Lord Jubjub
01-October-2006, 02:52 AM
Let's say you have a list of say. . .104 titles. You want to find the consensus top 5 from this list. How does one go about doing this?
Well, it occurs to me that a good series combines wide popularity with a smaller core of passionate supporters. But what is the best statistical balance of these two ways of measurement?
I think the obvious first course is to find the titles with the most widespread popularity--simply to ask whether a certain title is likeable. This allows the pollster to establish a baseline of which titles have the widest group of minimal fans. Asking for top flight material right off the top of a large list will likely result in a wide variety of titles getting a few votes. Asking for likeability first will get a few titles with a huge number of votes and decreasing numbers from that.
The second course is a bit trickier. A title that everyone likes but no one considers great will get a lot of votes for likeablility but may well get no votes for high quality. Using likeability as a filter should allow the second round of polling to be considerably smaller. But should the second round be exclusive (pick the best from this group) or should the choice still be wide open but with a higher standard of quality?
Consider the list of 104 titles. A poll determining simple likeability creates a list from which the top 25 are selected to be put into the second round. There may be a few titles that might have passionate support that do not make that 25, but that pool of supporters is likely too small to be of statistical significance.
But how should the top 25 be presented as to cut down the list to the top 5? The top title on the list of 104 may not be the highest quality. If presented as lists of 5 polls of 5, I think the best order would be to present the places 6-25 in 5 groups of four. Each of the top five is then placed into each those polls to fill them to the 5x5 matrix. The poll can only ask yes or no, it cannot put a grade to the answer.
Should the poll question ask for:
the best title of that poll (only one answer allowed)?
which of the titles would the pollee put in the top 5?
. . . . . . . . . . . . .would the pollee put in the top 10?
That last question is too allow a broader definition of quality that may define a better cutoff between the 5th and 6th place. I am also thinking that the votes for likeability should be brought into play if the first question is NOT used. Each vote for top 5 (or 10) would be worth . . . say five points . . .which would be added to the likeability score. The top five scores would then be presented with the exclusive question:
What is the best title of this list?
Sorry for the bit of rambling, but I thought that I'd rather have a system that has some general support rather than making something up as I go along.
Well, it occurs to me that a good series combines wide popularity with a smaller core of passionate supporters. But what is the best statistical balance of these two ways of measurement?
I think the obvious first course is to find the titles with the most widespread popularity--simply to ask whether a certain title is likeable. This allows the pollster to establish a baseline of which titles have the widest group of minimal fans. Asking for top flight material right off the top of a large list will likely result in a wide variety of titles getting a few votes. Asking for likeability first will get a few titles with a huge number of votes and decreasing numbers from that.
The second course is a bit trickier. A title that everyone likes but no one considers great will get a lot of votes for likeablility but may well get no votes for high quality. Using likeability as a filter should allow the second round of polling to be considerably smaller. But should the second round be exclusive (pick the best from this group) or should the choice still be wide open but with a higher standard of quality?
Consider the list of 104 titles. A poll determining simple likeability creates a list from which the top 25 are selected to be put into the second round. There may be a few titles that might have passionate support that do not make that 25, but that pool of supporters is likely too small to be of statistical significance.
But how should the top 25 be presented as to cut down the list to the top 5? The top title on the list of 104 may not be the highest quality. If presented as lists of 5 polls of 5, I think the best order would be to present the places 6-25 in 5 groups of four. Each of the top five is then placed into each those polls to fill them to the 5x5 matrix. The poll can only ask yes or no, it cannot put a grade to the answer.
Should the poll question ask for:
the best title of that poll (only one answer allowed)?
which of the titles would the pollee put in the top 5?
. . . . . . . . . . . . .would the pollee put in the top 10?
That last question is too allow a broader definition of quality that may define a better cutoff between the 5th and 6th place. I am also thinking that the votes for likeability should be brought into play if the first question is NOT used. Each vote for top 5 (or 10) would be worth . . . say five points . . .which would be added to the likeability score. The top five scores would then be presented with the exclusive question:
What is the best title of this list?
Sorry for the bit of rambling, but I thought that I'd rather have a system that has some general support rather than making something up as I go along.