Twenty-five crypto assets are worth 2.65 independent bets
60.4% of all variance in a basket of liquid perpetuals is a single common factor. Diversifying across coins mostly buys you the same bet in different colours.
Twenty liquid perpetuals, 293 common days. We measured how many independent bets that basket really contains, using the participation ratio of the correlation matrix.
| effective independent bets, raw daily returns | 2.65 of 20 |
| variance explained by the first component | 60.4% |
| effective bets after removing that factor | 14.24 |
Holding twenty coins feels diversified. Statistically it is closer to holding two and a half. Sixty percent of everything that happens is one factor moving.
Why the second number matters more
Strip the common factor and the residuals are 5.4× more independent. That is the whole argument for market-neutral construction: not that it earns more, but that it lets you learn faster, because you get fourteen observations a day instead of one.
For anyone testing a signal, that is the difference between needing 13 years and needing about eleven months to establish the same edge.
The cost side
Residual moves are 0.63× the size of raw ones and you pay two legs instead of one, so the break-even hit rate rises from roughly 50.24% to 50.76%. Still low. The statistical gain is far larger than the economic cost.