Can Jury Favorability Predict the Winner of Survivor?

I bang on a fair bit about jury favorability and the number of votes with jury members. This is because the number of votes with someone else is a good indicator of the strength of their alliance, and if one is a finalist and the other is on the jury, there is a good chance they will get their vote.

tl;dr

Of the 220 jurors, 115 voted for the finalist they shared the most votes with. We expect only 84. The \chi^2 test statistic is \chi^2= 11.4 and P-val = 0.0007. This is approximately 37% above random chance.

There is an association between the number of votes shared with the finalist and voting for them. It’s a good predictor but far from deterministic.

Here’s a short analysis.

Analysis

I’ll use the castaways table to find the jury members and finalists, and the vote_history table to calculate how many times players have voted with each other.

The number of votes a jury member has shared with a finalist is simply the number of times they have voted for the same person at the same Tribal.

I’ve summarised the data to see if the jury members voted for the finalists they shared the most votes with. This means the finalist will be flagged as sharing the most votes if, for example, they shared 7 votes and the other two only 6 votes, or they shared 7 votes and the others 1 vote each, the finalist with 7 votes is flagged as the most. The difference isn’t taken into account, but you can see how this is important.

Typically jury members are more likely to vote for the finalists they shared the most votes with. It’s most easily seen in the percentage chart. The more votes with the finalist, the more likely they are to vote for them in the final Tribal.

If a jury member has voted with each finalists equally, they are guaranteed to vote for the finalist they shared the most votes with. So, I’m going to filter the dataset to those cases where there is only one finalist they’ve shared the most votes with. I’ll then test if they are more likely to vote for that finalist.

Out of the 388 jury members and votes cast, 220 are cases where there is only one of the finalists they share the most votes with.

Of the 220 jurors, 115 voted for the finalist they shared the most votes with. We expect only 84. The \chi^2 test statistic is \chi^2= 11.4 and P-val = 0.0007. This is approximately 37% above random chance. That percentage increases as the number of votes with the jury increases.

I think it’s safe to say there is an association between the number of votes shared with the finalist and voting for them.

Final thoughts

During the season I look at the vote stats frequently to see who has positioned themselves well, has the strongest alliance, and who is the favorite to win. Often the data shows things that aren’t particularly obvious in the edit. The chart below is an example of how I view the votes shared between finalists and jurors.

As for predicting the winner, it’s far from deterministic but it is useful to know where votes are likely to land.

Other bits

Links:

All code to run the analysis is below.

Functions

Code

df_jury <- survivoR::castaways |>
  filter(jury) |>
  distinct(version_season, voted_with_id = castaway_id)

df_finalists <- survivoR::castaways |>
  filter(finalist) |>
  distinct(version_season, castaway_id)

df <- map_dfr(all_vs("US"), ~voted_with(.x)) |> # functions in append
  semi_join(df_jury, by = c("version_season", "voted_with_id")) |>
  semi_join(df_finalists, by = c("version_season", "castaway_id")) |>
  left_join(
    survivoR::jury_votes |>
      distinct(voted_with_id = castaway_id, castaway_id = finalist_id, vote),
    by = c("voted_with_id", "castaway_id")
  )

df |>
  group_by(version_season, jury = voted_with) |>
  filter(sum(max) == 1) |>
  group_by(version_season, jury) |>
  mutate(p = sum(max)/n()) |>
  ungroup() |>
  filter(vote == 1) |>
  summarise(
    n = n(),
    obs = sum(max),
    exp = sum(p)
  ) |>
  mutate(
    p = obs/n,
    p_exp = exp/n,
    chisq = (obs-exp)^2/exp,
    p_val = 1-pchisq(chisq, 1),
    index = p/p_exp
  )
Follow me on social media: