Yuku free message boards

getting real wrote:

---

RC said:

Upolu has been presented as a united group with interesting interpersonal dynamics to be resolved. The reason? Because those interpersonal dynamics will be crucial to the endgame - just like the Edgic model predicts.

The only Upolu dynamic that you need concern yourself with is that between Coach and Brandon.  Even the looming battle between Coach and Albert/Sophie will really be about the Coach/Brandon falling-out.   The key scene as far as Upolu is concerned is in Episode 4 which begins when Brandon tells Coach that Albert, Sophie and Mikayla have an alliance.

• Coach angrily tells him to "Stop it!" ;
• Brandon semi-dismissively asks him how many times he's been screwed over the other 2 times he's played;
• Coach tells him he no longer has his head in the sand.
• The scene ends with the following Coach confessional:

"Brandon is a great guy... and I want to be with Brandon every step of the way... but I'm worried that he's going to kick over the kettle at the most inopportune time (omnous percussion music)... and ruin the game for everybody ... and that's why I have to protect myself ... I have to remember that I'm out here for the third and possibly final time .. and this has gotta be my game... and I can't let anybody or anything mess it up.

Those are great sentiments to be expressing in a confessional, except the foundation they're based on is shown to be demonstrably wrong.  Brandon was right and an A/S/M alliance did in fact exist.  By denying it with such venom Coach was directly contradicted (the bane of his edit) and shown to still have his head in the sand.

Later that episode, at TC, Brandon says he wants to be somebody God's proud of , that he's going to reclaim the Hantz family name and that "it's not how you start it's how you finish."


This is the most important pre-merge scene for Upolu along with the Brandon v Coach black & white v. grey moral conflict later on. 

In combination, these scenes leave me in no doubt that the conflict between Coach and Brandon is crucial to Upolu and that, more than likely, Brandon will be the cause of Coach's downfall.

When I first mentioned this, you demanded to know which editor had told me.  Well, in a sense, they all have.  They've told everyone.  It's up to everyone to see it.  If I'm wrong about this, feel free to re-visit and exact the maximum possible retribution.

The issue here is that I just don't choose to play the odds the same way you do.

As we've all established, the sorts of hidden clues, hints, foreshadowing etc. most often referred to here are subject to the "played straight/ironic" coin flip. This has been covered dozens of times so I don't feel the need to reiterate it in full. One issue with this I've touched on in the past (but have only slightly detailed) is that the more of these you accumulate, the riskier your position actually appears to me. In the absence of some sort of clairvoyance (or insider information), there is absolutely no way to know what the significance of ANY Survivor scene is a priori - only in retrospect.

In isolation, then, the chance of one particular interpretation being correct is roughly 50/50 (unless there is some evidence to be shown that one interpretation is more frequent historically than another or that someone has special prediction skills). It could be right (Rob's voting confessional predicts his win) or wrong (Royal Treatment predicts Rob's loss). It's just one's opinion that determines which way to bet. Of course, few people actually make their bets on a single scene alone!

Instead, we get a series of multiple scenes that are interpreted in a coherent way. As we have seen many times this season, this is very possible and even popular. It seems quite convincing when you see a list of several hints all framed as pointing the same way. However, this is really just a series of connected scenes on which the writer has bet that the split will come down on the same side. If you list 15 key scenes for Cochran and suspect that the correct 50/50 interpretation is the same for all 15 (i.e., hints to Cochran win or hints to Cochran loss), the picture looks either quite rosy or quite damning. The size of the list, however impressive, actually has little connection to the strength of the case itself, as I see it.

Increasing the number of interconnected 50/50 hints does not increase your odds of correct identification. When presented, hint A supports hint K which supports hint P which supports... on and on. For example:
Hint A: Ozzy said he wants a Savaii to win (50/50 odds of success).
Hint K: Ozzy asks how to write Cochran's name, which foreshadows his FTC vote for Cochran (50/50 odds in isolation) - this is supported by Ozzy's desire for a Savaii win (again, 50/50 odds in isolation). In other words, given that Hint A is true (which we don't know), the chances of Hint K pointing to a Cochran win are high.

The problem here is that the connections between hints are themselves tenuous. It's true that Hint A being interpreted correctly DOES increase the odds of your Hint K interpretation succeeding. However, if Hint A proves to be false (ironic), the odds of Hint K don't look so good anymore. In fact, they probably get worse than a baseline 50/50!

I don't want to work through the details of a whole Cochran-supported argument (my conditional probability is rusty so someone please correct me if I'm wrong on this basic assertion) but the above, I think, demonstrates the issue. If your argument requires a large series of connected 50/50 bets to all come down the same way, it's not likely to be much better than betting based on one scene alone. People have detailed a large web of interconnected hints for why Coach loses, many of which involve Coach/Brandon ending badly for Coach. What if that bet is wrong? What if Brandon is doubt for Coach and not Coach's foil? The whole house of cards crashes. When the Brandon > Coach bet fails, the other bets one has made that rest on that assumption are also less likely to succeed. When the Brandon > Coach bet succeeds, the house can weather pretty much anything. You could have this exact same point made for every linchpin in the entire case (Stacey's boot was bad for Coach, every hypocrisy was highlighted prominently as a clue, the prayer scene was a hint Coach loses, etc.). You just need to hope like hell that none of your key bets fail you.

