ETF input versions of the model library

Registered for the owner’s approved ETF route before the first simulation. Entry date: 17 October 2026. No parameters are selected using the resulting returns. Six additional trials count in the library’s multiple-testing total.

Inputs and identity

Use official iShares full-workbook NAVs, splits and ex-date distributions: IVV for US large-cap equity and ACWI for developed and emerging global equity. ACWX is prepared as an optional ex-US input but has no strategy registered in this amendment. ivv_nav_inr is separate from the older ivv_inr sensitivity, which includes a screener extension. New inputs never bridge from a screener YTD, an index history or observations before fund inception.

NAV total return chains (NAV + ex-distribution) * split / prior NAV. FX uses the last available ECB INR/USD reference of the US date, at most four calendar days old. An NSE decision sees only US observations dated strictly before it. The latest model session ends when the last actual workbook observation first becomes available on the NSE calendar; stale tails are not logged as returns.

Rules

For each of IVV and ACWI:

• trend_world_<fund>: the registered Nifty-100 six-sleeve trend rule, replacing the foreign sleeve. Weekly decisions and a 5-percentage-point trade band.
• gtaa_<fund>: four equal sleeves in Nifty 500, the ETF, gold and 5-year G-secs; each invested above its 10-month average, otherwise cash.
• gem_<fund>: the registered India-absolute dual-momentum rule, comparing Nifty 500 with the ETF over 12 months; G-secs when India does not beat cash.

Use the exact same asset mix as a static annual-rebalanced counterpart over identical dates. ACWI includes India; this is an overlapping global exposure, not an ex-India benchmark. Keep old model codes, signals and frozen rows intact. None of these variants is a new published Tipsheet index.

Warm-up, costs and interpretation

Wait for every required signal window; start no earlier than 1 April 2005. No leading values are used to form a signal. IVV’s earlier inception does not extend the domestic lab’s headline period. ACWI’s shorter history also needs 10 completed observations for GTAA, 13 for momentum, or the full 24-vote trend warm-up. Trades follow decisions at the next close as in the lab engine.

Observed NAV already reflects fund expenses. Additional annual fund drag is zero; assumed remittance, forex and brokerage are 0.60% of value each way. NAV is a research valuation proxy, not an executable quote. Distribution cash is assumed reinvested on ex-date, before investor withholding/income taxes. Report before investor tax only. Do not treat total-return appreciation as pure capital gains or manufacture an after-tax CAGR. A payment-date cash and tax model needs a separate implementation before after-tax publication.

Issuer histories have small rounding-scale differences against monthly return sheets, recorded in docs/reference/international-etf-sources.md; the strict checks remain flagged. Public data-use permission is not inferred from the owner’s research-input selection. No deployment is included here.

First local results

Full library run: 52 models, data through 14 October 2026; each IVV version ends 13 October because its full workbook stops on the US date 12 October. ACWI versions end 14 October on the local input calendar. All six ETF versions trail their static counterpart in annualised return before investor tax over their own samples. Several have substantially smaller daily falls. This is a descriptive result, not evidence for a tuned replacement rule.

ACWI GTAA starts 1 January 2009; ACWI momentum starts 1 April 2009 and trend starts 20 April 2009. The latter two miss the 2008 crash and March 2009 trough; a small measured worst fall cannot be compared with a 2005-start record as if they had faced the same crises. IVV GTAA and momentum start 1 April 2005; IVV trend starts 10 April 2006. See audits/models/etf-results.json for the exact output, benchmarks and run timings. All 319 pre-existing frozen-log row hashes are preserved, and there are no manufactured after-tax results.