🏦⚡ stockloop desk +₹132.06

# #stockloop retro — 2026-07-03 (Kuber)

First retro since 2026-06-15. **The dead retro loop was itself the biggest
failure** — see below. Written as part of reviving the EOD routine.

## Result
- **Day realized: −₹597.4** (ADANIPOWER −553.5 swing stop + NATIONALUM −43.9 intraday).
- **Desk cum realized: +₹458.76 — GREEN** (after today's reconciliation).
- Week-to-date (W27): +0.65% vs 5% target — under target but positive.
- Book flat into the weekend (no carry).

## The headline: a MEASUREMENT failure, not a losing desk
The ledger showed −₹105 ("failing") for 3 weeks. Reconciling the **real Kite
Console tradebook** today revealed the desk had **banked ADANIGREEN +₹806 (5.6R)
and BHEL +₹355 (2.2R) on 07-02 and never logged them.** The desk was green all
along. Root cause: trades banked but not recorded + retros dead since 06-15 +
the self-improve loop starved of data. The trading was fine; the **bookkeeping
was broken.** FIX (now in place): daily Console-tradebook reconciliation via the
browser bridge is mandatory; this retro revives the EOD loop.

## Trades closed today
- **ADANIPOWER** −₹553.5 (−1R). Entry 226.96 (06-29, ema-pullback), held 4d to
  the structural GTT stop 220.9 (filled 220.81). **Held to plan, no discretionary
  cut — process CLEAN.** Drishti's order-flow scan flagged it seller-heavy
  (buy/sell 0.65) hours before the break → order-flow has EXIT-signal edge.
- **NATIONALUM** −₹43.9 (−0.21R). Owner-directed **intraday order-flow experiment**
  — bought the one demand-led name (book 1.05) on a seller tape, +1% target,
  tight bracket. Poked a fresh HoD 349.2 (peak ~+₹84) but a **33k-share sell wall
  at 348 capped the breakout**; book flipped to sellers (0.53) and it faded;
  squared at 15:19 for a small loss. **PROCESS FAIL (be honest): the position was
  GREEN (+₹50 at 15:11, +₹84 at peak) and I did not take it — I held for the fixed
  351.5 target and gave it all back to a −₹43.9 loss.** ~₹90-130 given up from peak
  to exit. Two errors: (a) the target 351.5 sat BEHIND the 348 sell wall — blocked
  from the start; (b) no trail / no take-profit-on-failed-breakout, so a winner
  round-tripped to a loser. This is the breakeven-stop mistake inverted.

## Scorecard snapshot (7/100 trades)
- **Expectancy +0.74R/trade** (positive), win 28.6%, avgWin 3.9R / avgLoss −0.53R.
- Per-setup: **ema-pullback E=+1.69R (n=4, 50% win) — THE EDGE**;
  intraday-orderflow E=−0.21R (n=1); sector-momentum(pre-pivot) E=−0.70R (n=2, the
  pre-rewrite churn). The committed swing strategy is working; the old churn and
  the new experiment are the drags, both tiny samples.
- Auto-improve: no changes (7/20, below the tuning gate — correctly avoids
  overfitting). Flagged the 2 old HoD entries + 2 intraday cuts as banned
  discipline breaches (all from 06-24 DRREDDY/TECHM, pre-rewrite).

## Lessons → guardrails
1. **RECONCILE DAILY against the Kite Console tradebook.** A 3-week logging lapse
   made a green desk look like a failure. Non-negotiable in every EOD now.
   (memory: [[kite-console-tradebook-reconcile]])
2. **Order-flow (buyer/seller book imbalance) has edge on EXITS, not just entries.**
   Drishti called ADANIPOWER's stop-out before it broke. Worth folding a
   book-pressure check into swing exit management.
3. **Before a breakout-TARGET trade, check the depth ABOVE.** NATIONALUM's +1%
   target sat right behind a 33k sell wall at 348 — the target was structurally
   blocked from the start. Read the overhead supply before setting the target;
   if there's a wall, set the target IN FRONT of it, not behind it.
6. **MANAGE THE WINNER — don't let it round-trip (new hard rule).** On an intraday
   trade: once +0.5R, TRAIL the stop to breakeven+ so it cannot become a loss;
   and TAKE PROFIT on a failed breakout / book-flip (a rejected HoD + buy/sell
   ratio flipping seller-heavy) rather than waiting for a fixed target or a clock.
   NATIONALUM was +₹84 and I gave it all back waiting for 351.5 — banked green was
   there and I didn't take it.
4. **5%/week needs OPEN scanning.** Midday order-flow scans arrive after the
   move (+4% already) and into a turned tape. Run Drishti's order-flow scan at
   the OPEN (09:15–10:00) Monday to catch lowball movers early — the first-45
   window is where the range is set ([[intraday-first45-edge]]).
5. **1:1 R intraday needs a high hit-rate to pay** — the NATIONALUM experiment is
   ONE data point, judged on process not P&L. Not adopting an intraday sleeve on
   n=1; keep it experimental, log more.

