Conviction Is Not a Strategy
Feeling very sure is a poor reason to make a very big bet. Research can uncover an advantage. Portfolio construction decides how much damage a mistaken estimate is allowed to do.
A lot of “high-conviction investing” is a Luke Skywalker fantasy.
The manager does deep research and acquires a special power: the Force. Systematic investing is the Death Star. Diversification is for people without his gift. He turns off the targeting computer, trusts his conviction, and makes the shot everyone else thought impossible.
In the movie, it works. The script says so.
Your portfolio does not have a screenwriter.
The fantasy is that enough research grants an exemption from uncertainty. Know the business well enough, meet enough people, build a detailed enough model, and ordinary rules about position sizes somehow stop applying.
They keep applying. The future can surprise a careful analyst just as easily as a careless one. Good research can improve the estimate. It cannot remove every way the estimate could be wrong.
Conviction means you feel sure. An edge means you have a defensible reason to expect a better result than the market at the price you pay. Those are different claims. A confident explanation of a company can establish the first while telling us very little about the second.
Here are five familiar arguments for concentrated bets, and the questions they tend to leave unanswered.
“I've done the research, so the risk is low”
Imagine an analyst studying a software business. He reads the filings, interviews customers, talks to former employees, and builds a model with revenue broken into more categories than the company reports.
He learns useful things. Customers like the product. Switching suppliers would be painful. Salespeople see room to sell more services to existing accounts. The business has a real advantage over its rivals.
Then the analyst moves from “I understand this business better” to “I can predict its next five years with great accuracy.” The second statement needs its own evidence.
The customer interviews might all come from the same part of the market. Former employees might describe the business as it used to be. Management might be honest and still underestimate a new competitor. The model might assume strong margins because the last three years had strong margins.
There is also the price. Other investors can know that the product is good. Your research might confirm the market's view rather than reveal a mistake in it.
The analyst has done more work. Whether he has gained a larger investing advantage remains an open question.
A great business can still be a bad buy
Consider a hypothetical company earning $10 per share. Earnings grow 15% a year for five years, reaching about $20.11. The business delivers the growth the analyst expected. At the end, investors value it at 25 times annual earnings, giving a share price of about $503.
Now compare two possible starting prices for that same future business outcome.
| Hypothetical purchase | Starting price | Price after five years | Annualized share-price return |
|---|---|---|---|
| Bought at 20 times earnings | $200 | $503 | 20.2% |
| Bought at 50 times earnings | $500 | $503 | 0.1% |
This made-up example excludes dividends, taxes, and trading costs. The two purchase prices are alternative starting valuations, not prices available simultaneously to different buyers.
The second investor correctly predicts that earnings will roughly double. The share price barely moves because the price paid already demanded so much. Being right about the business was insufficient to earn a good return.
The missing question was what the market had already priced in.
Research should identify the gap between a business's likely future and the future implied by its price. A report praising the product, the management team, and the size of the market does not finish that job.
It helps to keep a record. Write down the forecast before the outcome is known. State the range of possible results, the evidence supporting the view, and the conditions that would break it. Later, compare what happened with what was predicted. Check whether the most confident calls were actually more accurate.
That exercise is less exciting than another management meeting. It also answers a question the meeting cannot: does the process make better predictions?
More hours of research don't automatically mean better predictions. The market does not owe you a return because you did the homework.
“Why own anything except my best ideas?”
Your best idea could be only slightly better than the next one. You could also be wrong about which idea is best.
Suppose a manager ranks ten investments. His estimated annual edge over the market is 3% for the favorite and 2.9% for each of the others. For this illustration, grant him something generous: those estimates are correct in expectation. Each stock has the same market exposure and 20% annual stock-specific volatility, with those stock-specific returns uncorrelated.
Putting everything into the favorite gives an expected extra return of 3%. Spreading the money equally across all ten gives 2.91%.
| Hypothetical portfolio | Expected extra annual return | Annual stock-specific volatility |
|---|---|---|
| Everything in the favorite | 3.00% | 20.00% |
| Equal weights in all ten | 2.91% | 6.32% |
The manager correctly picks the best idea and gives up a large amount of diversification for an extra nine-hundredths of a percentage point of expected annual return. Market risk remains in both portfolios. These are assumed forecasts and risks, not measured manager skill or suggested weights.
Calling a stock “best” ranks it. It does not tell us how much better it is, how reliable the estimate is, or how much money belongs in it.
The same mistake can do very different damage
Take a simpler example. One stock falls 40%. There is no borrowing or trading, and everything else in the portfolio stays flat.
| Starting weight in the stock | Loss on the whole portfolio |
|---|---|
| 5% | 2% |
| 20% | 8% |
| 40% | 16% |
Same company. Same news. Same decline. Different consequences for the investor.
At the smaller weight, the manager can reassess the business with most of the capital intact. At the larger weight, a single mistaken estimate has consumed a substantial part of the portfolio.
