the person you're directing the question at interprets the guesses as a way to make them look bad,

Okay, I'm curious (sorry!): how do you figure out that was your interlocutor's interpretation? Do they come right out and say so? (I'm guessing not.) Do you infer that from their tone in subsequent messages?

Do you have specific recent cases in mind?

The attitude I try to project is "I'm open to changing my mind on this, but I'll need more solid grounds; please help me out".

Here's an example of how I asked the question on the Good Judgement Project, when I made a forecast that was radically contrarian compared to the team's. The question was "Will at least one individual be convicted of the July 2011 killing of Iranian nuclear physicist Darioush Rezaeinejad by an Iranian court of law before 1 January 2013?".

Either I'm going to get a lousy score on this question, or the team is. If I don't change my forecast, that is. So this seems worth discussing.

For the question to resolve true, there has to be not just a report of an unspecified set of arrest, but an actual arraignment and trial of one or more individuals and a conviction, for this specific murder (i.e. not Roshan's). What makes you 2/3 sure these will all happen in the question timeframe?

I'm not even sure that someone being convicted for merely "being involved" in the assassination would count (e.g. someone who stood lookout but didn't pull the trigger).

I'm ready to revise my forecast if and when a trial is announced, ideally with details of what the charges are and who is accused.