## Last-30-min fade STUDY (3 agents, 18 lowball names, ~13 sessions)
Owner asked: do these names fade in the last 30 min so we can short/take the
opposite side? **Answer: NO — disconfirmed across all 3 clusters.**
- Up-by-15:00 stocks **keep rising** into the close (avg last-30 positive almost
  everywhere: NATIONALUM +0.74%, VEDL +0.95%, RVNL +0.48%, NHPC +0.47%, BEL +0.45%).
- Blind "short at 15:00, cover 15:30" **loses on ~15 of 18 names.** Shorting a new
  post-15:00 HoD is a continuation trap (metals −0.9% to −1.3%).
- Genuine spike-and-round-trip (NATIONALUM-style) is RARE (~3 in the whole set,
  one on F&O expiry) — NOT systematic. **NATIONALUM 07-03 was an event, not a pattern.**
- **The real tilt is CONTINUATION**, not fade. Detail: [[last30min-study-no-fade-edge]].
- Caveat: 13 sessions, 0–4 up-days/name — directional, not statistically robust.

## STRATEGY STACK for Monday (by conviction — see MONDAY-PLAYBOOK.md)
**TIER 1 — CORE, LIVE (proven edge):** S1 ema-pullback positional swing (+1.69R,
the money-maker). Pre-open `swing_scanner.py` + regime gate + `stock-catalyst-recon`
(real NSE+Screener catalyst). LIMIT-on-pullback, extension gate, bracket, ≤1%/trade.
**TIER 2 — LIVE EXPERIMENT (tiny size):** S2 EARLY-LEADER catch-small-and-hold —
Drishti scans **09:15–09:45** for names up **SMALL (+0.5–2.5%) and TRENDING** (rising
5-min closes, not a gap-and-chop) on a confirmed catalyst + demand-led book; enter
small EARLY, bracket, and **HOLD the runner to the close** (continuation). Extension
gate: if already +4%, TOO LATE — skip. Data proof 07-03: AUROPHARMA +1.1% at 09:20 →
+3.8% close (I skipped it — judged "extended" at midday, wrong time to look); HFCL
+2.9% by 09:30 → +4.8%. Anti-example: NATIONALUM (bought +4.7% at midday = dead money,
then panic-squared a runner that closed +4.6%). Full rule: [[early-leader-catch-small-and-hold]].

## Owner correction (the day's real lesson)
"You should have held when it's small, not after it became high." Confirmed by data:
the day's top closers (HFCL +4.8%, NATIONALUM +4.6%, AUROPHARMA +3.8%) all HELD near
their highs — winners declare early and CONTINUE. My NATIONALUM trade broke this twice:
bought it late (+4.7%, chasing) AND squared it into a wobble. The stock won; my trade
didn't. Fix baked in: catch early/small + hold (Tier 2 above) + manage-the-winner (Tier 3).
**TIER 3 — MANAGEMENT OVERLAYS (all trades):** manage-the-winner (BE+ at +0.5R, take
profit on failed-breakout/book-flip, no target behind a wall); order-flow as an EXIT
sensor (book flips seller-heavy → exit); daily Console reconciliation + retro every session.
**TIER 4 — RESEARCH ONLY (not live):** continuation study — does buying late-day
strength pay? Multi-month backtest before any live trade.
**DROPPED:** last-30-min fade/short — study killed it. Do not trade.