How NexusTrade sources backtest data and checks pricing records

TL;DR

The January 2 to 3, 2024 SPY call run retained one observed-quote pricing, no synthetic pricings and 27 quote accesses with unknown timing. Those counters do not establish missing vendor quotes or bad accepted fills. Check the vendor's documented dataset, the application's retained contract/session rows and the run's pricing policy separately. ETF holding effective dates also differ from processing dates; the SPY/CZR sample needs an explicit availability convention. This page does not establish a product data-quality winner or complete quote coverage.

A holdings record has two dates worth checking

The current ETF Global constituents example contains SPY holding CZR, with effective_date September 18, 2025 and processed_date September 19. This is the vendor documentation’s sample, rather than a response fetched for this audit.

Using September 18 as the availability date would erase the later processing clock. For a conservative date-only worksheet, make the sample usable on the day after its processed date. That explicit convention permits it on September 20, not September 18 or 19. It is a worksheet policy, not a recovered issuer publication timestamp.

from datetime import date, timedelta

effective = date(2025, 9, 18)
processed = date(2025, 9, 19)
usable = processed + timedelta(days=1)  # Chosen date-only convention.

for decision in (effective, processed, usable):
    eligible = decision >= effective and decision >= usable
    print(decision.isoformat(), eligible)

assert effective < usable
assert usable == date(2025, 9, 20)

Inspect the SPY option’s pricing record

NexusTrade’s market-data pipeline integrates Massive (formerly Polygon.io), including its options quote files. The disclosures here explain what a particular run establishes; they do not by themselves establish inaccurate or low-quality Massive data, or a competitor’s superior data. The completed SPY option example evaluated January 2 to January 3, 2024. It selected SPY240202C00472500 and bought one call at 09:32 EST, after a 09:31 decision. The run finished with that call still open and retained a warning that quote coverage was not certified for strategy comparison.

The unknown quote-timing counter records quote accesses with unknown price-availability timing, rather than distinct source rows. Those 27 accesses do not establish 27 missing vendor quotes or accepted trades with bad fills. The pricing contract version is not a count of observed quotes. The saved result does not independently audit every upstream quote or certify a comparison with another platform.

Match the question to the evidence that answers it

Massive’s options quote documentation lists historical contract bid/ask quotes from March 7, 2022, with access depending on the options plan. January 2024 falls within that documented range. The page does not prove which quotes the application fetched or retained for the held contract.

Its OPRA quote-file documentation describes top-of-book bid/ask data with nanosecond timestamps across US options markets, including CBOE, NYSE and NASDAQ. The vendor’s dataset coverage is distinct from the rows retained by an application.

Its daily options aggregate files have a separate historical range starting June 2, 2014. Daily OHLCV bars and timestamped bid/ask quotes serve different pricing questions. Preserve the dataset name when stating the range.

Name the upstream source for the dataset you are inspecting

Massive’s source descriptions identify stock feeds from the CTA and UTP securities information processors, plus FINRA off-exchange prints; options from OPRA; and crypto from Binance, Bitfinex, Bitstamp, Coinbase and Kraken. These are vendor provenance statements. An application’s entitlement, ingested exchanges and selected run inputs need their own evidence.

The historical dividend endpoint retains declaration, ex-dividend, record and payment dates, amounts and adjustment fields. Its existence does not recover the version available at an old decision or identify an upstream dividend supplier. Keep the original source response and availability policy when using those fields.

Keep ETF holdings and official index membership separate

ETF Global’s documented holdings history starts April 3, 2017 and uses a separate paid expansion plan. It describes ETF holdings, rather than a dedicated official S&P 500 membership history. Fund holdings, weights and reporting delays must retain their own interpretation.

The original five-dividend-growers study begins in 2014. This dataset’s stated range alone cannot fill that earlier universe audit. Use the Tesla membership and Target dividend worksheet to inspect effective membership, public availability and the actual dividend clock before interpreting the original study.

Audit the same contract before comparing data quality

Freeze the requested contract, sessions, price fields, source version and decision/fill clocks. Recover the complete source response and retained consumer rows, then rerun the unchanged request only after verifying the data effect. Keep the original result and warning beside the new receipt.

Missing, stale or incorrect marks can be data-quality defects when established by source and consumer evidence. A certification warning alone leaves that finding unresolved. For a competitor comparison, request the same instruments, period, capital, sizing, costs and execution policy. Compare the returned curve, holdings and fills under that contract. The examples here identify the checks and the unresolved evidence; they do not assign a product winner from a warning alone.

Continue exploring