How scores are calculated

1. What is scored
AI extracts calls about a stock or asset rising or falling from video transcripts, then scores them using the actual price one month later. Only people with at least 20 scored one-month calls are ranked.

2. Market and beta adjustment
A rising market makes bullish calls easier to get right. For individual stocks, we subtract the market return over the same period (KOSPI for Korea, S&P 500 elsewhere) multiplied by the stock's Beta (sensitivity to market moves) to obtain the Market excess return. For example, if the market drops 5%, a stock with beta 2 is expected to drop 10%. That expected move is excluded from the assessment of stock selection skill. Index calls assess market direction separately. Calls about indices, rates, gold and Bitcoin are scored directly because they have no comparison benchmark.

3. Return-weighted hit rate
This measures the share of total absolute excess returns attributable to correct calls, rather than simply counting hits. Frequent small wins can be outweighed by large misses.

4. Peer excess
This measures performance relative to other YouTubers calling the same stock at the same time.

5. Accuracy score = Peer excess + Market excess return × 0.3 + (Return-weighted hit rate − 50) × 0.1
A score above zero indicates better-than-average forecasting; below zero indicates worse-than-average forecasting.

6. Bias, horizon and badges
Bias is the share of bullish and bearish calls. Best horizon is the most accurate interval from one day to one year. Performance badges include Hwasubun for consistently accurate forecasting.

Caution
Transcription and AI extraction errors, name variations and small samples (20–40 calls) can strongly affect results. This is not investment advice.