Data
Last updated: 2026-10-06
The SHALLOW ratings for 988 Survivor players (572 established) across 63 US and Australian seasons, the BEAST ratings for 874 players, and the PROBST forecasts are free to download as CSV and JSON. These are the files the site runs on, so they change with every update. They are licensed under CC BY-NC 4.0: use them for anything that is not commercial, with credit to Survivor ELO and a link to survivorelo.com.
| File | What it holds | Size |
|---|---|---|
| survivor_elo.csv | SHALLOW, one row per player. The columns are listed below. | 107 KB |
| survivor_elo.json | SHALLOW in full: every player with a season by season record of each rating change and the event behind it, and for each season the cast, tribes, episodes and the votes at every tribal council. | 16 MB |
| challenge_elo.json | BEAST: every player's challenge rating and record, and every challenge result by season. | 14 MB |
| predictions.json | PROBST: the forecast for the next episode of the airing season with the reasons behind each number, the season odds, voting blocs and the track record. Replaced after every episode. | 292 KB |
| forecasts/index.json | PROBST backtests: for US seasons 11 to 50, the forecast each week would have had from models trained only on earlier seasons. The index lists the seasons; each is at forecasts/{season}.json, for example forecasts/US20.json. | 1.4 MB in all |
The JSON files are large because they hold every rating change. For a list of players and ratings, the CSV is enough. The llms.txt file is a short text summary of the site for AI tools.
survivor_elo.csv has 1 row per player, 988 rows, sorted by SHALLOW rating.
| Column | Meaning |
|---|---|
rank | Rank by SHALLOW among every player in the file. Provisional players are ranked here; the site's rankings leave them out. |
castaway_id | The player's id in the survivoR data, the same in every file: US or AU, then a number. |
name | Full name with the nickname in quotes, for example Rob Mariano "Boston Rob". |
full_name | Full name. |
nickname | The name the show uses. |
elo | SHALLOW rating. Every player starts at 1500. |
z_score | Standard deviations from the mean rating of established players. |
tier | S, A, B, C or D, from the z score. |
provisional | True for a player with fewer than 5 tribal councils. |
tribals_attended | Tribal councils attended, across all seasons. |
latest_season | The season the player was in most recently, for example US50 or AU11. |
seasons_played | Number of seasons played. |
correct_votes | Votes for the player who went home, or with the side of a split vote that sent them home. |
incorrect_votes | Votes for anyone else, or votes that were nullified. |
vote_accuracy | Correct votes as a percentage of all votes cast. |
immunity_wins | Individual immunity wins. |
idol_plays | Idols the player played on themselves that saved them. |
outlasted | Players voted out at tribal councils this player survived. |
outlast_losses | Players who survived the tribal councils where this player was voted out. |
ftc_appearances | Final tribal councils reached. |
jury_votes | Jury votes received out of jury votes possible, as received/possible. Empty for a player who never reached the end. |
BEAST ratings are in challenge_elo.json, in the players list: elo, tier,
individual_wins, team_wins, individual_challenges and
team_challenges, with the same castaway_id. How each rating is calculated:
SHALLOW, BEAST and
PROBST.
The ratings and forecasts are licensed under CC BY-NC 4.0. You may copy, share and adapt them for noncommercial purposes if you credit Survivor ELO, link to survivorelo.com and say what you changed. Commercial use is not covered by the license. The forecasts are not for betting: PROBST knows only what has aired.
The votes, challenge results and boot order under the ratings come from the survivoR R package, which has its own license.
Survivor ELO, survivorelo.com, with the file or page and the date you downloaded it. For example: Survivor ELO, "survivor_elo.csv", survivorelo.com/data, downloaded 2026-10-06.