Conviction is overrated: does "backing your best ideas" actually pay?

Last reviewed 17 September 2026.

You have probably been told that the way to build wealth is to find your best ideas and bet big on them. "Concentrate. Diversification is protection against ignorance. Put your eggs in one basket and watch the basket." It is a seductive story because it flatters the part of us that wants investing to be a test of insight rather than of temperament. And it contains a real truth: over-diversifying into fifty half-understood positions is its own kind of mistake.

But there is a quieter question hiding inside the advice, and almost nobody checks it against their own record: when you feel high conviction, do you actually earn more? Not "did Buffett's concentrated bets work" — survivorship bias makes that unanswerable — but in your own closed trades, did the ones you rated a 9 or 10 out of 10 at entry beat the ones you rated a 5 or 6? That comparison is fully within reach with nothing more than a spreadsheet and some discipline about sequencing.

How to test conviction against return by hand

You do not need software to run this test — you need discipline about when you record the rating. The whole exercise falls apart if you score conviction after you know the outcome, because memory rewrites the past. So the manual method is strict about sequencing:

Step 1 — Rate at entry, never later. The moment you open a position, write down a conviction score from 1 to 10 and one sentence on why. Do it before the price moves. If you are honest, most trades will not be 10s; a 10 should be rare.

Step 2 — Freeze the rating. Do not "update" conviction as the trade runs. The rating is a snapshot of what you knew and felt at entry. Editing it later is exactly the hindsight bias you are trying to catch.

Step 3 — Record realised return at exit. When the position closes, log the actual percentage return, net of commissions and any tax you can attribute. Half-exits count as partial closes at the price you got.

Step 4 — Bucket and take the median. Group closed trades into conviction bands — say Low (1–4), Medium (5–7) and High (8–10). For each band, take the median realised return, not the mean, because one lucky moonshot drags a mean around and misleads you. Write down how many trades sit in each band. Sample size is the whole story here.

Step 5 — Don't trust a thin band. Suppose across a year you close 30 trades: 8 Low, 14 Medium, 8 High. You line up each band's returns, sorted, and read off the middle value. If Low medians at +2%, Medium at +9% and High at +7%, the naive reading is "medium conviction wins." But with only 8 trades in each outer band, the difference between +7% and +9% could easily be noise. As a rough floor, treat any band with fewer than about twenty trades as too thin to draw a conclusion from.

Why conviction and return come apart

The gap between how sure we feel and how things turn out is one of the best-documented findings in behavioural finance. Confidence is generated by the coherence of a story, not by its accuracy. A stock with a clean narrative — great product, loved brand, obvious growth — produces high conviction precisely because the story is easy to tell. But an easy story is often an already-priced story: if the bull case is obvious to you, it is obvious to the market, and it is in the price. The trades where you feel a nagging 6-out-of-10 are frequently the ones with a real edge and a messier narrative.

There is also a compounding problem: overconfidence, anchoring and the sunk-cost reflex all cluster around high-conviction positions. When you are a 9 out of 10 on an idea, you hold it too long when it turns, you average down without re-underwriting, and you dismiss disconfirming news as noise. The conviction that got you in becomes the bias that stops you getting out. Overconfidence specifically tends to show up not as a worse win rate but as a wider spread of outcomes — the trades you feel surest about are often the ones taken with the least hedging and the largest size, so when they are wrong, they are wrong more expensively.

None of this means conviction is worthless. It means conviction is information about you, not about the asset. The useful move is to make it measurable: rate it, freeze it, and check it against outcomes over enough trades that the noise averages out. Do that for a year and you learn something no market commentator can tell you — whether your conviction is calibrated. Some investors' high-conviction trades genuinely do outperform; most people simply never check, and assume they belong in the first group.

What this doesn't tell you

This is a behavioural observation, not a strategy. It does not tell you to concentrate or to diversify; that is a portfolio-construction decision with your own risk tolerance attached. It says nothing about whether any specific stock is a buy, because it is not about stocks at all — it is about the reliability of a feeling. Median return by band also ignores how long capital was tied up: a strong return held for two years is worse, risk-adjusted, than the same return held for four months, which simple bucketing by band does not correct for — a proper treatment would look at annualised, risk-adjusted return, which is exactly why metrics like Sharpe and Sortino exist. And if you day-trade, none of this applies at your time horizon — this method assumes decisions reviewed over months and years, not minutes.

Frequently asked

Does high conviction investing produce better returns?

Not reliably, and rarely in the way people assume. Confidence is generated by how coherent a story feels, not by how accurate it is — an easy, obvious bull case is often already priced in. Whether your own highest-conviction trades actually outperform your medium-conviction ones is a testable question, not a matter of opinion: rate conviction at entry, freeze the rating, and compare median realised return by band once you have enough closed trades.

How do I measure whether my own conviction is calibrated?

Record a conviction rating at entry, do nothing to it afterwards, and compare it to realised return once each position closes. Group your trades into conviction bands, take the median return per band, and note how many trades sit in each band. If your highest band does not beat your middle band across at least a few dozen closed trades, your conviction is not yet calibrated.

This article is for information only. It is not financial, investment or tax advice, and nothing here is a recommendation to buy or sell any security. StackViewer analyses decisions you have already made; it does not predict returns. Past performance does not indicate future results.