chore(sl-hunting): pre-open note for 01 OCT 2026 (+ 30 Sep comparison)

[DoRmAmMu1997]

Summary

The pre-open note for 01 OCT 2026 (ZrSyia9MGHc, uploaded 30 Sep 21:25 IST). The 30 Sep trade comparison is in this PR too, as a doc-only addendum.

One plan for every open: SELL. It flipped from 30 Sep, where a flat open bought. On 30 Sep, positive momentum on all three indices was followed by as much selling. After a rejection like that, few people get to join the selling, so there are no sellers to hunt.

index	resistance	support
NIFTY	22790 / 22900	22500 / 22400 (22,500 is his "important psychology number")
BANKNIFTY	55000 / 55200	54000 / 53790
SENSEX	73130 / 73450	72000 / 71500 (unchanged)

• All levels were read off the 1080p chart.
• The transcript said BankNIFTY's resistances farthest-first ("55 200 55000"); they're recorded nearest-first.
• SENSEX's weekly contract expires today. NIFTY's and BankNIFTY's don't, and the note marks that second part as not coming from his video.

Addendum: the 30 Sep comparison (doc only, no knowledge change)

Same read, same instrument, opposite result.

• IH bought the flat open at 09:18, held through a dip to 22,685.10, and booked at 09:45 for +2,51,147.85. 85% of that came from his two BankNIFTY legs.
• The agent bought the same NIFTY 22700 CE (184.75, against his 184.16 average) at 09:22:49. It was stopped four seconds later: −301.75.

Why the agent's trade failed so fast:

• The decision was the first after two restarts that switched the model, and it took 87.2 seconds.
• By the time the order fired, spot was 22,705.20, against the 22,727.70 entry and 22,701.00 stop the model had named. Only 16% of the planned risk was left.
• Then it held all 67 remaining bars. 46 of them cited the stopped long, while NIFTY rose to 22,766.00.

Three candidate rules, measured and refused:

candidate	what the data says	verdict
Loosen re-entry after a loss	After a losing trade that lasted 2 minutes or less, the 17 same-side re-entries made −13,448.25 with 18% winners. 12 of 14 sessions were negative, and both halves were negative.	Refused: the holds were right by the book's history
Refuse an order when most of the declared risk is already used up (extends SLH-016)	There's a real gradient (50–75% left: 23 trades, −8,259.00). But only a 75% cut-off holds up in both halves and without the worst two sessions; 70% and 80% flip positive.	Refused: one exact number that happens to work is curve-fitting
Refuse an order placed late in a long decision call	58 entries have call timing, and delay shows no pattern	Refused

A defect, offered but not built (SLH-021):

• SLH-012 journals a stale decision call's trade only if the position is still open when the call's result is collected.
• A trade opened and stopped inside its own call gets dropped from both the journal and the decisions log.
• I found this 4 times in the runner log (10 Aug, 4 Sep, 15 Sep, 30 Sep); three of them happened after SLH-012 shipped.
• Every missing trade is a fast stop, so the journal currently flatters the book.

Test plan

• The note test is retargeted to 01 Oct. Eight deliberate mutations of the note were all caught, and a control rewording passes. Premarket tests: 19 passed.
• The comparison addendum changes only the doc. SL Hunting suite: 283 passed; pre-commit passes. Every figure in the doc was re-checked against the scripts' output, and one wrong figure was fixed (the stop sat 1 point above 22,700, not below it).

🤖 Generated with Claude Code