Yuku free message boards

toomanynachos16 wrote:

---

But all this talk about how Chase being an idiot was qualified....by whom? By Brenda, Nay, Marty, and Benry? They're ALL villains! I just don't get Edgic's analysis of Chase this episode AT ALL. It's the exact opposite of what America is supposed to think, and you all just think he was proven to be a dumbass because you think he's a dumbass!

It's not so much by a specific player but more by the editing itself. The edited sequence of events clearly hammer home the "Chase is paranoid and annoying to his alliance" point. It is not subtle in the least that the editors want you to know and believe that Chase is paranoid and annoying to his alliance. It is not subtle that they want you to believe this is a mistake he is making in terms of his relationship with this individuals.

I agree that negative SPV from villains is less impactful than negative SPV from neutral or P individuals. However, I would argue that carefully edited negative SPV from the editors themselves trumps any sort of negative SPV from a character in the story. The editors have a sort of omniscient role in the storytelling. They are never portrayed to be wrong. If the editors tell you a character is lazy, the story does not later reveal that they are actually hardworking (regardless of their true behavior).

In this case, the editors are clearly telling us that Chase is isolating himself from his current alliance. Of course it IS possible that this is being done to show why he eventually turns on them, why he ends up beating them, etc. It's possible. However, my take on this (and the take many others seem to be taking) is that there are ways to do this without highlighting his paranoia and annoying nature in a bright pink highlighter. If he is our eventual hero that takes down the N characters, why reinforce the criticisms that these negative characters are making?

Obviously there is more to his edit than this one heavy-handed sequence of events but I really don't feel like the editors conveyed as strong of a message with his other behaviors (like seeing through Marty). I also don't remember him elaborating on this - did he tell us anything about what he thought of Marty's plan himself? We can infer his thoughts based on his actions but I don't remember him saying something like "Marty is trying to get Nay to use her idol by spreading a bunch of rumors. I think he's just blowing smoke though and I'm not buying it." This is a great strategic moment for him, yet the editors portrayed very subtly. Perhaps that's them slow playing Chase for the future. Given what they chose to highlight, however, I am skeptical of that.

CrystalOwnage wrote:

---

I doubt any "formula" will be able to predict anything other than the boring CP/MOR winner just like current Edgic.

I think it's important to point out that statistics of this sort (predicting human behavior - the behavior of editors in this case) do not necessarily aim to explain 100% of the variability in such behaviors. To do so is an unrealistic goal. In psychology, for instance, many empirically supported relationships among variables do not hold 100% of the time for 100% of individuals. It is more limited than that, yet still useful. For example, we know that certain types of parenting behaviors are associated with higher incidences of problems for children. Some children do not experience these problems despite the more general trend. Nonetheless, we can still say something about which types of parenting are probably a bad idea, given that you don't want to expose your child to the risk of following the most common trend. Same thing with something like smoking and lung cancer - not everyone that smokes gets lung cancer but a significant proportion do. Knowing this, most people choose not to smoke because chances are good that they will fit the trend. I see Edgic in a very similar way. I don't think any sort of statistical analysis will allow us to arrive at the winner with 100% accuracy. However, I believe it can get us close enough that we are more likely to choose the correct winner than not.

Say we develop a formula that assigns each person a score from 0 to 100. In every single season (S1 through S20), the winner has the highest predicted score on this formula out of everyone from his/her season. Additionally, every winner has a score of over 60 and no one who has lost the game has a score over 50 for their season.

Say this season (hypothetically) scores like this:

Chase = 80
Brenda = 45
Holly = 30
Sash = 42
Jane = 48
Jud = 32
NaOnka = 5
Benry = 15
Kelly = 0
Dan = 2

Who should we be looking at as our MOST LIKELY winner? Based on the data, the answer is obvious. If Chase were to win, it would conform to the data gathered up to this point.

"But what if Chase doesn't win?"

This is the beauty of such an approach. The formula itself should be driven by the guidelines (theory) of Edgic and nothing else. Given the above formula, if Chase does not win, the guidelines are either A) being misrepresented in the data somehow or B) INCORRECT and in need of revision. This approach formalizes the guidelines of Edgic and makes every season a direct test of the hypothesized "Edgic Theory." There is no "well the guidelines are probably correct." They either gave you the best chance to get it right or they didn't.

Using a formula also gives you an estimate of just how much the guidelines are explaining. If, say, 60% of the variability in "winner vs. no winner" is explained by your formula, this tells you that 40% of what predicts the winner is error variance. That is, factors OUTSIDE of the Edgic framework that predict the winner. If, as many naysayers would suggest, Edgic "predicts nothing," then the error variance should be huge and the amount of variance explained by the guidelines should be trivial. If Edgic explains 80% of the variance with its guidelines, it is very hard to argue that it does nothing to help in predicting the winner. Additionally, having an estimate of the error variance also tells you how much more we could learn by adding guidelines based on patterns in the data. If we explain 40% of the variability with the current guidelines, it encourages use to search for additional systematic variables that are out there. What ELSE can we use to predict the winner? You can even test for whether or not a new proposed variable explains something above and beyond the explanatory power of all previous variables - a direct test for new guidelines.

On a more practical side, you could also derive separate models for different types of winners or different genders. Winner predictions could take the form of "based on the traditional Edgic model, X is the best candidate, but based on the Danni/Bob/Natalie type model, Y is our most likely winner." Or even "W is the best winner according to the male winner model, while Z is the best winner candidate if the winner is female, and V is the best using the combined model." There are so many ways you could go with this and I think that almost all of them are of some value to us.

In any case, that's my argument for why statistics are valuable to Edgic. The goal is not to attempt to characterize the editors as robots, whose behaviors are entirely predictable based on some algorithm. Rather, the goal is to detect regularities in their behavior and use those regularities to give our best guess at a future outcome. In a way, this is what Edgic purports to do in the first place - this approach would simply try to objectify those efforts and use statistics to provide tests of their effectiveness.