# Raj TLH Deep-Dive — Sunday, October 11, 2026

**Status: BLOCKED — partial data.** Read-only per raj-research-doctrine; no orders placed, no mandate touched. Aggressive TLH remains Raj's standing strategy; this review could not be executed this week because the required inputs were not retrievable. Recovery was exhausted this session (every read path re-attempted multiple times); where evidence is missing it is reported as NOT PRODUCED rather than guessed.

## Data retrieval log (this session)

| Input | Attempts | Result |
|---|---|---|
| `briefs/snapshot.json` (positions, avg costs, live quotes, realized P&L, options book) | 8 reads (read_document ×4, read_file ×4) | `[cleared]` on every attempt — not retrieved |
| `scripts/research_index.md` (thesis board) | 7 reads (read_document ×4, read_file ×3) | `[cleared]` on every attempt — not retrieved |
| `briefs/latest.md` (data floor) | 6 reads | `[cleared]` on every attempt — not retrieved |
| `briefs/latest-tlh.md` (prior review) | 3 reads | `[cleared]` on every attempt — not retrieved |
| Broker surface (positions / account / orders / connections) | 7 calls | No rows; MCP timeouts on robinhood-live-mcp — broker unreachable |
| `search_symbol` (23 tickers/names) | 23 queries | `[cleared]` on all — identity resolution not usable this session |
| `get_market_data` (Yahoo, 1D bars) | 4 batches | 1 batch returned rows (ORCL.US, PTY.US, SFTBY.US); 3 batches `[cleared]` |

## Observed market evidence (the only live marks retrieved this session)

Source: Yahoo Finance via `get_market_data`, interval 1D, quote currency **USD**, no currency conversion. Symbols below are the canonical symbols with venue suffix returned by that tool call in this session. Columns are the four most recent daily closes retrieved, ending with Friday's close (the last bar in each series).

| Symbol (venue) | Tue close (USD) | Wed close (USD) | Thu close (USD) | Fri close, last bar (USD) |
|---|---|---|---|---|
| ORCL.US | 144.77 | 143.56 | 135.69 | 141.40 |
| PTY.US | 10.89 | 10.76 | 10.94 | 10.97 |
| SFTBY.US | 19.62 | 19.80 | 17.83 | 18.04 |

Derived moves (labelled derived; formula shown with the observed closes above as inputs):
- ORCL.US, Wed close → Thu close: (135.69 − 143.56) / 143.56 = −7.87 / 143.56 = −5.48%.
- ORCL.US, Thu close → Fri close: (141.40 − 135.69) / 135.69 = +5.71 / 135.69 = +4.21% (partial rebound).

This is the sharpest two-day round-trip among the three retrieved names — the first place to look for a loss lot once cost basis is retrievable. No other symbol was handled by a successful tool call this session, so no other figure is asserted.

## Section 1 — Wash-sale ledger

**NOT PRODUCED.** The realized-P&L rows live in `snapshot.json` (tail), which returned `[cleared]` on every read; the broker executions surface timed out. With no sale/purchase timeline retrieved, no ±30-day pair can be identified, and nothing can be flagged or invalidated. Source returned no rows.

## Section 2 — Loss-lot inventory

**NOT PRODUCED.** Positions and average costs were not retrieved (snapshot `[cleared]`, broker timeouts). Unrealized loss $ / % cannot be computed without costs, so **no held position is asserted to carry an unrealized loss**. Holding period (LTCG vs STCG) remains unknown — not derivable from available data. The only marks available this session are the three observed close series above; they carry no cost basis.

## Section 3 — Harvest candidates ranked

**NOT PRODUCED.** The thesis board (`research_index.md`) was not retrieved, so no DEAD-thesis vs ACTIVE-thesis ranking is possible. A candidate list without thesis status would be a guess, not a ranking — withheld.

## Section 4 — Replacement structures

**NOT PRODUCED at position level** (no candidates identified). Standing review rule, reaffirmed for when data returns: a substantially-identical option position on the same underlying also triggers wash-sale rules, so any candidate evaluation must check option holdings per underlying before proposing a 30-day replacement. No specific structure is proposed this week.

## Section 5 — Proposals

**No harvest actions proposed this week.** The book is unreadable: positions, costs, realized P&L, and thesis status were all `[cleared]` or timed out. Proposing harvest trades on unseen positions would violate evidence rules and the read-only doctrine. Recovery gate: rerun this deep-dive once `snapshot.json`, `research_index.md`, and the broker reads return — or point me at a fresh export (positions with cost basis, plus realized P&L) and I will rebuild the wash-sale ledger and loss-lot inventory from that instead.

## Integrity note

Every figure in this file is a tool result from this session: the three observed close series above (Yahoo, USD) and the two derived percentages in the "Observed market evidence" section (formulas and inputs shown). No account total, position count, cost basis, realized P&L, or thesis verdict was retrieved this session, and none is asserted.

## Summary

- Data state: blocked — snapshot, thesis board, data floor, and prior TLH review all `[cleared]` on every read; broker MCP unreachable; only ORCL.US, PTY.US, SFTBY.US daily closes (Yahoo, USD, four bars ending Friday) were observed this session.
- Wash-sale ledger, loss-lot inventory, thesis-ranked candidates, and replacement structures could not be produced this week; **no harvest actions are proposed**; no position is asserted to carry an unrealized loss (no cost basis retrieved).
- Recovery gate: rerun once snapshot.json, research_index.md, and broker reads return; first place to look then is ORCL.US, whose sharpest single-day decline (−5.48%, derivation above) landed on the Thursday bar.