ICP during major Bitcoin moves

If ICP moved in the same direction as Bitcoin in the window around an event, the correct description is that the two moved together in that window. The incorrect description is that Bitcoin's move caused ICP's. The event study has no mechanism for distinguishing the two, and the common-factor explanation is available for every case: a Bitcoin event large enough to be named is usually a market-wide event, and a market-wide event moves most assets in the same direction.

There is a further problem specific to this comparison. Bitcoin's largest moves are often driven by developments in Bitcoin's own market — a halving, a regulatory decision about a Bitcoin product, a large holder's activity. Those developments have no direct connection to the Internet Computer, so an ICP move in the same window is more plausibly a response to the market-wide conditions that accompanied the event than to the event itself.

The reverse claim is equally unsupported. Nothing in this method shows that ICP influenced Bitcoin, and the relative sizes of the two markets make that explanation implausible on its face. The point is not that one direction is more likely than the other; it is that the method cannot speak to direction at all.

This site's correlation page makes the same point about the statistical measure, and the design-intent comparison covers why the two networks are not substitutes for one another, which is relevant to how much a shared market factor should be expected to move both.

The sample is five events at most, and they are not independent. The November 2022 peak-to-trough and the exchange failure in the same month overlap, so they are two views of one period rather than two observations.

The windows are conventional. A different length would include different days and could change the description of what happened in the window, which is a reason to treat the record as illustrative rather than definitive.

The series are daily closes, so intraday moves are not captured. An event that produced a sharp intraday move and a partial recovery by the close will appear smaller in this record than it was.

Finally, the page does not assert a return for either asset in any window. The event dates and the method are stated so that a reader can reproduce the exercise from the source data, which is where the numbers belong.