Grayscale's research desk just published a number that will be screenshotted into a thousand group chats.
Over the trailing three years, they report, Bitcoin returned 225%. Remove the five best trading days and that collapses to 95%. Remove ten and you're left with 27%. Remove fifteen and the entire period goes negative โ minus eleven percent. The Nasdaq, offered as the sober comparison, turns 109% into 21%.
The conclusion writes itself: you cannot time this market, so you must never leave it. Time in the market beats timing the market. Hold through everything.
I have spent enough hours auditing return tables and smart contracts to be suspicious of a statistic this clean. The number is real. The framing is not. And the half of the arithmetic Grayscale omitted is precisely the half that would have changed how you size your position.
Let me show you the hole.
Context: what Bitcoin is actually being measured against
Start with the asset. Bitcoin is not a company. It has no earnings, no cash flow, no board. It is a monetary protocol with a hard cap of 21 million coins, an issuance schedule that halves roughly every four years, and a value proposition that rests entirely on scarcity and settlement guarantees. That design produces something the Nasdaq structurally cannot: violent, self-reinforcing volatility.
A stock index is anchored by discounted cash flows. Bitcoin is anchored by belief about the future of money. When that belief shifts, it shifts fast, in both directions โ which is exactly why its returns cluster into a handful of sessions.
This matters because the Grayscale comparison โ 225% against 109% โ flattens two completely different return engines into one chart. One is a profit-driven cash-flow asset. The other is a monetary-substitution narrative. Their risk sources are not comparable, and neither is the meaning of their concentration.
The publisher is not a neutral party, either. Grayscale sits under DCG, manages one of the largest crypto vehicles in the United States, and earns its revenue as a percentage of assets under management. After GBTC converted to a spot ETF, its 1.5% fee became conspicuous against cheaper competitors. Every research note that encourages clients to stay invested is, structurally, a client-retention document. That doesn't make it false. It makes it motivated.
The author, Zach Pandl, is a genuinely respected macro strategist who moved from traditional asset management into crypto. That pedigree matters. His framing โ long horizon, stay invested, ignore the noise โ is the standard posture of sell-side research, transplanted wholesale onto an asset that behaves nothing like a blue-chip equity.
Now the methodology. This is where the real defect lives.
Core: the asymmetry nobody screenshotted
Concentration of returns is not a Grayscale discovery. It is one of the oldest results in asset research. Study the S&P 500, study gold, study any high-volatility instrument, and you will find that a tiny fraction of trading days โ often under 0.5% โ accounts for the bulk of long-run performance. That finding is robust and true.
But look at how Grayscale presents it. They remove the best five days, then the best ten, then the best fifteen. A clean, escalating sequence of decay: 225, then 95, then 27, then minus 11.
What they never do is remove the worst days.
This is not a rounding error. It is the entire argument. Financial literature has repeatedly demonstrated that the best and worst sessions cluster inside the same high-volatility windows โ the same halving, the same ETF approvals, the same macro shocks. When you delete the best days, you are usually deleting the worst days sitting right beside them. A symmetric test would almost certainly show that removing the ten worst sessions lifts the cumulative return substantially, not crushes it.
Grayscale ran one half of an experiment and published the half that supports holding.
There is a second lever, quieter and just as powerful: window selection. The 225% figure is hypersensitive to where you start and stop the clock. Begin at a cycle bottom, end near a relative high, and concentration inflates. Invert those endpoints and the same method tells a different story. Grayscale never discloses the exact start and end dates. A three-year sample is roughly 1,095 trading days โ statistically short, and silently adjustable.
That is the tell of every motivated study: real data, honest inputs, and a frame that quietly decides which conclusion survives.
Here is the mechanism most readers miss. Return concentration is not evidence that Bitcoin is special. It is a mathematical byproduct of volatility itself. Any asset that swings hard enough will show a lopsided return distribution โ including the hundreds of tokens that eventually went to zero. Concentration measures turbulence, not scarcity. The scarcity narrative is being smuggled in on the back of a statistic that would look identical for assets that failed.
And the direction of the finding is not fixed. Re-run the identical method over a bear-market three-year window and it would produce the opposite headline: timing โ or simply not being present โ spared you catastrophic losses. The math does not choose a side. The window chooses it for you.
Contrarian: staying in has a price too
The narrative's emotional payload is the cost of missing out. What it never prices is the cost of staying.
Bitcoin has drawn down more than 70% multiple times in its history. A holder who stayed in through those windows did not avoid a cost โ they absorbed one, in drawdown, in forced patience, in sleepless quarters. The mirror proposition is just as true as the original: staying in has a cost, and it is the cost you haven't seen yet.
If you re-run the numbers removing the worst ten sessions, the cumulative return likely improves. That single line would demolish the implicit claim that you must endure every unit of volatility to capture the upside. Grayscale's framework jumps straight from timing is hard to therefore never time. It skips the middle ground entirely: position sizing, valuation bands, risk budgets. Those tools reduce drawdown without requiring you to predict the next five sessions.
Takeaway
History doesn't repeat, but it does rhyme with a window someone chose not to disclose. The concentration statistic is genuine, and it is worth knowing. The conclusion built on top of it is a sales document wearing a lab coat.
Run the symmetric test yourself. Publish the start and end dates. Then decide whether always-in is a strategy โ or a fee structure with a chart attached. Read the data. Discount the conclusion. Re-price your own tolerance for a seventy-percent hole.
Because the days that make you rich and the days that ruin you are the same days. The only question is whether you were sized to survive them.