Sniping vs. Statistics

[Matrim Cauthon123]

Hi,

while reading this thread, and more specifically NSA\'s statistical reply, I came up with an idea...

NSA brought up the median in the thread above. I think the standard deviation (another statistical \'device\') might be a way to stop sniping. Well actually, it won\'t stop sniping, but it will definitely show it to everyone.

A little theory:
The standard deviation (SD) is a measure of the degree of dispersion of the data from the mean value. Stated another way, the standard deviation is simply the \"average\" or \"expected\" variation around an average.
More theory: standard deviation

So, in more understandable language: the bigger the standard deviation, the more \'extreme\' scores have been given.

Short example: suppose a superb mini would get 10 votes : 9,9,9,9,8,9,10,9,9,9, the average vote would be 9, and the SD would be 0,47 which is rather small. It evidently hasn\'t been sniped yet.
If someone decided to \'snipe\' this mini, and give 3 one-votes, then the average will lower to 7,15 but the SD would rise to 3,53 which is HUGE and an obvious proof of sniping.

An answer to sniping would be to disable (or ignore) votes which are to far from the average (which has already been said in other threads), thus automatically lowering SD. Sniping would then be more difficult. Not impossible, but it will require MUCH more time, and everybody would see that it has been sniped.

So what do you think? I think that it would be good to know that a mini has been sniped (not that my minis are anywhere near good enough to actually have been sniped).

I will now end this this mathematical rambling, but hey, I actually do this for a living (I teach math).

Thanks for listening, you can now stop shouting \"NERD!!!\" at your monitor

 

[Naukhel]

Statistic Analysis is a beautiful thing.
What about the guys that have their friends padding their votes with high votes on garbage minis?
What if you justify a low vote with a lengthy and detailed comment on a mini that is otherwise good in other people\'s eyes?

Math seldom accounts for the human factors.

 

[bayrodney]

MATHS aye...!!!

lol well written and a great idea, a lot of work though.

Although, i also agree with Naukhel on his two questions.

 

[Brimshack]

Actually discounting anything beyond the standard deviation would also take care of unnaturally high scores as well. And as to honest scores that fall that far afield, well, such is life. It isn\'t uncommon to arrange a judging contest so as to throw out the lowest and the highest scores. It doesn\'t mean condemnation of the authors of those scores, just a device to norm the results a bit. And of course comments would remain in place. If someone has the balls to explain a really low score, then they can certainly do so. Their comments remain, and have more force than a simple stat anyway.

And as a further option, one could of course report both stats. People can decide which one they care about for themselves.

 

[Avelorn]

Isn\'t that how that anti-sniping device work here on cmon? It takes care of those scores after a while.

However, if people snipe more smartly.. like 6\'s or 7\'s It can\'t distinguish those from honest votes.

 

[Jericho]

But a six or seven has like 25% the impact of a 2 or 3. 1\'s can destroy averages if you\'re above 8.5 or so.

Assuming the 1 doesn\'t get erased by some method, it\'s effectively taking away 8 points from a 9.0 model. So 8 more people than usual would have to give it a 10 to make up for that single 1.

Once you hit the upper cieling of scores, you can only really go down, and it\'s a million times easier to snipe than to bring an average back up.

That\'s the point.

Scores in the middle can go up or down so they\'ll balance out extremely easily.

So perhaps the real way to curb these BS votes once and for all is to eliminate anonymous voting, allow 1 vote per model only, and track how far away from the average every user rates models. If someone is deviating far from average on a regular basis then remove the account. IMO the single biggest contributor to the sniping trend is the anonymity of rating. Public forum + anonymity = jackassery as far as the internet is concerned.

Only problem of course is the increase of database size this would require... and my expertise in the field is about equivalent to let\'s say dental surgery?

I\'ve used metal files to smooth out my tooth when it was cutting my tongue, so I can clearly have a good idea once in a while, but generally I have no idea what I\'m doing lol

 

[QuietiManes]

I dont really get the whole sniping thing. I mean, voting is tracked and it\'s already 1 vote per mini, even anonymous votes are tracked, there\'s a cookie or your IP or something right? So, even if I hated UberPainterXXX and voted 1 a thousand times without logging in, it would only get recorded on the system as one vote of 1 right? Unless I blocked cookies or refreshed my IP every time.

Not that I really want to know how to screw the system but it just seems like a story people tell themselves to make them feel better about their low scores. Or am I missing something? (Just to note, I\'ve read countless \"my score was 9.whatever then a few votes later it was 6.soandso, I\'m being sniped! so gay!\" posts, they dont really prove anything, it could just be a string of bad votes, deserved even, who\'s to say every mini doesnt get its fair share of the crap vote)

Only problem of course is the increase of database size this would require.

It\'s all kept in the database already isnt it? What the members vote. So unless there\'s a whole slew of people registering just to vote it shouldnt be to much of a problem. I cant see many people wanting an account just to vote anyway.

Scores in the middle can go up or down so they\'ll balance out extremely easily.

A middle score would still need people to vote above average to balance after a snipe. A few less since the allowable deviation is less due to the range but it goes both ways. If sniping does exist I guess I\'m of the opinion it\'s not any less of a problem for the people with lower scores.