This table does not identify the correct position size. It shows why the size needs a separate argument. “We know it well” cannot tell us whether the downside deserves 5%, 20%, or 40% of an investor's money.
There is evidence on the other side
The Best Ideas study finds that managers' inferred strongest ideas outperform, and its authors argue for more concentrated manager portfolios. That is meaningful evidence for specialization. It deserves a serious answer.
The main tests combine ideas across managers. They do not establish that most of a client's wealth should sit in one manager's few favorites. A specialist fund can play a useful role inside a diversified overall portfolio. The question is how much the client puts into it and what risks the client's other investments already carry.
Sometimes the strongest ideas really are much better. The point is to demonstrate how much better and size them with room for error. A preference ranking cannot do that work by itself.
“You have to bet big to beat the market”
To beat the market, a portfolio must differ from the market in useful ways. Simply being different does not make those choices useful.
Put half the portfolio in a weak business and the result will certainly be distinctive. Distinctive is a low bar.
Concentration makes each decision matter more. It magnifies a sound estimate and a mistaken one. It does not improve either estimate.
A Morningstar study published in January 2026, covering more than 5,800 Europe-domiciled equity funds and ETFs over the decade through September 2025, found wider outcomes and deeper drawdowns among concentrated funds, with slightly lower average returns across most regions. That comparison does not prove concentration caused every result. It does challenge the idea that taking more concentrated risk earns a general return bonus.
Three companies can depend on one outcome
Imagine a portfolio with 20% in an online retailer, 20% in a digital lender, and 20% in a software company. The manager has three separate investment reports. The companies serve different customers and sell different products.
Yet all three prices may depend on cheap funding, strong demand, and investors continuing to pay high prices for future growth.
Suppose a common shock cuts all three share prices by 40%, with the rest of the portfolio unchanged. The investor loses 24% of total capital: 60% exposed, multiplied by a 40% decline. There is no leverage in this hypothetical example.
Three reports have produced one very large shared vulnerability.
Now suppose the investor spreads money across five funds. One manager describes the same exposure as quality, another as innovation, and a third as long-term growth. The marketing language changes. The economic dependency can remain.
Five logos do not guarantee five different sources of risk.
The right question is what could hurt the holdings together. Higher interest rates? A slowdown in spending? A closed funding market? Customers cutting the same type of budget? A position that looks small on its own can enlarge a risk already spread across the portfolio.
A manager might reasonably specialize in a narrow set of businesses. The client still needs to understand how that specialty fits into the rest of the client's wealth. The manager's ability to research only a few companies is a limit on the manager's capacity. It does not settle how much money the client should put at risk.
“I invest for the long term, so price swings don't matter”
Patience can be valuable. It needs a reason to keep waiting.
Consider two hypothetical businesses whose shares each fall 40%.
In the first, customers keep renewing, the company generates cash, and its debt is manageable. Investors broadly decide to pay lower prices for growth businesses. The original business forecast still looks reasonable. The lower price might create an attractive opportunity.
In the second, customers are leaving, sales growth relies on heavy discounts, and debt comes due before the business can generate enough cash to repay it. New financing may require expensive borrowing or issuing many more shares. The original forecast is breaking down.
The percentage decline is identical. The investment decisions should depend on very different facts.
The phrase “short-term noise” cannot tell these cases apart. Neither can a promise to hold for five years.
Time does not refinance the debt
Suppose the second company's thesis depends on reaching profitability in year four. Its lenders need to be repaid in year two. The manager can have the correct view of the product's long-term potential and still be wrong about what today's shareholders will own by the time that potential arrives.
The company may survive through a rescue financing while the original owners suffer a large permanent loss. A successful product and a successful investment are separate outcomes.
A long investment horizon should lead to more work on debt, competitive threats, customer behavior, and the path to cash generation. It should also include conditions that would justify selling before the hoped-for future arrives.
“I am a long-term investor” tells us how long someone hopes to wait. It does not tell us whether waiting still makes sense.
The investor has a calendar, too
A fund may be willing to wait. Its investors may need cash for living expenses, a business, or a liability that falls due sooner. They may also have more of their wealth exposed to the same risk than the manager realizes.
The financial effect of the decline matters even when the business eventually recovers. A 50% loss needs a 100% gain to restore the starting wealth. An investor who must withdraw money during the decline has less capital left to participate in a rebound.
That is one reason position size belongs in the discussion before the loss. Telling investors to be patient afterward does not restore the options that an oversized bet took away.
The useful form of conviction stays open to evidence. Write down what would invalidate the thesis while the price is still going up. If every piece of bad news becomes another reason to believe, the thesis has become impossible to test.
“Look at all the great investors who got rich this way”
Big bets can produce spectacular winners. They can also produce spectacular losers. A strategy should be judged using both.
The visibility of the winners creates a trap. The successful investor gives interviews and explains the research that justified the bet. The person who made an equally confident bet in a company that failed is less likely to be asked for a masterclass.
Imagine 100 forecasters each making five independent calls with a 50% chance of being right on each call. Even with no forecasting advantage, we would expect about three to get all five right.
