NSE Large-Cap Turn-of-the-Month Window: Long an Equal-Weight Basket of the 37 Names That Were NIFTY 50 Members in January 2018, From the Open of the Month's Last Session to the Open of the 4th New-Month Session (INEQ 1-DAY, calendar-only, long/flat, unlevered)

Original Research Lead hypothesis (not seeded). It tests the turn-of-the-month (ToM) flow effect (Ariel 1987; Lakonishok & Smidt 1988; McConnell & Xu 2008) on the Indian cash market, venue INEQ, which no experiment in the corpus has used. The window rule is the one already promoted on CAEQ (TSX 16-name basket: 1,664 trades, PF 1.35, Sharpe 0.57, DD 9.4%) and on SPY. Per lesson L263 it is long/flat with fixed entry and exit sessions and has no short arm. India has a specific flow reason to expect the effect. Salaries are paid at month end, and domestic mutual-fund SIP (systematic investment plan) debits are heavily concentrated in the first days of the month. These SIP flows are now a structural monthly bid of tens of thousands of crore rupees, and the funds deploy that cash into large caps. UNIVERSE (fixed ex ante, no later-member look-ahead): the 37 names that were NIFTY 50 constituents in January 2018 and have INEQ daily bars. 2018 members we hold no bars for are absent: YESBANK, VEDL, ZEEL, INDUSINDBK, DRREDDY, HINDPETRO, UPL, IBULHSGFIN, TATAMOTORS (now split into TMPV/TMCV), AUROPHARMA, LUPIN and INFRATEL. That leaves residual survivorship bias, which the analyst should weigh. YESBANK collapsed in 2020, which is the main omission, and it would have hurt. No 2019-2026 entrant (ADANIENT, TRENT, BEL, JIOFIN, ...) is included. PRE-STUDY (Research Lead, raw open-to-open equal-weight basket returns on these names, Feb 2018 to Oct 2026, before costs): 104 month windows, mean 0.552% per window, median 0.81%, t = 2.66, 62% positive. That compares with 0.216% for an average 4-session open-to-open span. The year means were 2018 -0.74%, 2019 -0.01%, 2020 +1.41%, 2021 +1.92%, 2022 +0.29%, 2023 +1.17%, 2024 +0.65%, 2025 +0.14% and 2026 YTD -0.15%. The result is not knife-edge to the window. Every entry offset of 1-3 sessions before month end combined with an exit at the open of new-month sessions 2-6 gave t between 2.27 and 3.31 (for example [-2, open of session 3] 0.60%, t 3.11; [-3, open of session 4] 0.76%, t 3.15). The baseline stays the canonical cross-venue window so it is not chosen from the scan. COSTS (INEQ venue_rules): 0.1187% on buys and 0.1055% on sells, plus about 3 bp half-spread per side, so about 0.284% per round trip. Expected net per window is 0.552 - 0.284 ≈ 0.27% of notional. That is thin, about 1.9x cost. Lesson L238 asks for distinct periods to be counted: there are about 104 distinct entry dates, and the roughly 3,800 per-name trades overstate the evidence. EXECUTION: daily bars; decisions on bar N's close fill at bar N+1's open (next_open_fill, supported on session venues). Session boundaries come from the platform's INEQ calendar via src.strategies.helpers.sessions and FactoryStrategy.session_* helpers, never hard-coded dates. Any special session the calendar lists (for example Diwali Muhurat trading) counts as a session like any other. Account and metrics are in INR (USD view in fx_report). CASH account, long only, no leverage.