How the funnel works
How this site reads its own numbers: seven conditions applied to every covered name, ordered so the ones with a track record are applied before the ones without. Exposure is considered first; individual names last.
This page explains the procedure. The Funnel tab under Tools → applies it across the whole book, name by name. Every condition below carries the strength of the evidence behind it, and the numbers here are computed from the published data each time the page loads — never typed.
What this is not. A name meeting every condition is a description of what the published figures say about it against a stated rule — not a recommendation, not a rating, and not a suggestion to buy or sell anything. The conditions are filters over published data; they carry no view on whether any name is a good investment, and the largest of them has no track record at all. Educational analysis, not investment advice — and the preparer is not licensed by Egypt's FRA. Full disclaimer.
What the coloured rail means
Zone 1 — set the exposure
Exposure is considered before any individual name
provenThe procedure settles total equity exposure first, against the book's own measured worst case, before any name is looked at. On this record, being in the market at a size that survives its own drawdowns accounted for more than every selection condition below put together.
RULE Exposure judged against the median published 3-month p5, not against a target return.
Zone 2 — narrow the universe
Keep only what can be read at all
prerequisiteDrop a name only where an input does not exist: no fair value, no computable momentum, no band record. Nothing is dropped for being old or for being work in progress. Prices are entered by hand on a monthly rhythm and a study rebuild takes days, so both are printed on the row and neither ends the name — a tool that dead-ends everything it has touched answers nothing on the day you ask it.
RULE Stop only on a missing input. Age and rebuild-in-progress are flags.
Cap each name from its own band record
provenSet the maximum before forming a view, so the risk budget cannot be argued backwards from conviction. A name whose own 90% band has historically under-covered gets a smaller cap than one whose bands held.
RULE A name flagged "bands ran narrow" sizes down. Caps, cuts nothing.
Order by the gap — and read the two sides differently
ungradedA fair value well above the price is the ordinary shape of a name the fundamental lens reads as cheap, and the condition records that. A fair value well below the price is a high-prior-of-defect region — historically our own model missing something rather than a short — so those go to an investigation queue, never to a position.
RULE Condition met at a gap above +10% against the latest price held. Below −10% the study is audited rather than acted on.
FLAG A name whose study is being rebuilt is evaluated on the published fair value — the only number a reader can act on — and its row says so.
Use momentum as a veto, not a ranking
thinThe lean is right about 53 times in 100 against a base rate near 51 — enough to refuse a name, not enough to order a list. It is not used to rank: the cross-sectional version is refuted, and in a tape where most names print up, the sign is nearly free.
RULE A strong negative call fails the condition. It only ever removes; it never supports a position.
All three lenses must point the same way
ungradedThe fundamental, momentum and technical reads are built independently, so agreement is evidence rather than an echo — but tested as a stock picker it earned nothing. Its job is timing rather than selection: two of three is recorded as disagreement, not as a weaker version of agreement.
RULE Condition met only where all three lenses point the same way. It speaks to when, never to which.
Zone 3 — act, then loop
Enter and exit off the touch ladder
provenThe bands are calibrated, so a published level is a probability rather than a chart line. The procedure reads entries and exits off that ladder and controls risk by size rather than by stops — on this record a stop tighter than the typical drawdown fires on ordinary movement.
RULE Nothing above the name's own p95. Use size, not stops.
How often a level is actually reached inside three months, by how far it sits from the price — every published rung across the covered book:
Rebalance quarterly and grade every decision
provenMonthly rebalancing measured net-negative: the round trip is larger than the signal. Each quarter the gap and direction call are recorded against what actually happened — that record is the only thing that can move the gap condition from hypothesis to evidence.
RULE Quarterly or not at all. Re-run Gates 1–5 on the current state each quarter.
Tested, and it did not work
- Monthly rebalancing. Every one-month construction is net-negative after the round trip. The commission is larger than the signal.
- Sizing by cone width. Inverse-volatility weighting did not beat equal weight; the names co-move too hard for it to bite.
- Trading the centre of the cone. The centre is the carry — the same for every name. Reading it as a target is reading the risk-free rate as a forecast.
- Momentum as a ranking. Sign-unstable across samples. It survives as a veto and nothing more.
- Lens agreement as a stock picker. Zero excess over the panel. It is a timing gate, not a selector.
What actually earns the money
Gates 0, 2, 6 and 7 do most of the work and all of them are proven — exposure at a survivable size, a cap from each name's own record, execution off the ladder, and a quarterly clock. None of them selects a name. Gate 4 is a small measured filter. Gates 3 and 5 are where the upside lives and neither has a graded record, so they narrow and time, and never support a size that Gates 0 and 2 had not already allowed.
Nothing here is a rating or a price target. The procedure produces fair-value ranges, distributions, and a decision about size. Every figure on this page is computed from the published data at load — see the live table for the current state of every name, and Method for how the numbers underneath are built.