
20%	25.0%	about 3.3	£150,000 lost
30%	42.9%	about 5.3	£225,000 lost
50%	100.0%	about 10.2	£375,000 lost
78%	354.5%	about 22.4	£585,000 lost

The 7% a year is an illustration of steady growth after a fall, not a prediction. Real recoveries can be faster or much slower. Two real examples:

• The Nasdaq Composite closed at a record on 10 March 2000, fell about 78% by October 2002, and did not close above that record again until 23 April 2015, more than 15 years later.
• The FTSE 100 set a closing record of 6,930 on 30 December 1999 and did not close above it until 24 February 2015. Dividends made the real experience much better than the index price suggests, but the price took over 15 years to recover.

This is why the approach I teach is built around limiting large drawdowns rather than chasing the biggest possible gains: sizing positions so no single holding can do serious damage, rebalancing within set bands and trimming winners back to target, having pre-set rules for when to cut a position, and judging the market’s overall trend before adding risk. Avoiding the 50% fall does more for long-term wealth than catching the occasional 100% rise.

Behaviour: the part that decides whether any of this works

Every pillar above can be undone by the investor. Two Nobel prize winners spent their careers showing how.

What Kahneman, Tversky and Thaler found

• Loss aversion. Daniel Kahneman and Amos Tversky’s prospect theory (1979) showed that losses loom larger than gains. In many experiments, people need a potential gain of around twice the size of a potential loss before they will accept an even bet. Kahneman received the Nobel prize in economics in 2002.
• Myopic loss aversion. Shlomo Benartzi and Richard Thaler (1995) showed that combining loss aversion with checking your portfolio too often makes investors behave as though shares are riskier than they are, and hold too little of them. The more often you look, the more losses you see, and the more frightened you become. Thaler received the Nobel prize in 2017.
• Overreaction. De Bondt and Thaler’s work on past losers and winners showed investors becoming too pessimistic after bad news and too optimistic after good news. That overreaction is part of why value works.
• The disposition effect. Terrance Odean (1998) found that individual investors were much more likely to sell shares that had gone up than shares that had gone down: selling winners too early and holding losers too long, the opposite of what momentum research suggests.
• Overconfidence. Brad Barber and Terrance Odean found that the US households that traded most earned about 11.4% a year in the 1990s, against 17.9% for the market, largely because of costs and overconfidence.

How the process protects you from yourself

• A written investment policy that sets out your goals, risk limits and rules, with a cooling-off period before you are allowed to change it, so a bad week can’t rewrite your plan.
• A monthly review rhythm, not daily price-watching, directly addressing myopic loss aversion.
• Pre-trade checklists that every buy and sell must pass, so stories and tips can’t skip the evidence.
• Rules for selling losers and trimming winners, decided in advance, to counter the disposition effect.
• A decision journal, recording why you made each decision and how you felt, so you learn from your own patterns rather than repeating them.

Expectation management: what to expect, honestly

A process that you abandon at the worst moment is worse than no process. So it is important to know, in advance, what any evidence-based approach will feel like in real life:

• There will be losing months and losing years. Every factor in this guide has had long periods of disappointment: value through much of the 2010s, momentum in sharp rebounds like 2009.
• There will be periods when a simple index fund does better. That is the price of doing something different. Judge the process over years, against a fair benchmark, after costs.
• Drawdowns will happen. The aim is to keep them survivable, not to avoid them entirely. Decide now what size of fall you can live with, and size your risk to match.
• Back-tested results always look better than real life. Research uses historical data, without the emotions, costs and taxes you face. Treat any back-test, including mine, as evidence about a method, not a promise.
• Be wary of anyone promising guaranteed or unusually high returns. No legitimate process can, including this one.

“The evidence tells you what has worked. Your discipline decides whether it works for you.”Alpesh B Patel OBE

Before you change how your money is invested

If this approach appeals to you, the next step is to learn it, not to move money in a hurry.

What this taught me

When I first read Tweedy, Browne’s booklet, what struck me was not any single study. It was that the same handful of ideas kept appearing, decade after decade, market after market, discovered by different people with different data. Cheap tends to beat expensive. Good businesses tend to keep being good. Trends tend to persist for a while. And human beings, given the chance, make the same mistakes over and over again.

What this taught me is that you don’t need to predict the future to invest well. You need a process built on what has worked, the discipline to follow it, and realistic expectations about the bumps along the way. So the question to take away is: is your money invested by evidence, or by story?