They would have plenty to explain in interviews.
This is a probability illustration, not a claim that every successful manager is lucky. It shows why an impressive selected record cannot, by itself, establish the skill behind it. We need to know how many people tried, how many failed, how much risk they took, and whether the process continues to work on decisions made afterward.
A comeback answers one question
A stock can suffer an extreme decline and then make an extraordinary recovery. That establishes that holding worked for the capital that stayed invested through that particular path.
It leaves other questions open. What was the risk of a permanent loss before the outcome was known? Was the expected payoff large enough to justify the initial weight? How much room did the portfolio have if the recovery took longer or never came? Could the investor remain invested throughout?
Buying after the first spectacular run and selling during the collapse produces a different experience from the manager's full-period chart. The later recovery cannot repair a loss for dollars that have already left.
There is also a difference between reviewing a decision and admiring an outcome. A small probability of a wonderful ending can produce a wonderful ending. We still need to judge whether the bet was sensible when it was made.
A trophy recognizes the period that has ended. New investors are committing money to the period that comes next.
Make conviction earn its weight
The answer is a process that connects research to the amount of money at risk.
Before making a large position, ask:
- What does the price get wrong? Explain the disagreement with the market, as well as the strengths of the business.
- How large is the expected advantage? A modest edge calls for different sizing from a large one.
- What could we lose? Include financing problems, weaker profits, lower valuations, and the possibility that the estimate itself is wrong.
- What do our past forecasts show? Check the original predictions, including the failures, rather than the stories remembered afterward.
- What risks do we already own? Count shared dependencies across stocks, funds, and the investor's other assets.
- What would change the position? Identify facts that would make us reduce it, exit it, or admit the original thesis was mistaken.
A large bet can make sense when the evidence supports a large enough advantage and the investor can bear the downside. It should survive these questions.
Humility should have a visible effect on the portfolio. Allow for smaller edges than the model predicts. Consider larger losses than the base case. Leave enough money outside any one idea to keep investing after a mistake.
Research finds possible advantages. Portfolio construction turns those estimates into a plan that can withstand an unpleasant surprise. Both jobs matter.
Do the research. Keep the controls. You have not acquired the Force.
Humility means leaving room to be wrong. A well-built portfolio does the same.
Example assumptions and research notes
The numerical examples
All worked examples are hypothetical. They illustrate the difference between business analysis, expected advantage, and the consequences of sizing. They do not recommend a stock, a position weight, or a particular number of holdings.
In the valuation example, annual earnings per share grow from $10 to 10 × 1.15^5, approximately $20.1136. The assumed ending price is 25 times those earnings, approximately $502.84. Annualized share-price return is (ending price / purchase price)^(1/5) − 1. There are no dividends, taxes, or trading costs. The ending valuation is an assumption, not a forecast derived from earnings growth.
In the ranking example, all stocks have identical market exposure. The favorite's expected annual alpha is 3%, and each of the other nine has 2.9%. Each has 20% annual stock-specific volatility; those residual returns are assumed uncorrelated. Equal-weight expected alpha is (3% + 9 × 2.9%) / 10 = 2.91%. Stock-specific volatility is 20% / sqrt(10), approximately 6.32%. Market risk remains. Correlated residual risks would reduce the diversification benefit.
The loss examples multiply starting weight by the stock decline, assuming no borrowing, trading, or changes in the other investments. The three-stock shock applies a simultaneous 40% decline to three 20% positions. These are simple one-period loss calculations, not estimates of how likely those shocks are.
The forecaster illustration assumes independent 50–50 outcomes. The expected number with five successes is 100 × 0.5^5 = 3.125. It does not estimate how many actual managers possess skill or the probability that any particular manager was lucky.
Evidence for specialization
Antón, Cohen, and Polk's Best Ideas, April 2021 reports roughly 2.8%–4.5% annual outperformance for inferred best ideas, depending on the benchmark. The authors favor greater manager concentration. Their empirical portfolios combine ideas across managers, and their theoretical discussion considers clients spreading capital across managers. The client-level implications depend on the entire portfolio, including overlap, costs, and constraints. The paper provides a substantive counterargument, not a universal weight for a manager's favorite stock.
Evidence on concentration and outcomes
Morningstar's Bold Portfolios: Are They Worth Their Risks?, January 2026 includes obsolete products and combines stock, sector, and return-based concentration measures. Its performance tests compare subsequent returns across concentration groups. It excludes portfolios with fewer than ten equity positions. The comparisons are observational; fees, investment style, and market conditions can affect the results. The study does not establish a universally optimal stock count.
Framework and attribution
This essay develops the argument in Lukasz Tomicki's Alpha Generation, Betting Discipline, and Portfolio Construction: choosing investments and deciding how much to put into them are separate tasks. Owen Lamont's Acadian essay on concentrated managers helped shape the critique and the use of the Star Wars analogy. The worked numerical examples and the conclusions drawn here are illustrative arguments, not historical fund results.