How to evaluate a hybrid growth and money-flow scan

A growth screen, a money-flow timing screen, and a hybrid of both can be compared only if each scan date is treated as a sealed test window. The archive workflow collected four dated market scans, one in 2010, one in 2011, and two in 2012, then measured each selected stock’s subsequent absolute and market-relative return at regular intervals through the next two months.

Keep every scan date out of sample

Because the scanner could not backtest in place, historical scans were reconstructed from dated watchlists on machines frozen at earlier dates. The lists were then exported and scored only after the scan date so the evaluation window stayed out of sample.

That reconstruction is what makes the later statistical tests interpretable. If names are scored with information that arrived after the scan, a Random walk hypothesis test, a Chi-square test, and Monte Carlo simulation are no longer tests of the published rules.

Compare three rule sets on one protocol

The growth screen required 90-day dollar volume above 387,000, year-over-year quarterly earnings growth above 14.5%, trailing-four-quarter earnings growth above 11%, year-over-year quarterly revenue growth above 14.7%, trailing-four-quarter revenue growth above 12.3%, return on equity above 12%, long-term debt no more than twice common equity, a four-quarter net profit margin above 11.2%, and a price-to-sales ratio no higher than 5.

The technical screen reused the same liquidity floor and required MoneyStream near a three-month high, Time Segmented Volume near the top of its 40-day range, and price no higher than 96% of the 52-week high.

The same four-date, two-month-forward protocol was used to compare the fundamental screen, the technical screen, and a hybrid that applied both sets of conditions. Differences could then be attributed to the rule set rather than to a change in sampling interval or lookback.

Choose a market baseline that a few large names cannot dominate

Relative performance was scored against the S&P equal-weighted index rather than the capitalization-weighted S&P 500 so that a few large names would not dominate the market baseline. Editorial reading: that choice keeps the two-month comparison aligned with a broad list of selected names rather than with the largest capitalizations in the cap-weighted index.

Test whether the hybrid adds information

The statistical evaluation treated a random-walk null as the explicit baseline. If successive returns are unpredictable, dated screens should not systematically beat the equal-weighted market over the fixed two-month horizon. That is the Random walk hypothesis test against which the three rule sets are judged.

Chi-square test counts of outperform versus underperform outcomes across scan dates and market regimes were the planned test of whether observed win frequencies departed from the random-walk expectation.

Monte Carlo simulation, implemented as resampling of the same ordered price, volume, and breadth series, was specified as a robustness check on whether hybrid-versus-single-method differences survived a change in path rather than a single historical sequence.

Editorial reading: use the three methods in that order. First ask whether dated screens beat the equal-weighted market under the Random walk hypothesis test. Then use the Chi-square test on outperform versus underperform counts. Finally use Monte Carlo simulation to check whether hybrid-versus-single-method differences persist when the path of the ordered series changes. The archive describes that workflow. It does not establish that any screen is profitable.