On Monday 28 September, the Polymarket bet on whether the Fed raises rates by a quarter point at its 28 October meeting stood at 65%. On Monday 5 October it stood at 17%. Over the same week the ten-year Treasury yield went from 5.241% to 5.31%, a two-decade high, and the Fed's own measure of the dollar ended Friday 0.6% higher than it began.

When traders expect the Fed to raise rates less, borrowing is supposed to get cheaper. At the short end it did. By Glassnode's count the two-year yield fell and the ten-year rose, so the gap between them went from about 0.2 percentage points around the Fed's rise on 16 September to 0.42 by Friday. The ten-year is the rate behind mortgages, and it's what a bitcoin holder gives up by owning coins that pay nothing.

The week had its reasons for the odds to fall. On Thursday 1 October Philip Jefferson, the Fed's vice chair, said officials may need more time before deciding whether to raise rates again. On Friday the government reported 29,000 new jobs for September, far fewer than economists expected. The yield isn't a smooth line either: on Thursday it touched 5.34% and closed at 5.233%.

On Monday it closed at 5.31%. The Wall Street Journal put that down to a jump in the prices-paid part of a purchasing managers' survey, and to investors wondering whether the Fed would have to raise rates hard after all. Bitcoin closed Monday at $85,839, up 2.7% on the Monday before.

The Fed explains why the odds fell. Whether it explains why the yield rose is the question.

If the Fed were behind the yield, the yield would have followed the odds down, and over these seven days it went the other way. The test is simple. The Fed explanation passes if the ten-year comes back down while the odds stay low. If it doesn't, something else is charging lenders more, and it would have to be a borrower big enough to matter. On Monday a member of the European Central Bank's executive board described one.