Does Past Performance Predict Future Results? Bill Miller

Past performance does not predict future results in the way most people hope. A long winning record is no promise that the next decade will resemble the last one. What a record does reveal is the method and the temperament that produced it — and those persist. The same disposition that builds a streak is usually the one that ends it.

The clearest case on file belongs to Bill Miller. From 1991 through 2005, his Legg Mason Value Trust beat the S&P 500 — the standard benchmark of large American companies — in fifteen consecutive calendar years,¹ the longest such run a mutual fund manager had posted.² Money arrived on the strength of the streak, most of it late. The streak ended, and in 2008 the fund lost more than half its value.³ Founders who have just sold a business and are interviewing managers tend to meet records like this one at exactly the moment a record is most persuasive.

Does past performance predict future results?

Not as a forecast. Fifteen years of outperformance did not raise the odds that year sixteen would go well. What the record did carry was information about how Miller invested, and that was the genuinely useful part. He ran a concentrated fund — a relatively small number of positions held in large size — and he bought what others were selling, often companies in visible trouble, on the view that the market had overreacted. That approach can produce long stretches of being right. It can also produce a stretch in which the companies the market is abandoning are being abandoned for good reason. Both outcomes come from the same decision rule. So the honest reading of a record is never "this will continue." It is "this is what the person does, and here is what it looks like when conditions turn against it."

Why do fund advertisements say past performance does not guarantee future results?

Because a federal rule requires the line whenever a fund shows returns in an advertisement — and because the required words do less work than a warning should. "Does not guarantee" leaves room for a reader to hear "usually continues, just not for certain." The stricter and more accurate claim is that it does not predict. An advertisement can comply with the rule perfectly well and still be designed so the eye lands on the number and the caution sits in smaller type underneath. The instinct to trust a hot hand is far older than the footnote. For a family office weighing managers, the practical reading is harsher than the legal sentence: the record is finished history, the footnote is the actual point, and any pitch that needs its most important idea in small type is asking the reader to do the harder work unaided.

Was Bill Miller's fifteen-year streak partly an accident of the calendar?

By his own account, yes. The streak was a calendar-year count: in each year from 1991 through 2005, the fund's return exceeded the index's. Writing to shareholders near the end of the run, Miller called it a fortunate accident of the calendar⁶ — December-to-December was the only one of the twelve possible twelve-month windows that produced an unbroken string⁶ — and warned that anyone buying because of the streak was setting up for disappointment. A fund that beat the index in fourteen of sixteen years, with the two misses scattered in the middle, would have finished in roughly the same place and had a far weaker story. The record was real; its tidiness was partly luck. And the tidiness is what carried the weight. "Fifteen straight years" was easier for investors, the press, and the brokers selling the fund to repeat than any description of what he actually owned, or why.

Why did the streak end, and what did 2008 reveal about the method?

The run broke in 2006, when the fund trailed the index,⁴ and the real damage came in 2008. Miller had built large positions in financial companies as their shares fell — Bear Stearns, AIG, Freddie Mac, Washington Mutual and Countrywide among them⁵ — reasoning, as he had for years, that fear had pushed prices below what the businesses were worth. This time the businesses were not worth what he thought. Several were effectively wiped out, and the fund fell far harder than an already brutal market. None of this was a change in method. Buying heavily into fallen companies and adding as prices dropped — asked once when averaging down should stop, Miller answered that it stops when a quote can no longer be had⁶ — was precisely what had produced the streak, including early and unfashionable stakes in technology businesses that a traditional value manager would never have touched. What 2008 exposed was not a hidden flaw. It was the full range of a temperament that had previously only shown its good side.

If the fund beat the market, why did so many of its investors not?

Because a fund's reported return and the experience of the average dollar inside it are two different numbers. The published record tracks a dollar invested at the start and left alone. The dollars that actually arrived came mostly late — present for the end of the run and the collapse, absent for the years that built the record. Assets grew from well under a billion dollars to more than $20 billion around the peak,⁷ and the bulk of that money bought a finished history. The return those dollars earned was dramatically worse than the headline, which is the familiar gap between the fund's math and the investors' math. A long public streak is almost built to widen it, because fame opens the gates widest at the point where the record is longest and no more predictive for having gotten longer. Ask whose return a number describes before asking how large it is.

How much of a long record is skill and how much is luck?

More luck than the arithmetic suggests, and more skill than cynics allow. Miller's own view was candid: the length of the streak owed a great deal to chance, while the underlying approach was deliberate and consistent. That is the right frame for almost any long record. Over fifteen years, a concentrated, contrarian manager who is genuinely good at judging businesses can hold a real edge — but the year-by-year outcomes are noisy, and the odds that the noise lines up into an unbroken run of calendar wins are low even for a skilled person. Underneath sits a second trap: thousands of managers were running at the same time, and the one whose years happened to line up is the one who gets written about. Separating the two components lets allocators ask the useful question — is the method sound, and does it suit this capital — instead of the useless one about whether the dice keep landing the same way.

What does a track record actually tell an allocator?

Three things, if read closely. First, the shape of the method: how concentrated the portfolio was, how long positions were held, what kind of company the manager gravitated toward, and whether the stated philosophy matched what was actually owned. Second, the worst stretch. Every approach has an environment in which it fails, and a record containing one shows how the manager behaved under pressure — held, added, or abandoned the framework. Third, the behavior of the money: when assets grew, and what that growth did to the strategy. A manager who was nimble at a few hundred million dollars may be unable to run the same approach at twenty billion, because the companies that fit the method are no longer large enough to move the result. A record presented as good years only, or small-fund years only, answers none of this. It is an advertisement, not evidence.

What should founders ask a manager who leads with a streak?

The questions a record cannot answer by itself. What does a bad year for this approach look like, and has one happened? What was owned during the best and the worst periods, and does the explanation for both sound like the same person? How much of the record was earned at the current size? What would cause the manager to change the method — and what would cause them to stop? These are close cousins of the questions a serious allocator asks of any wealth manager, and they are the ones Miller's streak quietly discouraged anyone from asking. Founders will recognize the pattern from their own companies: the metric that looked best in the pitch deck was rarely the one that explained the business. The streak was that metric. The investing underneath it — concentrated, patient, willing to be badly wrong — was the business, and anyone who understood it could have described the bad year in advance. This is a historical illustration, not a recommendation of any fund, manager, or approach.