[EM] Efforts to improve on CR's strategy

[Bart Ingles]

Gervase Lam wrote:
> 
> 1) Identify the two frontrunners, and pick your favorite among them.
> 2) Give your favorite frontrunner plus the runners who you think are
> better than your favorite frontrunner a score of 2.
> 3) Give your less favored frontrunner plus the runners who you think are
> worse than your favorite frontrunner a score of 0.
> 4) Give the other runners a score of 1.
> 
> Step 4 is the "major" bit.  I had missed out the runners who are in
> between the two frontrunners.
> 
> However, I don't think 3-level CR is as useful in a 0-info context.  For
> the situation I was thinking of using a voting method on, which is why I
> joined this mailing list, I think the context is practically a 0-info
> situation.
> 
> In the 0-info case, the middle level is useful only for runners who are
> exactly in the middle or at least near the middle.  For the other runners,
> you want to maximise or minimise their score in order to ensure your
> ballot gains something.  But how near the middle is near the middle?
> 
> Hmmm.  Thinking about the full-info case, wouldn't it be better to give
> the runners identified in step 4 a score of 2?  This reduces the 3-level
> CR method to just 2 levels (i.e. Approval method).

Well, you could give those "Step 4" candidates whom you consider "better
than average" a score of 2, and the rest a score of 0.  I can't
guarantee that this hasn't already been proposed, though.

Bart