Most trading "strategies" on YouTube are locked inside long videos — often in another language, usually light on specifics, and frequently a runway to a paywall pitch. I wanted to see if I could pull the actual idea out of those videos and test it honestly, without watching hours of footage.
Here's a workflow I ran end to end, using Scribefy — MCP server for YouTube transcripts — as the front door.
1. Extract — turn video into something a model can reason over
I gave Claude a link to a trading video and had it call Scribefy to pull the full transcript. Because Scribefy handles non-English content, I worked through a Bulgarian creator's channel entirely in English — the intro video, then a live session, then the "theory" video where the mechanics were actually explained. Hours of watching collapsed into a few minutes of extraction.
2. Reverse-engineer — separate the mechanism from the narrative
With the transcripts in hand, Claude and I distilled the method to its codifiable core: a "Power-of-Three (PO3) dealing range" — a formula that partitions price into fixed levels. There's a lot of narrative wrapped around it in the videos; having the raw transcript let us isolate the part that's actually testable and set the rest aside.
3. Test — let the backtester decide, not the hype
This is the part that matters. We translated the logic into a Pine Script strategy for TradingView, hit the free-tier data limits, and moved to a MetaTrader 5 Expert Advisor (MQL5) for unlimited historical backtesting.
The non-negotiable: every version shipped with a built-in control — a deliberately misaligned grid — so we could tell a real edge from a good-looking fluke. (If a "strategy" can't beat its own scrambled control, it doesn't have an edge.)
The original idea came out marginal. But the process pointed us toward a session-based breakout variant that, on gold (XAUUSD), backtested at:
- Profit factor 1.23
- Sharpe 1.58
- Max drawdown 6.5%
- …and, crucially, it passed its falsification control.
The takeaway
Scribefy isn't "just a transcript tool" — it's the front door to a research pipeline. Extract the idea, codify it, and let the backtester (not the video's hype) decide whether it's real. The transcript is what makes the whole loop possible — especially when the best explanation is buried in a 40-minute video in a language you don't speak.
Try it: scribefy.app · setup guides at scribefy.app/guides
Not financial advice. These are historical backtest results on past data — not a promise of future performance, and nothing here is a recommendation to trade. The point of this post is the research **process, not the strategy.
Top comments (0)