I remember reading posts about sniping and the \"system\" in place to help protect against it WAY back a few years ago when I fist found CMON. Noone knew the exact system but Borg or Chern or someone confirmed there was a system in place but wouldnt give the specifics because that would just help the snipers bypass it. The general consensus was that you had to vote within 2 or 3 points of the current average for your vote to count towards the average, otherwise it was not saved or contributed to the score at that time.

Um...yeah...so there\'s my long winded rambling post for today...lollollol

 

[Ogrebane]

You know I heard that 73% of all statistics were made up. Score is good I guess and I like it when I get a good score but what I like better is the feedback. People that snipe are worms. Much like consultants they cant do it so they critise. Score is nothing painting a mini you are pleased with is everything (unlees you are hoping the score will help your ebay price)

 

[Modderrhu]

I suggested standard deviation analysiis in another thread, and I have no doubt that it will almost completely resolve sniping issues. But having thought about it more, I doubt it will be feasible.

Depending on the strategy, each time a vote is made or queried, the analysis would have to be done. The server would not be able to display or calculate with pre-calculated votes like it is able to now. Standard deviation cannot use pre-determined values, but must look at each vote every time an analysis is done. So whenever a vote is made or displayed, the server has more querying to do. Moreover, standard deviation does not identify aberrant votes. That is easily done, but still requires an additional operation.

Standard deviation analysis will, I believe, consume too much processor and disc time. Chern Ann recently cached some common queries, and that has considerably sped up the CMON server. I, for one, don\'t want it slower again.

 

[Ebonbuddha]

Drink!:flip::bouncy::flip:

 

[Itchy]

Well, what if somebody starts their mini out with a score of 10, but it\'s really a 6 mini. That\'s a SD of 4! Obviously the logrythm would see that and knock out 6 vote... even though that\'s where the mini belongs. So whatever the mini starts at will be, more or less, what the mini ends at because you can\'t deviate too much or else the vote will be kicked. That\'s the problem i see.

Is there any way to see what all the votes are on a mini? Can you go to a mini with 100 votes and see that there are (just making up numbers here);

14 eights
40 sevens
32 sixes
3 fours
11 ones

And then say, gee... that looks a wee bit sniped. i think that\'d be the only real way to do it... is to go in manually, but that\'s NEVER going to work out.

PS. On a side note... i\'d kind of like to see what all the votes are and how many of each i get on a particular mini. Kind of interesting IMHO.

PS Again. i WISH i was good enough to get sniped

 

[Chern Ann]

I can guarantee you will never see such an abnormal distribution after there are 50 votes and the server does its nightly cleanup.

 

[philologus]

I just finished Statistical Analysis this semester and I have to say: \"Knock it off everybody!\"

 

[Chern Ann]

One idea that is relatively simple to implement is to force hide scores for miniatures with votes less than 50. Although that may take some of the fun out of the site.

 

[DrEvilmonki]

@itchy. I\'ll snipe you if it will make you feel more appreciated. lol

 

[EricJ]

Originally posted by Chern Ann
One idea that is relatively simple to implement is to force hide scores for miniatures with votes less than 50. Although that may take some of the fun out of the site.

lol, what might work better is an \"option to make your mini inelegible for top 10 in last 7\"

although also might take some of the fun out of the site

 

[Klute]

I would second that Eric.
The few times I have made THAT list the figures dropped like lead weights.

 

[frenchkid]

I finaly get some vacation time to browse on the forum and ..... more maths !!! Damn it people, you want my death !!

( On a side note the SD of my class on mth tests is around 4, talk about level disparity )

 

[Ritual]

I think the ACTUAL sniping of the minis in the T10L7 is quite moderate. The biggest impact on the scores come from your mini getting much more exposure. Before a mini becomes activated it gains a lot of votes from friends and people who are more or less aquainted with the painter through other forums etc. When the mini hits the T10L7 a lot of people will vote on it and most people DON\'T have any sense of loyalty towards the painter. Furthermore, people have very different views on what makes a great mini. If you don\'t like a particular style it will influence your voting. If a mini is rated 9+, a couple of 8s will hurt the score.

So, I think the exposure to a large audience with very differing views is a more important factor to why things are the way they are. I\'m not saying there is no sniping, though, but I think the lasting effect of that is quite small compared to the fact that not everyone share the same views on what makes a 10.

 

[vincegamer]

Originally posted by Ritual
Furthermore, people have very different views on what makes a great mini. If you don\'t like a particular style it will influence your voting. If a mini is rated 9+, a couple of 8s will hurt the score.

exactly. For instance, the best painter in the world could do a diorama involving all of the fianas on some ridiculously elaborate base and I would never give it a 10 because I think Fianas look like crap.

As to SD, I think someone or sometwo have pointed out: if you post a mini and get 50 friends to vote 10 on it, it has a 10 score. Then you get a couple other votes that are more realistically placed at, say 7 or 8, the regular scores would get booted. You would have to discount them, but hold them in memory so at 100 votes when the score is 50 10s and 50 7.5s, they would end up with even weight.

I don\'t know why I replied here. I don\'t think I\'ve ever been \"sniped\" (unless you count the folks who gave Left Talon of Roc a 10) and if I were, I doubt I\'d really care.