Welcome to Ten Words of Wisdom, the place where a picture is worth a thousand words, so you get... um, one percent of the picture.

Cary Huang, creator

Ten Words of Wisdom, also abbreviated as TWOW, is a YouTube camp which tests contestants' ability to come up with a relevant and creative response that fits in with a particular prompt, in ten words or fewer. Labeled as BAGUETTE's reboot, the series began with a staggering 491 contestants. The first episode, TWOW 0A, was released on December 12, 2015.


Contestants in TWOW are identified by original YouTube username, alongside a book icon, either generated systematically or supplied by the contestants themselves.

In TWOW 14A and onward, living contestants may request a book icon change and submit a confessional, to be read aloud, given that it is of reasonable length.



Before the start of each round, Cary announces the prompt for the round. Eligible contestants will generally have up to three days to respond to this prompt, and will be eliminated if no response is submitted in time. Their responses are limited to ten words, along with any other restrictions given by the prompt. No official penalty is given if any rules are broken—penalties are applied entirely by voters.

Under the initial circumstances, contestants could submit only one response per prompt. TWOW 17 and onward, however, contestants may submit two responses each. These counts may be raised by having earned a prize.


Every "A" video contains a voting portion, typically one minute in duration. This voting portion consists of hundreds or thousands of screens, each with an identifier and a randomly selected set of responses. Viewers must pause within this region to receive responses to vote on.

Screenshot 2016-02-04 at 12.21.02 PM

One voting page displays at most ten entries, labeled with letters of the alphabet, typically A through J. Each response has a word count in blue to its right. Viewers must vote using the episode’s comments section before the voting deadline, which generally falls three days after the episode’s release. A vote consists of a screen’s keyword identifier and the letter labels of the responses for that screen, ranked from best to worst (from left to right). The keyword and the rankings must be separated with a space and surrounded by a pair of square brackets.

Screenshot 2016-02-04 at 12.40.06 PM

Any other text will be ignored by the vote counter, so viewers may write other text alongside the vote, and it will still be valid.

If more than one vote is received from a single user, distributed among the comments in any manner, then the voting power will be divided evenly among all the votes. In TWOW 16A going forward, a single user may submit up to ten votes. Before TWOW 16A, the limit was not present, leading to astronomically high vote counts that eventually caused problems with vote collection.

The unique identifier on each screen, seen to the upper right, is a random English word from a list (with some omissions[1]) that ranges from three through six letters in length.

The voting screens flash by at a rate of thirty-two per second (pre-TWOW 15A) or eight per second (TWOW 15A and beyond). Given a one minute voting period and a thirty-two screen per second rate, there are 1,920 unique screens in a this single voting period. The number is kept high to preserve anonymity for votes and to keep contestants from being pressured into favoring certain responses.


Average percentile

At its most basic, the average percentile of a response is a measure of how well voters viewed it against its competitors. The response adopts this percentage as its “score”, which is compared against the others to form the results. A greater average percentile signifies that it performed better.

More specifically, a contestant's percentile in one vote is calculated by counting how many other responses this response places better than, and then dividing this value by the total number of other responses in the voting screen. For example, if a viewer ranks a response third on a screen with ten entries, then it beats 7 out of 9 (~77.78%) other opponents, which is its percentile for this vote. Responses excluded from a vote have the remaining available percentage points distributed evenly among them.

Screenshot 2016-02-04 at 12.25.39 PM

With this system, the average of all the percentiles in a single vote should be 50%, and the overall average, among all the percentiles, should also be 50%.

This method of calculating percentiles does not match the formula under the traditional definition. By that formula, the denominator would not be the number of other responses, but rather the total number of responses. One characteristic of this change would be that it would be impossible to achieve a 100% percentile. The method used in TWOW would be more commonly described as a Borda count starting at 0.

The overall average percentile is found by taking a contestant's percentile in every vote that includes their response, and finding the average. If a single voter casts more than one vote, then the votes must be weighted accordingly.

Normalized ranking

Normalized ranking (or relative ranking) is a conversion of a numerical ranking to a percentage between 0 and 1, inclusive. This percentage measures how well a contestant placed relative to the others, regardless of the size of differences in average percentile.

The normalized ranking for a contestant can be calculated by taking one less than their numerical ranking and dividing it by one less than the total number of contestants, and then subtracting the quotient from one. This method always gives 100% to first place and 0% to last place, where it is defined.

