Methodology

Every ranking on this site is produced by the same pipeline, with no discretionary step anywhere in it. This page describes that pipeline completely. If something here doesn't match what you see on a screen page, the page is wrong and we want to know.

1. Universe

We track a fixed list of 53 US-listed ETFs spanning broad market, factor, sector, international, fixed income, and commodity exposures. The list changes rarely, and changes are noted in the week they take effect. Each screen draws from a subset of that list — sector screens see only sector funds, the defensive screen sees only bonds and commodities, and so on.

2. Price data

We pull three years of daily split- and dividend-adjusted closing prices for each fund. A run that fails to retrieve at least 80% of the universe is aborted rather than published, so a partial data outage produces no ranking instead of a misleading one.

3. Metrics

Each fund gets the same set of measurements, all computed from adjusted closes:

• Trailing returns over 1, 3, 6, and 12 months (21, 63, 126, and 252 trading days).
• 12-1 momentum — the 12-month return excluding the most recent month, the standard construction that avoids short-term reversal.
• Volatility — the standard deviation of daily returns over the trailing year, annualized by √252.
• Maximum drawdown — the worst peak-to-trough decline over the trailing year, expressed as a negative number.
• Trend — the percentage distance between the latest close and the 200-day simple moving average.
• Return-to-risk — trailing 1-year return divided by annualized volatility. This is not a Sharpe ratio; no risk-free rate is subtracted.
• Liquidity — the median daily dollar volume over the last quarter, used as an eligibility gate rather than a scoring factor.
• Expense ratio — from a maintained table of issuer-published figures.

4. Eligibility gates

Gates run before scoring. A fund that fails any gate is removed from that week's ranking entirely and listed in the exclusions, with the gate it failed. Gates are hard filters, not penalties — they never partially reduce a score.

5. Scoring

For each factor, every eligible fund's raw value is converted to a cross-sectional z-score — how many standard deviations it sits from the mean of the funds it is being ranked against, that week. Factors where lower is better (volatility, expense ratio) have their sign flipped. The z-scores are multiplied by their published weights and summed into a composite. Funds are ranked by that composite, descending.

Two consequences worth understanding. First, a score is relative: it says where a fund stands against its peers this week, not whether it is objectively good. Second, scores are not comparable across screens or across weeks, because the peer group and the distribution both change.

6. The screens

7. What this cannot tell you

These screens read price history. They do not read holdings, fund flows, index construction, tax treatment, tracking error, bid-ask spreads, or anything about your own situation — your time horizon, tax position, existing exposures, or how much loss you can actually tolerate. A fund at rank 1 is the fund that scored highest on a specific arithmetic formula last week. That is all it is.

Momentum-style factors in particular have a well-documented failure mode: they perform poorly through sharp reversals, and they tend to concentrate into whatever recently went up. A high rank during a late-stage run is exactly what you would expect to see just before a reversal. The screen cannot distinguish those cases and does not try to.

8. Corrections

Data errors happen — bad ticks, stale adjustments, wrong expense ratios. When we find one that changed a published ranking, we correct the page and say so on it. Expense ratios are periodically re-verified against issuer fund pages.