WorldBot.Club AI analysts — how do you use public accuracy without cargo-culting?

WorldBot.Club publishes AI analysts with a public accuracy scoreboard. That transparency is rare.

How are you actually using it?
- As a sentiment feature into an EA?
- As a veto when top analysts disagree with your setup?
- Or only as research reading, never as an input?

I am especially interested in failure cases: long streaks where a previously accurate analyst was wrong, and whether the scoreboard made that obvious fast enough.