Classic Tease Of The Month - December 2011

[RamseyS]

For what it is worth. I really liked the CH Flux 5 tease. My reason was not the video but the last picture on the voting page. This woman whoever she is, may be the hottest woman on earth.
But that is just me!

It would be nice to be able to incorporate even more videos but it is much easier to mach up other's photos with written text than it is to do an analogous thing with video.

But I do appreciate the effort.

Also for what it is worth. The cactusman tease with one of my all time favorites, Sydney Moon, is outstanding as well, in what I consider to be a VERY VERY good month. Surely it is one of the longest, most involved and enthralling teases ever posted.

It would be nice if everyone could win at last something for even getting nominated to the Tease of the Month voting competition. I have no problems however with there being only one winner.

Finally, I also do hope that all of the webteases on this site are backed up at least 2 or 3 times. It would be terrible to have a crash. Knock on wood. It would not be the first time that a site goes down for unknown reasons. I surely hope I have not jinxed it!

Thanks to all.

[cactusman]

Also for what it is worth. The cactusman tease with one of my all time favorites, Sydney Moon, is outstanding as well, in what I consider to be a VERY VERY good month. Surely it is one of the longest, most involved and enthralling teases ever posted.

Thanks very much for that praise of my work, by which I think you're referring to my tease
Tonight an Orgasm Will Cost You... The Christmas Gift (Part 1). I'm very glad you liked it.

For what it's worth, for a few days I was happy to see that tease rise to be the most highly rated tease on this site (with a 4.6 rating). However, a few days ago, it plunged to a 4.4 rating. I was wondering if it came under the same sort of attack described earlier in this thread. It's disappointing if that is true, but I have no way of knowing... Usually if one of my teases makes it very high in the ratings, it gets beaten down again, perhaps because there are some here who don't like my style quite a bit, so I don't worry about it.

In any case, Thanks!

[Human]

So yeah. All that statistics can show is correlations, which do not imply causation, especially since all these other explanations haven't yet been filtered out. Maybe you should get off your "jump to conclusions" mat.

Since there was an attack in November, and I was the whistleblower, I'm entitled to a little jumping

And note that I used the word "consistent" when I used the statistics.

[Human]

This is where you fellow Milovanians come in. Can any of you come up with a better way to nominate teases for TOTM/FTOTM that doesn't involve software?

Wyatt

Have a tease nomination round.

5 teases are entered into the TOTM competition as usual.
3 additional teases get into the TOTM competition by Milovana member nomination; so finally the TOTM competition will have 8 teases.

To get these 3 teases do the following:
1. Create a thread indicating the 5 teases which are automatically getting in by the ratings process.
2. In that thread ask for nominations from members (thread stays open for one week). Only members with 25+ posts can nominate. Any numbers of nominations allowed per user. Tease creators cannot nominate their own teases.
3. At the end of the week, take the three teases with the most number of nominations.
If there is a tie, also include the tie candidates.
4. Create final TOTM thread with the 8 canditates (more in case of ties).

This way, the input of all the users (milovana members or not) gets counted throught the 5 automatic candidates. Also, if some exceptional teases have been attacked, or felt to have been attacked, the community can explicityly nominate them.

[Alliteration]

Since there was an attack in November, and I was the whistleblower, I'm entitled to a little jumping

You merely suspected an attack. You proved absolutely nothing - that was seraph0x:
http://www.milovana.com/forum/viewtopic ... ck#p105899

Furthermore, nothing entitled people to jump to conclusions. Unsupported assumptions are bad regardless. At best, such assumptions are reasons to investigate further, not reasons to conclude anything.

And note that I used the word "consistent" when I used the statistics.

You said two things that come to mind:

1) "But make no mistake CH flux 5 = under 4 rating = 100% attacked."
2) "Seraph confirmed that there was an attack on CH teases in Nov. Since the number of votes is higher in december, it would be consistent for there to have been an attack on CH in December."

Number 1 implies that you think one piece of data ("CH flux 5 has a rating under 4") is sufficient evidence for an attack. This is false.

