← All posts

Survivor ELO Blog

BEAST and the New Challenge Data

Published: 2026-11-14

survivoR adds full finishing orders to 169 older challenges, and 469 past players get a new BEAST rating; PROBST's log loss on 583 past boots stays at 2.273

TL;DR: This week's survivoR update adds the full finishing order to 169 challenges from Survivor 1 to Survivor 48 that had recorded only the winner. BEAST scores a full finishing order as a matchup between every pair of players, so a last place finish in a field of 12 now counts as 11 losses instead of 1. Of the 853 past players BEAST rates, 469 have a new rating, 93 by more than 50 points. PROBST's forecasts barely change: across 583 boots in 40 past seasons its log loss is 2.273 with either version of the data.


What Changed

survivoR records every challenge result for every player, including the order of finish. Until this week, 20 of the 635 challenges BEAST scores as individual had a full order, all from Survivor 49, 50 and 51 and the 12th season of Australian Survivor. The update fills in 169 more, with 1,424 player rows, in 47 of the first 48 seasons; Survivor 17 is the only one without a new order. Of the 169, 160 are individual challenges and 9 start in teams and end individually.

The new orders are consistent with the results survivoR already had. In all 150 of the new challenges with a single winner, the winner is in 1st place; the other 19 record more than 1 winner.

The same update corrects a few results. In challenge 24 of Survivor 38, the duel Chris won to return from the Edge of Extinction, survivoR had listed Reem and Wardog as winners too; it now lists only Chris. It also fixes results in a team reward challenge in Survivor 39 and a tribe challenge in Survivor 50, who sat out in 8 seasons, and team labels in 3.


How BEAST Scores a Finishing Order

BEAST rates challenge performance with Elo. It scores a challenge as a set of matchups between pairs of players, all worked out from the ratings before the challenge, with a K of 14 for each individual matchup. How many matchups a challenge makes depends on what survivoR records.

What survivoR records Matchups in a field of 12 Last place
The winner only 11: the winner beats each of the others 1 loss
The full order 66: 1st beats 2nd to 12th, 2nd beats 3rd to 12th, and so on 11 losses

With the new orders, the challenges BEAST scores in full go from 20 to 189, and the matchups from individual challenges go from 4,857 to 8,852. Team challenges, where BEAST divides K by the size of each side, are unchanged at 33,070 matchups.

Wardog shows how it adds up. In Survivor 38, Wardog finished 11th of 13, 12th of 12, 11th of 11 and 6th of 6 in 4 immunity challenges that now have a full order. BEAST used to count 4 losses there, 1 to each winner. It now counts 36 losses and 2 wins, and the new orders cost Wardog 130.0. The corrected duel costs another 76.3. Wardog goes from 1508.8, 247th of 874, to 1302.5, 873rd.


Who Moved

Player Seasons Before After Change Rank
Wardog 38 1508.8 1302.5 (F) -206.3 247th to 873rd
David Wright 33, 38 1445.7 1595.3 (A) +149.7 791st to 58th
Nick Wilson 37, 40 1516.5 1648.3 (S) +131.8 201st to 12th
Lauren O'Connell 38 1415.4 1547.2 (B) +131.8 853rd to 135th
Ben Driebergen 35, 40 1411.6 1540.2 (B) +128.6 855th to 146th
Erik Reichenbach 16, 26 1596.2 1479.8 (C) -116.4 36th to 592nd
Hannah Shapiro 33 1444.8 1333.0 (F) -111.9 794th to 869th
Leif Manson 24 1464.1 1569.3 (A) +105.2 708th to 93rd
Carolyn Wiger 44 1432.1 1328.7 (F) -103.4 827th to 871st
Marty Piombo 21 1445.2 1546.8 (B) +101.6 793rd to 136th

These are the 10 largest changes. Apart from 76.3 of Wardog's drop and 14.1 of David Wright's gain, which come from the corrected duel, they come almost entirely from the new orders.

The ratings spread out. More matchups per challenge means larger moves in both directions. The standard deviation of BEAST goes from 51.8 to 61.0, and the tier cutoffs move with it: S tier now starts at 1621.1, up from 1602.7. Of the past players, 209 change tier, 10 of them only because the cutoffs moved. Among them, S tier grows from 26 players to 30, and F tier from 5 to 11.

The top of the list holds. Simon Mee, Joe Anglim, Kyle Ostwald and Joe Hunter are still the top 4. Joe Anglim gains 59.8, and Ozzy moves from 12th to 5th. No player who played only Australian Survivor moves.


What It Does to PROBST

PROBST uses BEAST for its immunity odds and as "challenge threat" in its odds of going home. Each backtest below trains PROBST's models on earlier seasons only and forecasts the next, once with the old orders and once with the new. Lower log loss is better.

Backtest Old orders New orders
Who wins individual immunity, 309 challenges: log loss 1.859 1.864
Picks the immunity winner 26.2% 23.6%
Who goes home, 583 boots in 40 seasons: log loss 2.273 2.273
First pick went home 18.2% 19.4%
Season winner going into the merge, 40 seasons: log loss 2.395 2.378
Winner in the top 3 at the merge 50.0% 47.5%

The differences are small and point both ways. The immunity forecast is a little worse with the new orders; with no orders at all, its log loss is 1.863. In the immunity model, BEAST's weight per standard deviation goes from 0.226 to 0.212. The new data changes who BEAST rates highest more than it changes how well PROBST forecasts.


Survivor 51

The new orders do not change any Survivor 51 rating: every individual challenge this season already had a full order. They do move the cast's ranks among all players. Mike, at 1637.2, is 15th of 874, down from 13th; Mike Holloway, Cole Medders and Nick Wilson moved above him, and Boston Rob moved below.

PROBST's Episode 9 forecast, built once with the old orders and once with the new:

Player Goes home, old Goes home, new Wins immunity, old Wins immunity, new
Jelly 36.8% 36.6% 8.9% 8.8%
Devin 26.5% 26.2% 10.2% 10.5%
Mike 6.6% 6.6% 18.7% 16.8%
Lewis 5.1% 5.3% 15.9% 14.9%
Carter 2.2% 2.2% 12.5% 12.4%

No player's chance of going home moves by more than 0.3 points. Mike's immunity odds fall from 18.7% to 16.8%. The immunity model learns from past seasons how much a BEAST point is worth; with the ratings spread wider and the weight a little lower, Mike's lead counts for less. Every player's odds to win the season move by less than 1 point, and the odds of reaching the final 3 by up to 1.5.


What Did Not Change

BEAST's rules are the same: a K of 14 for individual matchups, and 12 for team challenges, divided by the size of each side. The ratings on the site use the new data from this week.


Explore the full rankings, player histories, season snapshots and predictions at survivorelo.com.

Thanks to Dan Oehm (@danoehm.bsky.social) for his survivoR R package, which provides the underlying data.

Player Profiles