Computed
Performance, volatility, Sharpe, Sortino, beta, maximum drawdown, tracking error and similarity. Calculated by TradeAlphaAI from observed daily closing prices, refreshed on every build.
Methodology
ETF Score Methodology
How the TradeAlpha ETF Score is calculated: every component, its weight, the inputs behind it, the assumptions stated openly and the limits acknowledged.
What this measures
The TradeAlpha Score
The TradeAlpha Score describes the structural quality of a fund — what it costs to own, how broadly it is spread, how easily it trades, how closely it follows its index, and how it is structured. It does not describe what a fund is going to do next, and it is not a ranking of expected return.
Every component is computed from inputs this platform can evidence. Where an input does not exist, the component is marked indeterminate, dropped from the total, and the remaining weights are rebalanced so the result is never quietly padded with an assumption.
Funds covered39
Funds scored39
Verified fields195
Awaiting verified data287
Coverage floor50%of model weight required
Components
Weights and inputs
| Component | Weight | What it measures | Where the input comes from |
|---|---|---|---|
| Cost | 25% | The published total expense ratio, mapped across the 0.03%–0.75% range that index-tracking funds occupy. Cheaper scores higher. | Requires a verified expense ratio. No free provider publishes one today, so this component is currently indeterminate for every fund and its weight is redistributed. |
| Diversification | 20% | How broad the mandate is, reduced where the fund has actually behaved far more volatilely than the broad market — a concentrated fund is concentrated regardless of how its mandate reads. | Mandate breadth tier plus observed volatility versus a broad-market reference. |
| Long-horizon composite | 20% | Combines the cost and diversification components with the length of observable history and the depth of the worst recorded drawdown. | Derived from the other components plus observed price history. |
| Liquidity | 15% | Median daily turnover over the trailing year, on a logarithmic scale spanning 1M to 10B units of the listing currency. | Computed from observed exchange volume and closing prices. |
| Tracking quality | 10% | Annualised tracking error against a same-currency proxy for the index the fund states it tracks. Scored only where such a proxy exists, and never against the fund itself. | Computed from observed prices; indeterminate for most funds, which is stated openly on each page. |
| Structural tax characteristics | 10% | Fund domicile and distribution policy — the two structural features that change withholding treatment most. | Distribution policy read from the provider fund name where it states one; fund domicile is awaiting a verified source. |
Bands
How the label is assigned
| Label | Overall score |
|---|---|
| exceptional | 85 – 100 |
| strong | 72 – 84.9 |
| solid | 58 – 71.9 |
| adequate | 42 – 57.9 |
| limited | 0 – 41.9 |
| indeterminate | below the coverage floor |
Data policy
Where every number comes from
Every figure on an ETF page belongs to one of three tiers.
Verified
Fund name, trading currency, exchange and first trading date, returned by a data provider and stored with the hash of that response. Issuer and distribution policy are read from the verified fund name, which is quoted alongside them.
Awaiting verified data
Expense ratio, ISIN, domicile, replication method, fund size and inception date have no free verifiable source. They are shown with that status on every surface rather than estimated, and the fund still appears in every ranking and table.
No value anywhere in this section is asserted from prior knowledge. Every field is traceable to a provider response, a computation over verified prices, or a labelled TradeAlphaAI classification — and a validator rejects the build if any factual field lacks one of those. Connecting a verified data source will populate the awaiting fields automatically, without changing a single page.
Currency
Why some comparisons are withheld
Returns are stated in each fund's own trading currency. Benchmark-relative figures, beta and tracking error are shown only where the fund and the benchmark share a currency, because a cross-currency comparison would measure the exchange rate as well as the fund.
This is why a European fund shows its returns but not a benchmark-relative figure: comparing a euro-quoted fund against a dollar-quoted index would report the exchange-rate move as if the fund had produced it.
Stated assumption
The tax perspective used
Structural tax characteristics are assessed from the perspective of a non-US investor, because domicile and distribution policy change withholding treatment most for that audience. It is not tax advice and does not model any individual circumstances.
Tax treatment depends on where an individual is resident and on rules that change. The component describes fund structure only; it cannot and does not model any individual situation.
Limits
What this does not cover
- Tracking quality is indeterminate for most funds, because a same-currency proxy for their index is not available from free data.
- Holdings and country weights are not yet published for every fund. Where they are absent the section is omitted rather than estimated.
- Fund size is a point-in-time figure shown with its date, not a live number.
- Turnover uses adjusted closing prices, so it ranks relative liquidity rather than stating an exact traded amount.
- Historical measurement describes what has already happened. It carries no implication about what follows.
Reproducibility
Checking the numbers
Each component on a fund page lists the evidence it used — the observation count, the window, the source document. Because the weights and bands on this page are published, any score shown on the site can be recomputed from those figures.
TradeAlphaAI ETF intelligence describes observed structure, cost and historical measurement only. It is not a trading signal, execution instruction or investment advice.