As of TWOW 14B, Cary’s simulations, used to predict contestants’ future performances based on past ones, use normalized rankings to form the normal distribution curves of probabilities. Before then, these curves relied on average percentiles, which favored good performance on earlier prompts, where high percentiles were easier to achieve.

Normalized ranking typically is not displayed on the round leaderboards.

Standard deviation

The standard deviation of a response quantifies the amount of variation in how different votes ranked it.

Standard deviation is calculated using the population standard deviation formula: $ \sqrt{\frac{\sum_{i=1}^{n} \left(\left(x_i - \bar{x}\right)^2\right)}{n}} $ where $ n $ is the number of votes that affect the response, $ x_i $ is the ranking given by each vote, and $ \bar{x} $ is the mean ranking. Standard deviation is a secondary statistic and strategically holds no value outside of tiebreaking.


As explained in TWOW 3A, if two contestants score the same average percentile, then the response with the higher standard deviation prevails, “to encourage exciting responses over bland ones.” Should standard deviation tie, then higher vote count wins. If this matches as well, then votes that dealt with each entry will be iterated through by ascending order of the time they were cast, and the first response to have a higher percentile in a singular vote will win.


If the contestants continue to tie, then they will share the ranking.[2]


In every "B" part, the fifth of contestants with the lowest average percentiles must leave the show. Following TWOW 16B, contestants are eliminated one by one.

In "A" parts, contestants may leave too, by having not responded to the prompt in time. Contestants eliminated through this system are ranked at the bottom by their placements relative to each other on the previous round. Special cases include Juhmatok, who left voluntarily, and Sam Billinge, who was disqualified for being an alternate account of TheMightyMidge.

If a message was sent but not received (as with Meester Tweester in TWOW 4A) or Cary deems an elimination to be unfair (as with fryUaj in TWOW 10B), Cary may bring contestants back, though this mechanic is used sparingly. Returning contestants are placed at the bottom of the safe zone for the round in the order of their rankings on the previous round.


Occasionally, Cary may record a response incorrectly or forget to accommodate for limitations in his graphics rendering programs.

In TWOW 3B, he explains that errors by a few characters may generally receive a 5% percentile boost, and mistakes where entire words are missed may receive a 10% boost. Missing a response entirely, thus resulting in elimination in the A part of an episode, can be made up by letting the mistaken contestant rejoin. Compensation boosts may not push otherwise safe contestants into elimination. In TWOW 11A, this policy was revised so that any misrecording receives only a 3% boost.

Minor mistakes caused by a lack of support for non-ASCII characters are also no longer compensated with a boost, since Cary has warned contestants about it numerous times.


The prize for TWOW is supposedly "wisdom", but will likely be changed.


  • The font used for TWOW episodes and graphics is “Interstate”.

Season 1 Episodes

Episode Name Contestants Remaining

(after 'B' portion)

Link to episode Prize
TWOW 0 491 A:


TWOW 1 393 A:


Double response prize
TWOW 2 284 A:


Choose a word for TWOW 4
TWOW 3 211 A:


A +5% bonus for you and the three contestants of your choice on TWOW 4.
TWOW 4 162 A:


Double response prize

(originally had a twist, but was removed)

TWOW 5 122 A:


Choosing a sidekick for TWOW 6
TWOW 6 97 A:


Double response prize
TWOW 7 74 A:


7% bonus

(originally going to be 3% - 12% bonus)

TWOW 8 56 A:


Guess your score

(originally going to be a prompt that will be given to you 24 hours earlier than everyone)

TWOW 9 44 A:


Choose a word for TWOW 11

(originally going to be a double response prize that uses 11-20 words)

TWOW 10 34 A:


Double response prize
TWOW 11 28 A:


Double response prize
TWOW 12 21

A: B:

Double response prize
TWOW 13 18



Double response prize
TWOW 14 14


B Part 1:

B Part 2:

Double response prize
TWOW 15 12 A: B: Double response prize
TWOW 16 10 A:

B Part 1:

B Part 2:


(all TWOWers have double response in all future episodes but in 17A, Spicyman33 and Joseph Howard submitted three entries)

TWOW 17 9 A: B: Triple Response
TWOW 18 8 A: B: Triple Response
TWOW 19 7



Triple Response
TWOW 20 6 A: B Part 1:

B Part 2:

Triple Response


5 A: B: Triple Response
TWOW 22 4 A: B:

Three Mentions


3 A: B: Three Mentions


  1. Cary’s reply to Diamondcup67 on TWOW 7A
  2. “If there’s a tie for tenth place, everyone in the tie will get the prize.” (TWOW 1A)