Personalities Methodology About
How it works
The OTR Score
Methodology

Every take is scored on four dimensions. The result is an OTR Line — points at stake on every call. Correct: bank it. Wrong: lose it. No partial credit.

OTR Line = Base(100) × Difficulty(1–3×) × Specificity(0.2–1.0) × Horizon(1.0–1.5×)
OTR Score = Running total of all banked and burned Lines · Cumulative · Permanent · Cannot be deleted
Accuracy Rate
40%
The foundation of the OTR Score. Correct calls divided by all resolved takes. Adjusted for category — score predictions weighted differently than long-term narrative calls. The clearest single signal of who actually knows their stuff.
Claim Difficulty
20%
Sets the Line value. Safe takes earn less and risk less. Bold takes put more on the line — in both directions. Difficulty cross-referenced against market consensus and odds at the time the take was made. For sportsbook takes, the posted odds set the difficulty automatically.
Specificity
20%
Hedge language kills your Line. "The Lakers will struggle" scores 0.2×. "Lakers finish 9th in the West with 41 wins" scores 1.0×. Words like "might," "could," and "wouldn't surprise me" are automatically detected and penalized.
Time Horizon
10%
Short-term takes settle fast and earn less. Long-horizon takes are in play longer — but the Line reflects the conviction. A career or era narrative call that lands is a major OTR moment.
Accountability Behavior
10%
The integrity layer. Tracked from transcripts and social media. Doesn't change the Line but adjusts the overall OTR Score. Positive: revisiting calls unprompted, public corrections, updating stance with evidence. Negative: deleting posts after a verdict, retroactive reframing, pretending a take never happened. Deleted posts are archived by OTR permanently.
Safe
1.5×
Consensus
Bold
2.5×
Contrarian
Wild

The vast majority of takes are classified and scored automatically. When our system has lower confidence — ambiguous subjects, multi-speaker clips, satirical framing — those takes are flagged for editorial review before publishing.

Every editorial decision is logged with a reason code and timestamp. Reviewed takes are marked "Editorially Reviewed" on the platform. We show our work.