rankings on here are pointless, just provide a link to gamestats.

problem solved.

top 5 percentile is what i meant. so, say, for game X there's 100 votes, listed below (with the middle truncated:

0,1,1,1,6,6,7 ... [etc] ... 9,9,9,9,10,10,10,10,10,10,10 (total: 100)

5% bottom: 0,1,1,1,6 gone

5% top: 10,10,10,10,10 gone

left with:

6,7 ... [etc] ... 9,9,9,10,10 (90 left) and take the average of these 90.

======================

histogram:

as in a histogram chart. so in this case, the scores are discrete, counting the number of votes for each score would work. so, for instance (100 votes):

11 votes of 0

0 votes of 1

0 votes of 2

2 votes of 3

4 votes of 4

6 votes of 5

10 votes of 6

27 votes of 7

26 votes of 8

10 votes of 9

4 votes of 10

on the x axis we would have 0, 1, 2, ... 10 and on the y axis a bar that's proportional to the number of votes gathered for that score.

it's just a bar chart... histogram is i guess a fancy name for it. this visualization works very well! easy to read, easy to understand, and everybody can see exactly what's going on.

==================

rich text editor: works for me. firefox 2.0.4.

================

another word on top 5% of bottom 5%: the other way to eliminate outliers is to eliminate those that lie outside of some number of standard deviation from the mean.  but it's unlikely we'll have things close to a "normal" distribution (don't actually need the quotes, i do mean the normal distribution, i.e. the Gaussian distribution).  with nintendo fans and sony fans, we could easily have some bimodal distribution.  so a percentile cutoff would be A) more robust and B) much simpler to interpret.

I think this is a cool feature in theory, but it's just not working out in practice. The game rankings system has sadly degenerated into rampant fanboyism, making it impossible to draw any kind of reasonable information from it. For the past week, all of the top games were Nintendo ones (and pretty much exclusively Mario/Zelda), a poor representation of historic top titles. Now today all of the top titles all Playstation ones, including some laughers. Ninja Gaiden Sigma hasn't even been released yet, and it's in the top 5. Fanboys are handing out nothing but 1s and 10s left and right, deliberately trying to manipulate the system to support games on their system of choice. It's a broken system.

ioi, I don't want to tell you how to run your website, but I think this needs to be addressed in some way.

I think kingdom hearts should top out at #1, RATE KINGDOMHEARTS 10/10, and this game was truly epic kH2 has no boxart? wondder why

He's pretty right I agree with him, sony fans are just rating nintendo games 1's it iisnt right, but nija gaiden ddi come out in the ps store and was a good game even though it wa only 1 level. Resistance is a good game but not that good to be #3 38 or #9 I  would understand, and kingdom hearts isnt even in the top 10 thats just rediculouse.

I think they rating system needs to be reset and limited to one vote per account (or IP), its getting very abused right now.

Croc 2 (PS) 9.50 (10 votes)
Bob The Builder (PS) 9.10 (10 votes)
American Idol (PS2) 9.58 (12 votes)

It was working pretty well up until today. Now its like someone is going around hitting all the Nintendo and Xbox games with 1s and all the PSOne, PS2 and PS3 games with 10s

 yeah... with more sophisticated methodology it can be smoothed out somewhat.  say, if you want to automatically categorize "nintendo fanboy" and "sony fanboy" and you can reduce how much their vote counts.  so for a start, you can use a clustering algorithm (which is comparatively easy) to determine if a certain voter belongs to a certain voting pattern.  you can then run statistically tests to test out your hypothesis.  this voter then would have the weight of his votes reduced by some amount.

there're lots of methods to figure out "fanboys", but truthfully, none are "automatic".  you'll need to run tests and actually study the results and adjust parameters or alrogithms, which is definitely much, much, much more work than it is worth, with no guarantee that results would be much better.  what's worse is that people can accuse the person who implemented such a system of favoritism.

so, when machines fail, humans step in.  as vgcharts is a community, not some dead immutable numbers some statistician gather, one can enforce fair and unbiased rules to improve the quality of votes.  that to me is the most promising route in eliminating extreme votes.

since there ARE more nintendo fans than any other, those games will be rated higher.  since we want to leave out statistical methods (which could draw accusation of favoritism), this is just a systematic issue that would be very hard to address.  

instead of giving SCORES, voters can be asked to RANK the games they want to play.  i don't really like this but it is an alternative way...

ioi said: Yeah I know what a histogram is, just wasn't clear how you intended to use it I still think that dropping the top and bottom 5% doesn't really work all that well. If I like a game enough to give it a 10 then I want my vote to count, same if I give it 1 - I think the system falls down (as does trying to find a decent normal distribution). Everything should even itself out on its own anyway - the scores are all relative. If we get rid of the top and bottom 5% then games with a score over 5 will go up a little and games with a score below will go down, but the poisitions won't really change.

 hehe, i kind of thought that :)

 yeah, in the long run, hopefully the extreme votes get drowned out.  whether top and bottom 5% elimination works better, it really needs to be tested to know for sure.  but it might not be worth the time to test it.

 again, the easiest way is to ENFORCE a set of strict voting rules.  and "full disclosure", or "partial" disclosure by using a histogram.  which i would absolutely LOVE to see!!!!!

The fact that LoZ:OoT has been shifted from 8.6 to 7.6 in one day when it was at several hundred votes shows that the system doesnt work.

Basically I think you should be able to vote once on each game and ONLY if your logged in to stop people with dynamic IPs vote spamming. Currently its a contest of who can click the most, its just being manipulated.

I reckon a reboot and a one vote per account system, there will be a smaller sample but it will be more representative of what people on this site in general actually think.

Your vote would still count. The median is a damn decent measure of central tendency for which only a couple of votes "count", the way you put it... Other measures of central tendency are far better than the arithmetic mean for skewed distributions. Be it the median, the trimean/trimedian (name varies), or the method he proposed (which is used at the Olympics). All are significantly better than the mean, which should really only be used for symmetric distributions.

Yes things may even out eventually, but that may take thousands of honest votes. For every fanboyish vote of 1 you have to "overcome" in Ocarina, to get it to the 9+ most reviewers seem to think it deserves, you need 10+ votes of 10. Ask yourself if honest users to this site clearly outclass fanboys by more than 10:1. And don't forget to factor that hate voters are allowed to double vote, while honest voters just won't (or they aren't honest at all).

Because in all honesty, what happened to Ocarina today when it already had a few hundred votes to begin with, tells us you're plain wrong, and the current system is a joke, really. It will likely take a few thousand votes to "correct" that, and this site is far from being that popular.