Methodology

Every instrument is read through several independent lenses. Each lens scores 0 to 100 from its own sub-scores, and the composite is the weighted mean of the lenses that could actually be computed for that instrument, renormalised over them. A lens with no data does not score zero — it is absent, and the missing weight is redistributed. A fund is not bad at something nobody measured.

Every weight, band and cut point behind those numbers is a row in the thresholds table, not a constant in code. Change a row and the next run scores differently, with no deployment. That is why this page describes the rules and the page above it shows the values: there is one source of truth and it is the database.

3 of 5 lenses have a producer: technical, cost & liquidity, risk. Every score on this board carries its own n of 5 lenses label, so a one-lens composite can never be mistaken for a complete one.

The rest are not estimated, approximated or filled in. They are absent until their data is ingested, and the score says so. The sources are identified and free — company financials, corporate events, insider filings and fund holdings from public filings — but data that has not been ingested is not data, and a number invented to fill a column is the one failure this board is built to prevent.

A gold-miners fund, a Treasury fund and an S&P tracker have nothing to say to one another. Ranking them in one table produces an ordering that describes no decision anyone would make. So every fund is ranked inside its peer group: its asset class crossed with what it actually does — a sector fund against sector funds, a country fund against country funds.

A group too small to support a distribution cannot support a percentile either: ten funds cannot have deciles. Those funds fall back to their broader asset class, and each row records which grouping was used, so “fourth of thirty-four sector funds” is never confused with “fourth of two hundred equity funds”.

Within a peer group the composites are cut into ten equal bands. Decile 10 is the strongest tenth of that group and Leader means exactly that — top decile among its own peers, never top decile of everything. Deciles are computed when the page is read, never stored: a rank is a statement about a population on a date, and a stored rank goes stale the moment a score moves.

The conviction tier is a second, stricter reading: it asks not only how high the composite is but how many independent lenses agree. The top tiers require several lenses to be present, so while only 3 of 5 has a producer, no instrument can reach them however strong it looks. That is the methodology working, not a defect — conviction means agreement, and there is only one opinion in the room today.

Every figure labelled RS compares an instrument to the S&P 500 in the form (1 + r) ÷ (1 + r_SPY) − 1. +8% means it beat the index by eight percent over that window — not that it rose eight percent. A reader who takes the second meaning has learned something false, which is why the form is stated wherever the number appears.

Geared and inverse funds are classified and listed but never scored. They answer a different question from “what should I hold”, and a two-times daily fund topping a ranking is the kind of wrong nobody notices until after they have bought it. Funds below the liquidity floor are excluded for a plainer reason: a fund you cannot get out of is not an instrument you should be shown as an option.

Nothing excluded is hidden. Every list can show the full set with the reason each row is out, because “we chose not to rank this” and “this does not exist” are different statements.

A fund’s peer group comes from its own registered name, read by an ordered set of rules. An issuer states what a fund does in its title and is not free to misdescribe it, so the title is evidence rather than a guess — and every classification records the rule that fired and the words it matched, so any grouping can be checked.

Where no rule reads a name, the fund is marked Unclassified and shown as such. It is not quietly filed under a broad bucket: one mis-filed fund corrupts the peer group it lands in, and the whole point of the group is that its members are comparable.

The current weights are seeds — reasonable starting points, not conclusions. A lens keeps its weight only once its signal has been measured against forward returns on real history; anything that fails to clear the floor goes to zero weight and becomes an overlay until it earns one. Until that work is done this board states what it measures and how, and does not claim the ordering has been validated.

Risk is computed on that basis today: volatility, drawdown, downside deviation and beta are shown beside the score rather than inside it. Folding them into a composite makes a dull instrument outrank a strong one, which is not what “what should I buy” asks.