I don't know if you've given any thought to this, but as I answered the poll for this month, it occurred to me that I couldn't be very accurate with some of the lower earning sites because of the way we're instructed to answer.

It says you must select the number where you have at least that amount, not the number closest to the amount you earned.

As an example, in September on PhotoDune I earned just shy of the $50 payout. I had to answer $20 as that was the next lower choice, but I was within $4 of $50, so really answering $50 would have given a much better picture of where PD stood. I answered $20 for BigStock too - they were about $30 for the month. So for the poll, the two sites look equivalent, but they really aren't (although if BigStock continues to be un-dead, things may turn around in a month or two!)

you have a point but when things are going to be rounded one way or another they are going to be less accurate than they could be.  Even if we just rounded to the nearest answer we'd see the same problem.  If you earn $36 on site A and $74 on site be they would both get the $50 answer.  When people get close to voting the next level though, there will be others who did mange to break the next level.  There will also be those who had bigstock below the $20 mark so I think all in all the average will work out to give us some good pointers.

Quote from: leaf on October 02, 2012, 12:50

.. Even if we just rounded to the nearest answer we'd see the same problem.  If you earn $36 on site A and $74 on site be they would both get the $50 answer.  When people get close to voting the next level though, there will be others who did mange to break the next level.  There will also be those who had bigstock below the $20 mark so I think all in all the average will work out to give us some good pointers.

A. if you round to nearest number, the biggest error is half of the interval. $15 for [20 to 50]
B. if you round to low boundary, the biggest error is interval (-1 cent), i.e. almost one interval. $29.99 for the same case

average error, for method A would be 1/4 of interval {(min + max) / 2 = (0 + 15) / 2 = $7.5}, method B average error is 1/2 of interval = $15 , so accuracy of A is better

Yeah, you are right, A is more accurate, and perhaps using method A in the polls would be better - if for nothing else but our sanity.
In the polls though, we aren't looking for an accurate number per say, just an accurate ranking, one site to another.  Both A and B method have the same potential to have two voters with nearly $40 difference in earnings vote the same amount.
.. anyhow the difference isn't much and i agree it would make more sense to do it the more mathematically correct way so I'll change the instructions.