Unconnected hints don't do you much better. If your pick is based on 10 separate 50/50 hints, none of which are supported by the others and all of which are crucial to your case, you're actually very unlikely to have all of them go your way. The more you have, the less likely you are to bet correctly on all of them.

Essentially I see it as an extremely high-risk high-reward way to play Edgic. If your key predictions ring true, you have an extremely high chance of success, as the effect will make your connected predictions more likely to be true. If your key predictions fail you, you have an extremely high chance of failure, as the effect will make your connected predictions look more unlikely to occur.

I prefer not to play this way. If you do, then great. If you end up making the right bet I'll be glad for you and work to refine my own methods (i.e., the Edgic model, which should theoretically stack the odds in my favor). That's the good thing with the Edgic model, by the way - the more data we get, the better we can presumably stack the odds (given that people actually use it). Alternatively, the odds remain roughly the same for every season prediction hinged primarily on a series of hints and clues - if you guess right on the key ones you win, if you guess wrong on the key ones you lose. This is why I will not simply abandon an Edgic-supported Coach argument due to these points, however well-argued (and potentially correct) they may be. Unless you start detecting patterns in these hints/clues to increase your odds of correct interpretation (i.e., a type of Edgic analysis), the number of seasons you see has nothing to do with success.

If you're wrong, the plan is not "retribution." More just to continue to support the strategy which I believe gives us the best odds of victory with the lowest level of risk. Win or lose, the Edgic model suggests that Coach is a low-risk pick. Given that the story also has the possibility to be supporting him, rather than demonizing him (especially in the very important episode recaps), you can't make a much better gamble right now, IMO. That's all I mean, really.

getting real wrote:

---

(i.e., the Edgic model, which should theoretically stack the odds in my favor).

Of course, there is no one universally accepted statistical model,  a problem which results in different practitioners arriving at different conclusions.

The probability problem which you identify for those who seek to find story-telling patterns is replicated in your approach in choosing between statistical models.  If everyone correctly identified, say, Phoenix's MORD approach as appropriate this season they should all arrive at the same conclusions as Phoenix did above; however, choosing between this model and, say, Types A-D model seems to me to be as subjective as identifying a pattern based on story analysis.  I say should because diffferent interpretations are sometimes possible even when one settles on a particular model.

This is a question that can, and ideally should, be addressed empirically. However, the panic-meter in Edgic hits the ceiling every time something even remotely hinting at this approaches.

Different statistical techniques are not quite so interchangeable. The full complement of statistical patterns at our disposal is not so subjectively different than "pick whoever has a CPP in episode 1." There are clear objective differences in what the models can explain and what they can't. The fact that no one has actually sat down and articulated which aspects of the patterns are most predictive is probably based on A) a lack of time taken to do so and B) a lack of motivation (people here are just going to dismiss it as "opinion" or "flawed" no matter what mountain of empirical evidence backs it up).

Ultimately, it seems to me that the level of success to be had with either approach depends on the right combination of inspiration and perspiration their practitioners bring to their work.

I should also repeat here my fascination at the fact that, this season at least, there seems to be greater agreement between those who employ the story/theme approach than those who use the statistical approach.  When I first mentioned this you brushed it off as some kind of - correct me if I'm misrepresenting you - sub-conscious desire for consensus.    Given the people involved, however, I think this to be very unlikely.    I am extremely intrigued by this as, on the surface of it, one explanation may be that there may be aspects of the editing that, given their independent confirmation, might not correspond with the 50% probability that you have nominated.

What I opposed was this idea that consensus is a good gauge of "correctness." I don't care if you find 5 billion people who claim 2 + 2 is 6. It has nothing to do with whether or not 6 is the correct answer to such a question. Likewise, I think the number of story people backing Cochran has nothing to do with whether or not their interpretation is correct. What I brushed off was the idea that consensus = correct. Edgic had a near consensus that Marcus won Gabon - that didn't make it any more right, did it?

Anyway, as you confim, you use story to confirm your stattistical conclusions.  We all work with some combination of the two but, unfortunately, you seem to be frustrated by differences in emphasis.  Let's just enjoy the ride.

I'm just explaining why I don't consider a wall of clues to be sufficient evidence against Coach. You argued that Upolu (Coach specifically) was doomed because of *list of clues.* I told you why I don't consider that evidence strong enough to alter my pick. If you think I'm frustrated, that's an interpretation you've misread! ;)

Cuauhtemoc Cosby wrote:

---

Man I really hope Cochran wins so RedCoffee can stuff it a bit.

Remember everyone: I'M the guy with an attitude problem that's rude and vindictive. Me.

The rest of the post seems succinctly summed up as:

- There are ZERO patterns in Edgic: "How many seasons of statistical data do you need to realize every season is different and there IS no pattern?"
- Thus, every season is different and this search for statistical patterns is a fruitless waste of time.
- HOWEVER! When it comes to storytelling techniques? Oh yeah we can DEFINITELY use the editing of past seasons there! Now those consistencies from season to always-edited-differently season are things that are crucial! "This kind of analysis that panurge does and VerucaSalt did was ABSOLUTELY required on the knowledge of how the editors used past seasons."

Please forgive me if I do not think it is worth my time to refute every point your contradictory rant of an "argument." I'm the giant idiot for applying the widely accepted techniques of behavior science to a TV show's footage. You're the genius who continues to misconstrue those techniques and goals in every ignorance-broadcasting post with the now-emerging argument that past patterns are useless when I use them and helpful when you use them. I see.

I don't know if there's a big enough for that one!