I find it a little odd that I am even in that list. I believe I have more votes on this site than anyone, and I am still listed in the top 100 people who don't vote? I even spent a couple weeks doing nothing but digging through questions specifically looking for those people who were new to the site, and needed rep to be able to effectively use StackExchange's features, and gave them up votes...on both questions and answers. I also go through the top users list every week or two, and look through the top up-and-comer. If their answers (or questions) are solid, I go through them all and give an up vote (or downvote) where deserved.
Given that, I believe that query might suffer a little bit from some volume anemia. Statistical calculations often require a large pool of source data to provide usefully accurate results. We still have a fairly small user base, and I don't know that we can really state conclusively one way or another yet whether people are adequately voting or not. Granted, the top users from Rowlands query have a high enough ratio that they probably needed a kick in the butt, but the lower half of that list seems to be doing a decent job. The bottom half of users had both up and down votes in decent quantity for the volume of our site, and given the age of many of their memberships. We can't forget that people vote what they see...as time goes on, there will be more and more new members who may indeed be active voters...but whom will likely not go way back in time to vote on old questions. That will ultimately skew the results of such a query if it is not restricted to a more recent time periods.
Statistically, I think that list is interesting, but not particularly useful at the moment. At least, not by selecting the top 100 users. It might be more useful to select the top 25 users for now, and as our membership grows, select more. From a statistical standpoint, I think a smaller aggregation (25, maybe even just 15) would provide a higher ratio of useful results to distinct data points than an aggregation of the top 100 users.
I think the fundamental premise behind the aggregation here should be examined. The basic algorithm used is as follows:
Ratio = (100 * Rep/10) / (Upvotes+1)
First, this does not take into account all of a persons voting, it only takes into account up votes. Down votes count too according to the discussion here, and should be factored in:
Ratio = (100 * Rep/10) / (Upvotes + Downvotes + 1)
Now, this divides rep by 10, then multiplies it by 100. That is effectively the same as multiplying it by 10. I think this might have been an error in parentheses, though.
A more useful rating might be the following:
Ratio = (((Upvotes + Downvotes + 1) / (Reputation/10)) * 100) - 100
This equation provides a useful division between all of the members involved in the query. Anyone with a rating over 100% has a positive voting ratio when compared to their reputation, where as anyone with a rating under 100% has a negative voting ratio when compared to their reputation. In the case of myself and Rowland, we would both have a positive ratio, which I believe is more appropriate. The top few members from Rowlands current query would have a negative ratio (which might give more punch to that kick in the butt. ;)) By turning ratings into scalar numbers, you now effectively have a voting index, which ranks voters from best to worst (or worst to best, however you choose to sort) by what I hope is a more useful value.
Here is a comparison of the results between the ratio of the top few people in Rowland's query, and Rowland's & my own ratio:
| Ratio | "Voting Index"
First Entry | 1323.08% | -93
Second Entry | 1214.29% | -91
Jon | 68.72% | +47
Rowland | 45.59% | +121
Rowland and I, as well as many other people here, are good voters. Rowland in particular is STELLAR, and has put his time into tweeting and sharing links to boost the sites visibility as well (I'm not a tweeter myself.) I think it is only appropriate that Rowland shows up in a voting index at the top, and is 121 index rating seems accurate to me. ;) There are also several other people who are considerably better at voting than either Rowland or myself, and a simple percentage ratio won't really bring those people to light. Here is a new query based on this new formula, that uses the rest of Rowlands query, which shows some intriguing results. I think @whuber deserves some major props for having an index rank of 775, and @Grant Palin comes in a close second at 539. In the inverse, negative rankings are about the same as they were in Rowlands query.
Here is a full explanation of my query, the nuances of which have come out while discussing weighting and grading on a curve with @Lindes. There are some interesting points about the equation I've used that those of you who like statistics might find intriguing:
- This index ranks the relative performance of each member's voting.
- Differentiating between those who are voting well vs. those who are not is easy:
- Positive rankings indicate that you have given more votes than you get
- Negative rankings indicate that you have been getting more votes than you give
- The worst ranking is -100, kind of like absolute zero on the kelvin scale:
- If you are a new user with zero votes, you'll be ranked -100
- If you are a big time answerer with 10,000 rep but have not voted on anyone, you'll be ranked -100
- If you ask a ton of questions, but never vote or accept answers, you'll be ranked -100
- There is no cap to how high you can be ranked:
- If you have a rank of 100, your a good community member
- If you have a rank of 500 or more, you are a stellar voter, but you might want to answer more questions or ask some of your own
- Some common grades can be given to various ranks:
- F: -100 to -35
- D: -35 to 0
- C: 0 to 35
- B: 35 to 100
- A: 100+
- The query sorts from best voters to worst voters by default
- The goal is to provide an incentive to vote
- The antithesis would be to sort by worst voter in order to "punish" or "shame" those who don't vote
- I believe incentivising voting through positive reinforcement is a better approach