How Breakout works

Last updated October 2026

Breakout shows aspiring actors how they land with two different audiences: verified working professionals and the public. This page explains exactly how the ratings work and how we try to keep them fair. The questions below are the ones judges and actors ask most.

The monthly challenge

Each month everyone performs the same speech (a public-domain play or an original written for Breakout), films it in one take, and posts it on YouTube. Performances are approved by hand, then rated for about a week or two. When the challenge closes, results are published to everyone at once. Free for actors, always.

Accounts

• There are no passwords. You enter your email and we send a link that signs you in. It works once and expires in 15 minutes.
• Everyone is 18 or older. You confirm it when you sign in.
• Judges are verified by hand before they can rate anything: we look at their credits (IMDb, a website, a festival or theater listing).
• We store your email, your name, and what you do on the site. We don't sell it. See the privacy page.

Profile pages

Performers can have a shareable page with their reel, a short bio and a few extra videos. It starts private; you choose private, members only, or public, and whether Google may list it. Judges have an optional profile too, hidden by default and never connected to any rating. Profile pages never affect scores.

Managers and agents

Talent managers and agents can join as judges. If a judge represents a performer, they cannot rate that performer, and every representation must be declared. Nobody on Breakout is charged to be represented, reviewed or ranked.

Messages

Judges can write to performers they rated once a challenge's results are published, and performers can write to judges who have chosen to allow it. Nobody can message anyone while a judge is judging that performer's entry. Messages are private between the two people, can be blocked, muted or reported, and links and phone numbers stay blocked until the other person has replied. Real casting never asks actors to pay.

How a judge rates

• About 10 reels a month, under an hour. Each is a performance of the same speech.
• You rate four skills from 1 to 5: delivery, emotional truth, range and screen presence. You can leave a private note and flag anyone you'd like to see more from.
• You never choose who you rate, and performers never choose their judges. Reels are dealt at random, spread evenly across judges.
• The performer's name is hidden until after you rate.
• If you recognize the actor, tap "I know this actor". It doesn't count and another judge gets the reel.
• If you teach or represent anyone who might enter, tell us ([email protected]) and we'll record it. Those performers are then never dealt to you, and if one already was, the reel is taken back and given to another judge.
• Your ratings are never shown with your name. Notes appear to the actor signed only with your role ("Independent filmmaker · verified"). You're named publicly only if you say yes.

How the public rates

• Two performances side by side. You watch both (playback is measured; skipping ahead doesn't count) and pick the one that landed. Optionally pick a winner on delivery, emotional truth and screen presence.
• The server deals every matchup at random. You can't look anyone up, so you can't target a friend or a rival. Under-seen performances are dealt more often so everyone gets a fair number of matchups.
• Performers can't vote in their own challenge, and judges rate in the judge queue, not the public game, so the two audiences stay separate.
• Nobody sees any scores, or how anyone else voted, until the challenge closes. There is no live leaderboard to chase or attack.

How the scores work

• Industry score: for each skill we take the median of the judges' ratings (one very harsh or very generous judge can't swing it), then average the four. It's published only when at least 5 judges rated that performance, and it always shows how many.
• Public score: from all the head-to-head votes we estimate each performance's chance of beating a typical one (a standard method called Bradley–Terry, lightly smoothed so a 5–0 start isn't treated as certainty). It's published only once a performance has enough matchups.
• The gap: the two audiences use different scales, so we compare ranks, not raw numbers: "ranked 12th of 40 by the public, 31st of 40 by the industry." The gap is shown only when both scores exist.

Keeping it fair

• Random dealing for judges and the public: nobody picks who they rate.
• Voters are weighted by consistency, never by agreement with the industry. About 1 in 8 matchups quietly repeats an earlier pair in the opposite order. Careless or hostile voters contradict themselves and count less. (Weighting by agreement with the pros would erase the very gap the site exists to show.)
• Fake accounts gain nothing. Accounts that appear to come from one network share the weight of three voters, so a vote farm of twenty accounts counts as three. A real household or dorm of two or three people is unaffected. Very new accounts count a bit less.
• Shared-link votes don't count. A friend who arrives from an actor's share link can become a "fan," which the actor sees, but fans never change any score.
• Videos are locked once voting starts, so no one can swap a video after it's been rated.
• Rate limits and checks: a daily cap on matchups, minimum time per vote, and automatic flags when a suspicious group votes on one performer very differently from everyone else. A person reviews every flag before results are published.
• Reports: anyone signed in can report an entry. We remove entries that break the rules and ban accounts that manipulate ratings.

What we're honest about

• Small samples are real. A performance rated by 5 judges is a signal, not a verdict. That's why every score shows how many people it's based on, and why early challenges are small, invited cohorts sized to the number of judges.
• Ratings contain taste and bias. "Screen presence" in particular can reflect how someone looks. We ask judges to rate skills, hide names while rating, and watch for outliers, but we can't make bias disappear.
• Playback measuring isn't perfect. It's one layer among several, not a guarantee.
• Breakout is a new service. If a rule here turns out to be wrong, we'll change it and say so on this page.

Questions?

Email [email protected]. A person (Michael Krewson) reads and answers every message.