feat(sl-hunting): v5i — when nobody is seated, your own entry is what seats them (+ 23 Sep pre-open note)

[DoRmAmMu1997]

The rare part: both books agreed, and both lost

IH and the agent took the same side, on the same premise, on the same session — and both were stopped. That agreement is what makes today worth a rule: the loss can't be pinned on the read, because two independent readers produced the same one.

IH bought the flat open across all three indices (BankNIFTY 56,500 CE, SENSEX 75,000 CE, NIFTY 1,430 qty on its expiry day), on the note's reasoning almost word for word — the previous day rose "slowly, slowly", so buyers came but "would not have gone holding the trade", so "the buyers' SLs are not available… we can continue to follow the market."

The agent's two entries cited the same FLAT → BUY branch off SLH-017's verdict (+0.17% NIFTY, +0.03% BankNIFTY):

#	time	entry	stop	dist	lots	exit	outcome
1	09:19	23,452.80	23,434.00	18.8	2	09:26 AI_STOP @ 23,432.30	+₹280
2	09:52	23,459.45	23,450.00	9.45	4	09:57 AI_STOP @ 23,450.65	−₹2,931

Day −₹2,651.

The +₹280 on a 20.5-point adverse move is a mirror divergence, not a rounding artefact: NIFTY CE −767, BankNIFTY CE +1,047 — the indices moved apart inside those seven minutes. Which also means the NIFTY-spot stop closed a basket that was in profit at that moment. That's the leg tie working as designed, but it's the first measured case of it.

What's net-new

IH narrated himself becoming the crowd he entered because it was absent:

"When the market takes support and then goes up slowly, it is possible other people also start buying. Sellers move away — and we had already bought here too. All of their stops are around the same place, near the closing price. So the market came down to target them."

Every rule in this family keeps you outside the crowd — v5f (a one-way day seats nobody), v5c (a retracement recruits the crowd you hunt next), v4e (don't forecast one into existence). None of them covers the follow trade that fills the book it was entered on. v5i also supplies the mechanism under v5h's "a standing premise is not a second reason": the branch doesn't refresh because acting on it consumes it.

Three repairs, priced first, all refused

IH attributes the whole thing to time — "the more the market stalls, the more the chances of opposite momentum in your trade." That's testable, so it was tested on all 151 closed journal trades before any of it could become a live rule.

1. A time stop — the book says the reverse.

hold	n	mean	win %
0–5 min	86	+₹25	36%
5–10 min	41	+₹225	44%
10–20 min	20	+₹1,562	80%
20–40 min	3	+₹1,386	100%

r(hold, P&L) = +0.21. Winners averaged 7.4 min, losers 4.0. A cap was negative at every level — −8,535 at 30 min through −53,184 at 5. Causation runs the other way: a working trade is held to target, a failing one hits its stop fast. Duration is a symptom, not a dial — the same shape as v5a's correction yesterday.

2. No re-entry after a stop. n=9, +₹358 a trade, ahead of the book's own first-trade-of-day mean of +₹128. Refusing them would have cost ₹3,222.

3. Tightening v5h's new-extreme test. After a stop that test is vacuous — the exit sits on the adverse side, so any recovery clears it. The obvious repair (require the re-entry to extend past the prior entry) is the worst of the three: it forgoes ₹36,580 of a ₹55,337 book, and the trades it blocks average +₹1,219 against −₹341 for the ones it keeps. The inversion is the finding — re-entering at a price not beyond your last entry is the better trade here.

All three are written into the rule as refusals, so they can't be quietly re-derived. A mutation deleting any of them is caught.

The honest limit

v5i would not have changed either of today's trades. Both were stopped inside six minutes — far too fast for any ageing mechanism to be the cause. It's carried on IH's evidence; what this book proved is the three refusals. That's stated in the rule itself, and the mutation that turns it into a self-proving claim is caught.

Worth flagging separately: this is the second consecutive session where every entry cited the standing pre-open premise (18 Sep had four, today two). v5h named that failure; it hasn't changed the behaviour yet.

Negative-tested 10 ways, all 10 caught, plus a passing control.

Gates: 606 master · 28 market-data-health · 1581 pytest · ruff 0.16.7 · mypy (80 files) · compileall.

---

Addendum: the pre-open note for 23 SEP 2026

Source: Prediction For 23 SEP 2026 (FBDx9zcN9Sw, uploaded 2026-09-22 21:25 IST).

Same word, opposite premise — which is exactly the case v5i was written for.

	22 Sep	23 Sep
why BUY	nobody seated → follow	sellers seated → hunt
gap down	SELL	small: still BUY · large: not stated

Yesterday's follow premise was consumed by the trades that acted on it. A note that just said "BUY again" would invite the agent to cite it a third time, so this one names both premises side by side and says outright that the follow reason is spent.

The premise: selling ran all day with "no big retracement" on a market "already negative from before", so sellers are seated. A retracement did come near 2 PM, but selling resumed after it with plenty of session left — the evidence they're seated, not a reason to doubt it. Flat to gap up → buying-side setups "to make these sellers our target": a squeeze, not a follow.

There is no sell branch tonight. A small gap down keeps the same BUY plan; a large one simply isn't covered. The test asserts the absence, because an absent branch is what gets filled in by analogy with the night before.

One garble, resolved from the chart rather than guessed. NIFTY's supports came through as "23 270 230". Read naturally that's 23270 / 23230 — and it's wrong. A 1080p frame at 1:29 tags the lines 23,271.75 and 23,201.20, so the levels are 23270 / 23200. Resistances cross-checked on the same frame (23,500.15 / 23,567.20), recorded as spoken.

	resistance	support
NIFTY	23500 / 23560	23270 / 23200
BANKNIFTY	56540 / 57000	56100 / 55910
SENSEX	75200 / 75500	74350 / 74000

Negative-tested 10 ways, all 10 caught, plus a passing control — including the natural misreading of the garble (23230), carrying the follow premise forward, and inventing the missing sell branch.

Gates: 606 master · 28 market-data-health · 1581 pytest · ruff · mypy (80 files) · compileall.

🤖 Generated with Claude Code