A Model for Winning Survivor | Hacker News Reader

A Model for Winning Survivor

victoriaritvo.com

53 pointsevakhoury28 comments

> I loved using it while watching Season 50, and it identified several interesting threads early on.

The ideas in TFA are interesting, but it’s important to keep in mind that when we watch Survivor, we’re not watching something that happened in real time and then was simply compiled together. Survivor is edited after the fact, once the winners and all the arcs of contestants are known.

And don’t get me wrong! I love watching the show even knowing that in my mind. But a model built on the edited material is necessarily biased due to the format of the show itself.

The first and third most correlated stats are completely out of the player’s control. The confessional stat is reasonable if you’re trying to guess the winner, but as a player, there’s no obvious way to control it.

Interesting to see Russel as the biggest upset. He is probably my favorite and completely changed some aspects of the game (idol hunting without a clue or game information about an idol even being present). He burned everyone, but his strategic game was flawless while his methods were still novel. Once they were known, however, his game collapsed.

Russell is famous for playing a different game than survivor, and then getting mad that he wasn't rewarded for it. His game didn't really collapse, because he was never going to win.

The stats say otherwise. Of the stats that are most correlated, with winners he was the biggest upset.

He was a rich asshole who personally wronged everyone on the jury in his first season, and carried by luck/ignorance/Parvati Shallow in his second season. Which stats capture that?

>Survivor is edited after the fact, once the winners and all the arcs of contestants are known.

That might make it easier for the model to pick up on clues from the editing process. Like winners get more on-screen confessionals or something like that.

It’s been a very long time since I watched Survivor, but in the early seasons at least it was obvious that the producers were also manipulating the game as it went along to increase the drama.

Tribe shakeups, double/no immunity, second chances… usually just when a particular alliance seemed set on a strategy that was going to make the next few tribal councils predictable.

Interesting, but it misses a lot in not factoring social dynamics (which the author admits). IMO the greatest Survivor player of all time was JT Thomas in Season 18. The guy was a down-to-earth cattle rancher from Alabama, and he was so good because his personality did all the work. He had that rare perfect combination of traits that you see in true born leaders, and was so loved by the group that no one ever cast a vote against him, not even once. It’s almost like they subconsciously let him win because they believed he deserved it more than they did.

The model in reality is “Hope the producers like you enough to write you into the show as the winner.” Survivor is hyperreality, not an actual competition.

doesn't mean that can't be predicted by visible side-channels

Not everything is in the hand of the producers though, and especially not the final vote.

> By the finale, Kalshi’s “Survivor” market had reached a volume of $32.7 million

I don't get this. How are people "betting" on something that is "known information" for other people? What's the point? I get betting on stuff that no one can know, but who's taking "the other side" on these bets? Why would you put money into something knowing that there are people who already have the absolute answer?

> Why would you put money into something knowing that there are people who already have the absolute answer?

https://youtu.be/Hm3mDatFpNE?is=vKkBJs6XsypuD833

That's exactly what "prediction markets" sell themselves as. Super-predictors.

Insiders have an edge and can in theory shape the overall distribution of bets such that the real world outcomes are priced in better.

Yes but in this case, tens or hundreds of people know the outcome months in advance. I'm very surprised that the market doesn't immediately converge to 100% for one candidate. The winner is not really a forecastable 'random event' in that by the time the show is airing, it's not being pulled from a distribution.

In theory, an insider willing to take any positive-expectation bet would drive the market price arbitrary close to the correct answer, but in practice they might be under NDA and want to avoid getting caught, or they limit their bet amount for other reasons, so the market ends up dominated by gamblers placing uninformed bets.

Indeed, I'm just surprised.

Maybe the people who could take advantage of these markets are just not sufficiently aware of their existence.

I predict that in 5 years "who wins Survivor" will be completely dead though, because the markets will have become common knowledge and there'll be enough people willing to violate their NDAs/whatever for the money.

You say in practice, NDAs. What we've seen so far: classified military ops, people with access to Presidential speeches...

In none of those cases did the market converge as a result of the insider betting, it was only discovered with hindsight that someone had placed a suspiciously well-timed large bet. So that's where "they limit their bet amount for other reasons" comes in. I guess it's difficult to get a large loan for "I know a secret that will allow me to double my money in a few days, honest!" so they have to make do with their savings instead.

"There are three ways to learn. The most noble is by thinking and getting to the truth yourself. The second, which is still pretty good, is by imitating those who do. And the third one is through pain."

In prediction markets, imitation is the best policy

Love this!

I played with doing something similar, where my features were based on Edgic charts I found online[1]. This ends up implicitly factoring in people’s _perceptions_ of the players/editing, which can be useful. Ultimately it was just a bit of a toy still, somewhat better than random.

Like other people mention, Survivor is edited with a lot of intention, I think if you really want to predict winners you need to somehow capture cues from the editors.

I wonder if you could derive per-player 1st and 2nd person sentiment from transcripts of the episodes. Could be interesting to factor that in.

[1]: Example - https://www.reddit.com/r/Edgic/s/2v0nA93odi

>"What that means: if you pick a player at random, you’ll get the winner right about 10% of the time (this varies with how many people are left, but 10% is an average). My model’s #1 pick is right just over 20% of the time. That 2x improvement holds for most of the game, basically after two episodes have elapsed (once the features have enough data)."

Ok... but no one picks a winner at random. How does this do against anyone who has watched the show and picks someone based on their knowledge of the game? Seems like a really complicated way of just barely doing better than a dartboard.

I haven't followed these sort of programs, but aren't they promoting and celebrating psychopathic manipulative behaviour.

In real history, often what's at stake is survival of a community and people want to genuinely work together. Here for some reason it's an arbitrary forced conflict if only one individual can win.

It's a game. Everyone is aware it's a game and uses language like who was a good and bad player, was that a good or bad play, etc. Everyone understands that deception and betrayal are possibilities.

Your objection is like saying playing a game of Monopoly is likely to turn people into economic rent seeking slumlords.

Sometimes things are just what they are and nothing larger.

This is an actual rift in game play that watchers debate over: the modern "manipulate everyone" game play vs. the old way of "cooperate and be decent". You can infer from my statement in which direction the game is heading. Even in recent seasons the non manipulative players reach the final three, or even win, but it's less and less likely.

Well I didn't want apoilers so I stopped reading. If the author is here, please remove the navigation floater from the side. The article is so narrow. I had to zoom all the way in on my tablet to get one graph to be full sized. Impresaiveely it wasn't pixelated. I want to see your graphs better.

I did not have this problem using reader mode

It looks to just be a problem on my tablet. Phone is narrow so it adjusted for that.

Reply on news.ycombinator.com