What research says: what works in the long run and what does not

Transcript

316. That is how many patterns three researchers counted in finance papers up to 2012, each supposedly predicting which stocks will do better. If all of them were true, investing would be easy. It is not. I'm Peter. Today we look at what research has really found over decades: what does not work, what holds up surprisingly often, and what that means for how you pick.

Let's start with the professionals. S&P Dow Jones Indices has been comparing active funds with their index for more than twenty years. Over the fifteen years to the end of 2025, about nine out of ten US large company funds trailed the S&P 500. And in not a single fund group they looked at did a majority beat its own benchmark over that time. One important reason is unspectacular: costs. Every percentage point in fees is missing every single year.

And all those patterns? Researchers McLean and Pontiff re-checked 97 of them. After publication they delivered on average 58 percent less than in the paper. Part of it was chance in the data, part was traded away once everyone knew. That is why Harvey and his colleagues ask for a much stricter test before anyone trusts a new pattern. Remember: a nice look back is not yet proof.

Now for what holds up. Back in 1977, researcher Sanjoy Basu showed that stocks with a low price-to-earnings ratio did better. In 1992, Eugene Fama and Kenneth French found the same picture with data from 1963 to 1990: companies with a low price relative to their book value earned more on average than expensive ones. Fama won the Nobel Prize in 2013. But to be honest: this edge does not show up every year. From 2007 to 2020, cheap stocks trailed expensive ones for about thirteen years. Buying cheap takes patience.

The second finding is almost the mirror image. In 2013, Robert Novy-Marx showed that companies earning a lot of gross profit relative to their assets did better, even though they usually look more expensive. Combined with a low valuation, it worked even better. Asness, Frazzini and Pedersen reached a similar conclusion in 2019: good companies, meaning profitable, growing and soundly financed, earned more than weak ones relative to their risk. And on average they were only a little more expensive, not enough to make up for it.

A third pattern is surprisingly robust: trend. In 1993, Jegadeesh and Titman showed that stocks that did well over the past three to twelve months often kept going for a few more months. The catch: the trend sometimes turns abruptly, and chasing it all the time costs a lot in fees. That is why the watchlist uses the trend only as a sign to take a look, not as a starting gun to trade.

So how often do you get both, cheap and good? We counted on October 19, 2026: more than 28,000 stocks with a fair value and a quality score. 44 percent were below their fair value. 30 percent had a quality of 60 or more out of 100. Both together: only 17 percent, about one in six. You can find exactly this overlap in the Pro screener with two sliders, instead of clicking through thousands of stocks.

In short: most funds lose to their own costs, many patterns vanish once people know them, and what has held up for decades is surprisingly simple: a low valuation, good quality, patience. None of it works every year. Important: this is not financial advice.