Buy & hold vs exit strategies.

We spent months backtesting every "smart" exit strategy you can imagine — trailing stops, profit targets, time-based exits, regime-based exits, Sharpe-optimized exits, combinations of all of them. The data was unambiguous. Buy & hold beat every one of them. Not by a tiny margin. By a wide one. This article documents what we found, what we changed, and what the implication is for how a signal-based newsletter should actually be used.

The conventional wisdom — "let your winners run but cut your losers fast" — is one of those things that sounds like sophisticated risk management and turns out, on data, to be wealth-destroying for the typical case. The intuition is wrong because it ignores how much of a portfolio's return comes from a tiny minority of multi-baggers, and how often the smart-exit rules clip those exact positions out.

What we tested

We ran every published signal in our 87-thesis history through five different exit-management strategies, plus a custom "combined" row. The same entries, the same prices, the same data. Only the exit logic varied:

Why this is true (the math)

Portfolio returns are convex: dominated by the right tail. Across any large basket of stock picks, the median outcome is roughly flat-to-slightly-positive, but a handful of multi-baggers (5×, 10×, 30×) drag the average up massively. Any exit rule that truncates the right tail — and almost all "smart exits" do — hurts portfolio CAGR disproportionately.

Consider: in a basket of 87 names, if 5 of them go up 10× and the rest are flat, your average return is roughly (5 × 900% + 82 × 0%) / 87 ≈ +52% per name. Cap each winner at +200% (a "respectable" profit target) and your 5 winners now contribute 5 × 200% / 87 ≈ +11% — you've cut the basket's return by 80%. That's the cost of a profit target on a convex distribution.

Trailing stops are even more insidious. Most great theses involve mid-trade drawdowns of 30%, 40%, sometimes 50%. NVDA had a ~70% drawdown in 2022 inside a multi-year megathesis. A 25% trailing stop would have exited NVDA at the 2022 bottom — and almost no one buys back into a name they just stopped out of, especially after a painful loss. So the trailing stop doesn't just give up the recovery; it psychologically prevents re-entry.

What this means for how you use the system

Several implications fall out of this finding:

• The entry decision is where the edge lives. If exits can't be improved through cleverness, then the leverage is on selectivity at entry. That's why the conviction model matters: a HIGH-conviction setup held to thesis invalidation is the engine; a MED setup held the same way is the median; a no-conviction setup held the same way is the dilution. See how the conviction score is built.
• Position sizing matters more than exit timing. Since you can't reliably improve outcomes by exiting cleverly, the lever you have left is how much capital you put in at entry. Bigger size for higher-conviction setups; smaller for lower. The Dragonfly recommended-strategy field on each signal card encodes this.
• One catastrophic-stop exception. Buy & hold doesn't mean "ignore everything that happens." If a company's actual financials disintegrate (the auditor resigns, the CEO is arrested, debt covenants break) — the thesis is dead. The Catastrophic Stop is for that, not for price movement.
• Thesis invalidation is the real exit signal. What kills a thesis isn't a 30% drawdown. It's a fact pattern that contradicts the original case: the FDA approval got rejected, the contract went to a competitor, the regulatory tailwind reversed. Exits should track thesis state, not price action.

The honest counter-argument

Some serious investors disagree with everything above. The most common counter:

"Buy & hold is fine for institutional money with a 10-year horizon. Retail investors need to manage drawdown risk because they don't have the emotional capacity to sit through a -50% paper loss without panic-selling at the worst moment."

This is a real point. The data says buy & hold is superior if you can actually hold. The historical record of retail behavior is that most people can't — they sell at the bottom of every drawdown and buy back near the top. If that's the realistic behavior, then a mechanical trailing stop might genuinely save them from worse outcomes than buy & hold would.

Our view: the answer isn't a worse-but-survivable exit rule. The answer is position sizing that lets you sit through drawdowns without panic. If you're sized correctly — meaning no single position can blow up your portfolio — a 50% drawdown on one name is uncomfortable but not catastrophic, and you can wait it out. If you're sized aggressively, the drawdowns force you out at exactly the wrong time, regardless of what the data says.