This article describes a structural property of crypto markets, using our own experiment and public history. It predicts nothing about the current regime. Nothing here is investment advice. Decisions are yours.
Crypto's correlation structure is asymmetric: in bull tape the board disperses, assets have their own seasons, and picking matters; in stress the market collapses onto one factor and everything becomes a single BTC trade wearing a hundred tickers. Every diversification claim in this asset class lives or dies on that asymmetry.
The two regimes of togetherness
Watch the board through a full cycle and the correlation weather is unmistakable. In sustained uptrends, participation broadens and disperses: sector families take turns, leadership rotates, and the gap between the board's best and worst performers widens enormously, which is alt season, measured. In drawdowns the dispersion vanishes: pairwise correlations converge toward one, and the daily question stops being "which assets" and becomes "how much crypto," because the marginal seller is exiting the asset class, not rebalancing within it. The asymmetry has a plain mechanism: buying is a choice among stories, selling is a flight from a category, and forced selling, meaning liquidation cascades, is the least selective flow in finance.
Our own demonstration, paid for in public
The transplant experiment ran the cleanest version of the lesson we could have designed, though not on purpose: four concurrent positions, individually sized, individually stopped, spread across different alts, drew down 74% together, behaving as one leveraged BTC position from the moment the tide turned. Every piece of per-position risk arithmetic was intact and every piece of it was answering the wrong question, because the unit of risk was not the position, it was the regime. That finding reshaped the sizing arithmetic this library teaches, heat capped as one correlated bet, and the ruin math that follows from it.
What the one-factor problem breaks
Three standard tools, each quietly voided. Diversification: spreading across twenty alts diversifies the bull experience, narrative exposure, and nothing else; the tail risk remains one bet, so the risk-reduction equities buy with breadth simply is not for sale here at any number of tickers. Correlation-based sizing: models calibrated on calm-period correlations understate stress co-movement by construction, the classic it-worked-until-it-mattered failure. And stock-picking defensiveness: "quality" alts fall with the junk in the flush, differing in recovery, not in participation, which is why drawdown depth across the board clusters so tightly in winters. What survives is exactly one distinction: BTC and the largest majors fall less at the index level than the alt tail, which is a statement about magnitude, not independence.
Designing for the asymmetry
The structural answer is to measure both faces of the regime separately, which is precisely how the dial is built: a BTC trend condition, because the one factor's own state decides the stress regime, and a breadth condition, because dispersion-versus-collapse is visible in participation before it is visible in any single price. Bull breadth says selection is being paid; bear correlation says selection is theater and exposure is the only decision. A participant who carries one habit out of this piece should carry the interrogation habit: whatever the portfolio looks like, ask what it becomes on the day the one factor turns, because that day, not the average day, is when its true position count is revealed. Most crypto portfolios, honestly counted, hold one position. Decisions are yours.