First off, the basics: Our scores are
not necessarily a means to measure performance within a season
(even though we often treat them as such). Instead, they're an
attempt to quantify a Survivor contestant's
likelihood of future success, much as predictive stats like
WAR do in baseball. Essentially they assess: How likely is
this contestant to win the next time they play Survivor?
(Assuming no drop-off in performance over time). The theory is
as follows:
The ideal Survivor contestant should have
all three "tools" required to play: (1) challenge ability
(easiest to quantify, just by counting wins), (2) social
ability, and (3) strategic ability. The last two are difficult
to measure, but both are reflected in a couple of measurable
ways: tribal council performance, and receiving jury votes.
Tribal council performance: A good strategic player will guide
the vote in such a way as to (almost) always know, and vote
for, the person that gets booted. Similarly, part of a good
social game is being sufficiently in the loop to do the same
thing. Conversely, a good social player should rarely receive
votes against him- or herself (and it makes no strategic sense
to receive lots of votes against, either). Getting voted out
is not good strategy. Jury votes: The ultimate test of both
strategic and social prowess is, of course, being able to
convince the jury to award you the million dollar prize. So
jury votes are, despite the small sample size, fairly
important.
To put all this together, we created the Survival
Average (SurvAv).
It's a simple sum of fractional Challenge Wins (ChW),
weighted TC Ratio (wTCR), and (eventually) a weighted Jury%.
This has two advantages over SurvSc, measured below: (1) It's
always a positive number (or zero), since tribal council votes
are treated as a weighted ratio (instead of a potentially
negative sum), and (2) it has a larger spread, and in looking
at past seasons, we just feel it better represents the games
we saw. As such, we've ordered the contestants based on this
number in the table of complete Season
3 totals.
Our other metric, which we less favor, is the Survival
Score (SurvSc). It's simply the sum of three
components, expressed as percentages: Challenge Win% (ChW%),
Tribal Council% (TC%), and Jury Vote%. So Survival
score = ChW% + TC% + JV%, for a max score of 2 (3
after the finale).
Neither ranking is perfect, however, and we're open to
suggestions. Feel free to comment below with potential
improvements. We're keeping both scoring systems active
through this season to see how they shake out.
Glossary of terms
- Challenge
stats
- ChW: Challenge Wins. For
tribal challenges, a contestant earns a fraction of 1
win, depending on if they participated (no points for
sitting out). So in a five-person tribe's win, each
participant gets (1/5) of a point, or 0.2 points.
Duels (or individual RCs as in Ep1) at Redemption
Island count as half a challenge (and half a win).
Individual challenge wins count as a full point.
- ChA: Challenge Appearances.
Used to calculate ChW%. Fractional for tribal
challenges (same as ChW), except sit-outs get charged
for an appearance, because they could have
participated.
- ChW%: Challenge Win%. Simply,
ChW% = ChW / ChA.
- SO: The number of times a
contestant sat out of a challenge.
- Tribal
Council stats
- VFB: Votes For Bootee. The
number of times the contestant has voted for the
person who was ultimately voted out. Applies only to
initial votes (no points for revotes in case of a
tie). Special case: In a final three TC, where only
one vote is cast (by the F3 IC winner), only that vote
counts.
- VAP: Votes Against the Player.
The total number of tribal council votes cast against
the contestant. Again, only initial votes count (no
penalty for revotes), and here a hidden immunity idol
(if played) erases the votes. In the special case of a
final three tribal council above, only the F3 bootee
receives a vote against.
- TotV: Total votes cast during
the tribal councils the player has attended (again,
only initial votes count). Used to adjust for
different vote totals as tribes shrink.
- TCA: Tribal council
appearances. The number of times a contestant has
attended tribal council (at which they voted).
- TC%: Tribal Council percent.
Attempts to reward voting for the bootee (which
players controlling the vote almost always do), while
punishing receiving votes yourself. The formula is:
TC% = [VFB - (VAP/TotV)] / TCA.
- wTCR: weighted Tribal Council
Ratio. Very similar in intent to TC%, but calculated
as a ratio of VFB to VAP, while also scaling to a
uniform number of TC appearances. The formula is as
follows: wTCR =2* [VFB / (4+VAP)] x (14/TCA). I
originally tried (1+VAP) to avoid dividing by zero,
but this overly rewarded getting zero votes against
relative to just one vote against, which seemed silly.
(4+VAP) scaled that effect back comfortably. 14 was
used as the scaling factor for TC appearances because
there are usually 14 episodes, then a final scaling
factor of 2 to bring maximal scores up to roughly even
with ChW and JV% high scores.
- Jury
stats
- JVF: Number of jury votes for
the contestant to win. Maximum nine (Earl Cole, Fiji),
theoretically.
- TotJ: Total number of jurors.
Necessary to not punish unanimous 7-juror winners (JT
Thomas, Tocantins).
- JV%: The percent of total jury
votes cast for the contestant, or Jur% = JVF/ TotJ.
This number is used, raw, in SurvSc, and is scaled in
SurvAv (multiplied by six) to make it similar in size
to ChW and wTCR.
- Overall
scores
- Survival Score (SurvSc). It's
simply the sum of Challenge Win% (ChW%) + Tribal
Council% (TC%) + Jury Vote%, for a maximum possible
score of 2 (3 after the finale).
- Survival Average (SurvAv). It's a
simple sum of fractional Challenge Wins (ChW),
weighted TC Ratio (wTCR), and (eventually) a weighted
Jury%. The latter two max out at six points total, for
a theoretical maximum score of around 18 or so.