Number 2 implies that you think one piece of data ("number of votes is higher in december") correlates with "CH flux 5 is being vote-attacked". It does, but it's a rather weak correlation, and also correlates with other conclusions, such as the ones I've mentioned previously. So, let's do this the RIGHT way:

First, a word on probability distributions. The probabilities of *all* possible explanations for a set of data are going to always add up to 1. So let's assign some prior probabilities (that is, what we think the probabilities are before we consider any evidence):

Code: Select all

People's tastes changed: .30There was a ratings attack: .20Error in the voting system coding: .10Bunch of people misclicked: .10Separate vendettas: .20Objectionable content: .10----Number of votes was higher in december: .80Number of votes was not higher in december: .20(we'll need these later)

I've assigned here what I think are some pretty reasonable prior probabilities for each of these things - if you disagree, feel free to plug in different numbers below and see what you get.



Here we have a simple form of Bayes' Theorem. First let's assign some variables:

Code: Select all

A: "There was a ratings attack on CH flux 5"B: "Number of votes was higher in december"

Plugging in numbers:

Code: Select all

P(A|B)= (.80 * .20) / .80 =P(A|B)= .16 / .80 = P(A|B)= .2

So, you should only believe, based on this single piece of data, that there was a ratings attack with 20% confidence. (Note: I've inserted .80 as the value for P(B) as well as P(B|A) because I can't see a ratings attack affecting the probability of the number of votes being higher, as the site is growing regardless.)

Now, let's see what happens if we look at another explanation, such as "A: peoples tastes changed"...

Code: Select all

P(A|B)= (.80 * .40) / .80P(A|B) = .32 / .80P(A|B)= .40

Oh look...it seems that if this is your only evidence, it's more likely that peoples tastes simply changed than that there was an attack!

Now let's look at what happens if we use our original A (the attack), and take another piece of evidence, such as "B: seraphox has confirmed IPs voting multiple times"...

Code: Select all

P(A|B)= (.95 * .20) / .30P(A|B)= .19 / .30P(A|B)= .63

Wow, that's MUCH higher! And that's exactly why I insist on this type of evidence. And once again...if you disagree with my original probability assignments, please explain why and feel free to insert your own. But don't argue with the math :D

[Alliteration]

For what it is worth. I really liked the CH Flux 5 tease. My reason was not the video but the last picture on the voting page. This woman whoever she is, may be the hottest woman on earth.
But that is just me!

Google image search says her name is Alice Goodwin.

[Alliteration]

@alliteration
            your comment lends more credence to creating a separate section.
If it becomes a "SOPA" or other problem then it can be cauterised very simply.

I'm not so sure, as once the intellectual property owners decide to take action, I don't think any "cauterization" will be possible...the damage will have already been done. As things stand now, however, the site can just use the "user created content" defense.

[Wyatt]

For what it's worth, for a few days I was happy to see that tease rise to be the most highly rated tease on this site (with a 4.6 rating). However, a few days ago, it plunged to a 4.4 rating. I was wondering if it came under the same sort of attack described earlier in this thread.

I noticed the anomaly on the day it happened. SeraphOx confirmed it was probably an attack. IP addresses from the UK.

I will continue to monitor the situation. We will never know for sure on CH 5 as there is just to much data for SeraphOx to sort through. I am trying to catch attacks within twelve hours. It makes it much easier to check out.

Wyatt

[cactusman]

For what it's worth, for a few days I was happy to see that tease rise to be the most highly rated tease on this site (with a 4.6 rating). However, a few days ago, it plunged to a 4.4 rating. I was wondering if it came under the same sort of attack described earlier in this thread.

I noticed the anomaly on the day it happened. SeraphOx confirmed it was probably an attack. IP addresses from the UK.

I will continue to monitor the situation. We will never know for sure on CH 5 as there is just to much data for SeraphOx to sort through. I am trying to catch attacks within twelve hours. It makes it much easier to check out.

Wyatt

Very interesting. That makes me feel better, knowing it might just be some jerk who down-rated it.

[Human]

Alliteration, I dont agree with your probability assignements at all.

