Methodology & rubric

Methodology & rubric

The rubric is the product. Everything below is public, fixed, and applies to every file.

01How a verdict is made

STEP 1
Transcript
The video's captions are fetched (Hindi + English). Every rule is quoted from them, with a timestamp. If no transcript exists, the file is stamped RECONSTRUCTION and can never receive a full verdict.
STEP 2
Labeled rules
Each rule is graded: Said quoted in the video · Said · not mechanized stated but not codable (e.g. DMI without parameters) · Said · discretionary judgment-based (e.g. OI-profile context) · Inferred filled in from context. Nothing is guessed silently.
STEP 3
Real backtest
The rules run on real market data with real costs: real bid/ask for options, measured charges for futures, one trade per day, stop-wins-ties, entries never better than the trigger. A rule-variant grid (e.g. target 8/11/15/20/30) proves the result is not tuned.
STEP 4
Verdict
A stamp and a score out of 100 — backed by re-runnable evidence. Every rule is challengeable, and the published verdict is always the strongest evidence we have.

02The score and the stamp

The 0–100 score grades the video's central claim — the strategy as promised — not the person, and not every idea inside the video. A video can fail at 25/100 while one of its setups shows promise; the verdict card says exactly which part was judged.

Failed
Failed
25 / 100
rules: transcript · data: real ticks
The claim loses money after real costs, in the tested configuration.
Passed
Passed
75 / 100
rules: transcript · data: real ticks
The claim survives real costs on a meaningful sample. (No file has earned this yet.)
Unverified
Unverified
— / 100
rules: — · data: —
Not enough data or not enough transcript fidelity to judge — an honest "no verdict" instead of a guess.
Reconstruction
Reconstruction
— / 100
rules: reconstruction
Built without the transcript. Never a full verdict; replaced by a transcript-verified revision when captions are obtained.

03The correction policy

Every rule is challengeable — that is how you know the verdicts hold. A challenge with corrected rules goes back through the pipeline, the file is re-run, and the published verdict is always the strongest evidence we have. No verdict is fixed in place against new data: files are re-verified as larger datasets land. What you read on this site is the result of that process, not the process itself.

Challenge a verdict

04Data & cost honesty

  • Options: real tick data (timestamp, LTP, bid, ask) — entry at ASK, exit at BID. The spread is real, never modeled.
  • Futures: measured NIFTY round-trip charges (~7 pts/trade from our own futures studies), applied to every trade.
  • Synthetic or approximate inputs, when used, are labeled on the file itself.
  • Every trade is downloadable. Every script is re-runnable. If the numbers disagree with the data, the data wins and the file is corrected.