It’s probably helpful to actually look at evidence and not solely rely on your prior, yes.
Part of the point of my comment is to suggest that “trend continues with some high probability”, and the “trend breaks tomorrow” should be the exception, rather than “”trend breaks tomorrow” should be the default outcome, with “trend continues tomorrow” being the unlikely edge case.
But I agree there are adverse selection effects. If you go around looking for “interesting” straight lines on graphs you are disproportionately likely to trick yourself by bad data and/or the one trend that’s more likely to break (because they’re selected to be interesting).
It’s probably helpful to actually look at evidence and not solely rely on your prior, yes.
Part of the point of my comment is to suggest that “trend continues with some high probability”, and the “trend breaks tomorrow” should be the exception, rather than “”trend breaks tomorrow” should be the default outcome, with “trend continues tomorrow” being the unlikely edge case.
But I agree there are adverse selection effects. If you go around looking for “interesting” straight lines on graphs you are disproportionately likely to trick yourself by bad data and/or the one trend that’s more likely to break (because they’re selected to be interesting).