1. People tastes dont change month to month. For individual people they do, but for an aggregate population no.
2. The other factors you mentioned, all legit, however, they too are relatively constant month to month. What we are trying to analyze is the change in CH ratings, not the absolute value of the ratings.
3. Dec votes being higher --- you should "legit votes higher" otherwise formulation does not make sense.
4. As I elaborated earlier, if Dec votes are to be considered legit, then we are considering the event that 50% higher votes in Dec to CH. I see no reason why this probability should be a super high 0.8. I would say 0.2 max. 50% more votes is a lot.

Pluggin in numbers from my assumptions, we would get a high CH attack number

[Alliteration]

1. People tastes dont change month to month. For individual people they do, but for an aggregate population no.

They most certainly do. People (in general) are always looking for the next thrill, the next "big thing". The same old stuff over and over tends to get boring. Furthermore, if you disagree with my prior probability distribution (remember, it all has to add up to 1) and think "People's tastes changed: .30" should be lower....where would you redistribute? And exactly how low would you set this?

2. The other factors you mentioned, all legit, however, they too are relatively constant month to month. What we are trying to analyze is the change in CH ratings, not the absolute value of the ratings.

Umm...what? We're trying to determine the probability that there was a ratings attack, given some piece of evidence. And if you agree to my probability assessments for everything else, you must agree with my probability assessment for "people's tastes changed" - for the Nth time, everything has to add up to 1. You simply cannot change the prior probability for one of them, and make no changes to any of the others. That's not how it works.

3. Dec votes being higher --- you should "legit votes higher" otherwise formulation does not make sense.

Fine, but the numbers for that are the same. Present another piece of evidence please.

4. As I elaborated earlier, if Dec votes are to be considered legit, then we are considering the event that 50% higher votes in Dec to CH. I see no reason why this probability should be a super high 0.8. I would say 0.2 max. 50% more votes is a lot.

I don't see anything so unusual about the number of votes being 50% higher - especially considering that there was a huge ruckus about cock hero and votes last month. This counts as advertising. Furthermore, when you look at the actual number of votes, it's not that many more compared to how many people use this site. I can quite easily envision a few dozen additional voters based on this advertising + regular site growth. Why can't you?

Furthermore....change it all you want, it doesn't matter as long as P(B) and P(B|A) are the same. To change the number, you have to somehow justify a disparity in the prior probability of higher vote numbers, and the probability of higher vote numbers given that there was in fact an attack. Knock yourself out.

P(A|B)= (.0000000001 * .20) / .0000000001 =
P(A|B)= .00000000002 / .0000000001 =
P(A|B)= .2

Pluggin in numbers from my assumptions, we would get a high CH attack number

Err, show your work. Always, always show your work.

[Human]

. And if you agree to my probability assessments for everything else, you must agree with my probability assessment for "people's tastes changed" - for the Nth time, everything has to add up to 1. You simply cannot change the prior probability for one of them, and make no changes to any of the others. That's not how it works.

Correct.
Who said I agreed with your probability assignments for the rest?
Probabilities must add to 1 if the events partition the sample space.
Consider an ideal static world. Probabilities of events 1-7 = 0.
Or an imperfect world where Prob of 2nd event = 1, rest = 0.

[les]

                         


                               Cock Hero Poll



                         

[Alliteration]

. And if you agree to my probability assessments for everything else, you must agree with my probability assessment for "people's tastes changed" - for the Nth time, everything has to add up to 1. You simply cannot change the prior probability for one of them, and make no changes to any of the others. That's not how it works.

Correct.
Who said I agreed with your probability assignments for the rest?
Probabilities must add to 1 if the events partition the sample space.
Consider an ideal static world. Probabilities of events 1-7 = 0.
Or an imperfect world where Prob of 2nd event = 1, rest = 0.

1) You said that: "The other factors you mentioned, all legit, however, they too are relatively constant month to month." If that's not what you meant, what the heck DID you mean?

2) This doesn't even begin to address everything I said. Try again.

3) What the heck are you talking about?

[Human]

1) You said that: "The other factors you mentioned, all legit, however, they too are relatively constant month to month." If that's not what you meant, what the heck DID you mean?

I meant we are interested in the change of CH ratings, and thus must look for the change in probabilites...