How an asset gets scored.
Seven dimensions, one weighted sum, and the weights published in full. Nothing here is a black box, and where the system has not yet run against real money we say so rather than describing it in the present tense.
Last updated
What a score is.
Every dimension below is a percentile rank against peers, not an absolute rating. An asset scores 0.80 on quality because it ranks in the eightieth percentile of comparable companies on the quality inputs, not because it cleared some fixed bar.
Peers are companies in the same sector and market-capitalisation tier. Where a sector is too thin for a meaningful comparison, the system falls back to a broader sector grouping and records how many peers the comparison actually used.
Scores run from 0.0 to 1.0. We state that explicitly because an earlier generation of this system stored the same column names on a 0 to 10 scale, and a page that quoted the wrong range would misrepresent every figure a customer sees.
The seven dimensions.
Each is computed independently, from the inputs named. Sector conventions are respected where a metric does not apply: return on invested capital is undefined for financial companies, for instance, so for those the quality weighting is redistributed across the remaining inputs rather than scored as a zero.
Quality
Business quality. Return on invested capital, return on equity, net margin, operating margin, and interest coverage.
Risk
How much volatility and drawdown the asset has historically carried. Weighted most heavily in the conservative strategy and almost out entirely in the aggressive one.
Value
Valuation cheapness: free cash flow yield, EV/EBITDA, price/earnings and PEG. A lower ratio ranks higher, because cheaper is the thing being measured.
Growth
Rate of expansion in the underlying business rather than in the share price.
Momentum
Recent price performance relative to the asset's own history and to its peers.
Sentiment
Published news, scored by a financial-language model and aggregated per ticker. This is the one dimension fed by a separate service, and it is range-checked on the way in: a value outside 0.0 to 1.0 is discarded rather than clamped, so a broken upstream feed reads as missing data instead of quietly blending a wrong number into the composite. That has happened before, which is why the check exists.
Analyst consensus
Published analyst ratings and targets. Carries the smallest weight in every strategy, and zero in four of the seven.
The published weights.
A strategy is a weighting profile, not a fixed list of holdings. Choosing one changes how much each dimension counts toward an asset's fit, and nothing else. These are the actual weights the scoring service applies.
| Strategy | Quality | Risk | Value | Growth | Momentum | Sentiment | Analyst |
|---|---|---|---|---|---|---|---|
| Conservative | 0.35 | 0.30 | 0.15 | 0.05 | 0.05 | 0.05 | 0.05 |
| Blend | 0.20 | 0.15 | 0.20 | 0.15 | 0.15 | 0.10 | 0.05 |
| Aggressive | 0.10 | 0.05 | 0.10 | 0.25 | 0.30 | 0.15 | 0.05 |
| Value | 0.25 | 0.15 | 0.45 | 0.05 | 0.05 | 0.05 | 0.00 |
| Quality | 0.45 | 0.20 | 0.20 | 0.05 | 0.05 | 0.05 | 0.00 |
| Opportunistic | 0.05 | 0.05 | 0.10 | 0.25 | 0.45 | 0.10 | 0.00 |
| Core | 0.25 | 0.20 | 0.20 | 0.10 | 0.10 | 0.10 | 0.05 |
Each row sums to 1.00. The heaviest weight in each strategy is emphasised: it is the single fact that most determines what that strategy will pick.
Missing data.
Not every company reports every input. Rather than treat an absent metric as a zero, which would rank a company badly for not filing rather than for performing badly, the system tracks how complete each asset's data actually is and carries that through to the final fit.
An asset with thin coverage is not silently promoted on the strength of the two metrics it does have.
What this method does not do.
- It does not predict returns. A percentile rank is a statement about how an asset compares to its peers today on measurable inputs. It is not a forecast, and a high score is not a prediction that the price will rise.
- It does not learn from your account. The weights are fixed and published above. They do not adapt to how your particular portfolio has performed, and no model is trained on your holdings.
- It does not model correlation between your holdings. Each asset is scored on its own merits against its peers. The scoring stage does not measure how a candidate would interact with what you already own.
- It has not yet run against a live account. Everything here describes what the scoring service is built to do, read from the code that does it. When it has run for real, we will publish what it actually did.
The decisions that follow from these scores, and the limits they are checked against, are described on how it works. What we record about each decision afterwards is on transparency.