What the score cannot tell you

Read this one before you trust anything else on the site.

The Divergence Score is a first, reasonable, unvalidated weighting. It has not been back-tested. It is not investment advice. What follows is what it genuinely cannot do, written down so you do not have to discover it yourself.

It is a sorting mechanism, not a prediction

The score asks one question: do several independent signals point the same way while the price does not yet reflect it? That is a reason to look, not a reason to buy. Nothing in the method estimates a probability, a return, or a horizon, and any number that looks like one is a rule-derived level rather than a forecast.

A high score can rest on one signal

If a stock has no insider filings and no analyst coverage, those two pillars are Unknown and the remaining weight is redistributed. A stock can therefore score 86 and read "Strong" on the strength of one pillar out of three.

This is deliberate — scoring missing data as zero would punish exactly the under-covered small caps the product is built to find — but it means the tier alone is not enough. Always read pillarsUsed beside it: the score says how many signals it had, and one out of three deserves a different kind of attention than three out of three.

The weights are a judgement, not a finding

Insider conviction carries 45%, analyst valuation 30%, market behaviour 25%. Insider weighs most because an open-market purchase by someone inside the company is the hardest of the three to fake. That is a defensible argument. It is not a measured result, and we have not demonstrated that these weights beat any other set.

Analyst data is thin exactly where it matters

The stocks worth finding are the ones nobody covers. That is the premise — and it means the valuation pillar is weakest precisely on the stocks the product exists for. One or two analysts is barely a consensus. The rating still counts, at a reduced weight, with an amber dot to say so, but a thin signal downweighted is still a thin signal.

Absence of insider buying proves nothing

Most companies have no Form 4 purchases in any given window. "No insider buying" is the normal state of the world, not a negative signal. We treat it as Unknown rather than as bad news, and you should read it the same way.

Market behaviour measures liquidity, not direction

The market pillar scores how easily you could get in and out — daily turnover as a share of market value. It does not say the price is going up. A stock near its 52-week low can score well on this pillar. Whether that is a bargain or a falling knife is a judgement the score does not make for you.

The track record is short, and short records say little

Every outcome is measured from the first sweep a stock appeared in against the most recent close we have. It is honest arithmetic on a small number of days. Under 20 sweep days the page says so itself. A handful of days mostly measures how recently we started, and past moves say nothing about future ones.

Data can be wrong, and sometimes is

Prices arrive from a third party and occasionally arrive in the wrong currency or unit. We check for that in three ways and refuse to score a stock whose price we do not trust — see how the score is built — but the guarantee is only that we catch what those tests catch.

What we do not do at all

No position sizing. No time horizons. No "buy" or "sell". No personalised advice of any kind: everyone sees the same sweep and the same numbers. Those are deliberate omissions, not features we have not got to yet.