HOW TICKR THINKS.

Most quantitative signal services operate as black boxes: a score appears, a badge glows, and the user is asked to act on faith. We think that's backwards. A signal you cannot interrogate is indistinguishable from a guess.

This document explains how every system on Tickr thinks — in plain language, with honest limits. It deliberately does not explain how the systems are built: the specific measurements, weights, and cut-offs are the product of years of testing, and published edges decay when they are copied. That trade-off is stated here once, openly, instead of hidden.

Everything that follows describes pattern-matching against public data. Nothing here predicts the future; the systems measure conditions under which certain outcomes have historically been more likely, and say so out loud.

After each market close, the pipeline ingests several thousand U.S. stocks and refreshes the inputs the engines reason over:

Price and volume behavior — trend, momentum, how stretched a stock is from its own averages, how volatile it has been. Fundamentals — profitability and balance-sheet quality from SEC filings. Insider activity — actual open-market purchases by executives and directors, from Form 4 filings. Short interest — how heavily a stock is bet against. Analyst posture — where Wall Street's targets sit relative to price. Institutional ownership — the full equity books that the largest, most concentrated investors disclose each quarter. And the market itself — breadth, trend, and stress measures that describe the overall tape.

Every input is a measurement, not an opinion. The systems never read headlines, tweets, or message boards — we tested those sources and found no durable edge in them.

MACRO starts every evening by asking a question most stock screeners skip: what kind of market are we in? It classifies the tape into one of seven recurring market climates — grinding rallies, broad selloffs, volatility spikes, calm drifts, and so on. This matters because the same stock setup behaves very differently in different climates.

Only then does it grade each stock — across eight independent dimensions: macro fit, momentum, fundamentals, volatility posture, analyst positioning, insider flow, short interest, and a machine-learned composite. Think of them as eight separate skeptics who don't talk to each other.

A stock earns a tier only when several skeptics independently agree and the agreement matches a pattern that has worked in that market climate before. Most stocks, most days, convince no one — and are shown as a dash. That silence is the system working, not failing.

The second engine is a specialist: it watches for the three stressed climates — active selloffs, volatility spikes, and crash conditions — and does nothing at all the rest of the time.

When those conditions arrive, SECTOR looks for a specific historical pattern: quality names inside specific sectors being sold indiscriminately alongside everything else. It stacks its own sector-aware checks on top of MACRO's grades, with tighter risk gates, and fires only when both agree.

An empty SECTOR page during a calm market is the system honoring its own rules. We would rather show you nothing than manufacture confidence where the evidence doesn't exist.

Corporate insiders — CEOs, CFOs, directors — must report every share they buy of their own company to the SEC within two business days. The INSIDER engine reads those filings nightly and asks one question: is this purchase unusual enough to mean something?

Decades of academic work and our own testing agree on the rough shape of the answer: who is buying matters (an operating executive's buy carries different information than a fund's), how much matters (size relative to role, not headlines), and clustering matters most — several insiders buying the same stock within the same rolling window is a far rarer and historically stronger event than any single purchase.

The engine scores each event against those validated rules, discards the overwhelming majority, and surfaces the few that pass — both on its own page and as one of MACRO's eight skeptics.

Large institutional investors must disclose their full equity book to regulators every quarter. The HEDGE FUNDS engine tracks sixteen of the most concentrated and most studied managers, and reconstructs each portfolio — every new buy, add, trim, and exit — by comparing one quarter's holdings against the last.

Most of that activity is noise: a fund nudging a position, a manager rebalancing. The validated signal is narrow — the cluster: two or more of these managers independently initiating the same position in the same quarter. Independent agreement, not a single famous name, is what the history rewards, and those clusters are flagged on the dedicated HEDGE FUNDS page.

Like every system here, it earned its place the hard way. Cluster buys are the one pattern in this data that cleared the time-machine test — outperforming the market in the weeks after each disclosure, in every separate year we examined, the bad ones included. The funds' own selling, tested the same way, carried no edge — so we don't publish it.

Every surfaced name carries one of three labels, and they form a ladder of historical evidence strength:

WATCH — the pattern has a real, validated historical edge, but a modest one. Worth attention, not conviction. HIGHLIGHT — multiple independent confirmations align; historically a clearly stronger cohort. TOP PICK — the rarest label: broad agreement across the skeptics in a climate where that agreement has performed best.

Two honest footnotes. First, tiers are calibrated per index — a small-cap must clear a higher bar than a large-cap to earn the same label, because small-cap patterns are noisier. Second, the ladder describes historical frequencies, not promises: a TOP PICK can and does lose. What the label tells you is how strong the evidence stack was at the moment it was assigned — never an outcome.

Every rule on this site survived the same gauntlet before going live, and it has three gates.

Gate one: the time machine. Patterns are tested walk-forward — the model is repeatedly placed at a point in the past and allowed to see only the data that existed at that moment, then judged on what happened next. This kills the most seductive failure in quantitative research: rules that explain the past beautifully because they secretly peeked at it.

Gate two: repeat across years. An edge must show up in year after year of separate market conditions — not just in aggregate, where one spectacular stretch can hide four mediocre ones. A pattern that only worked in a single regime is treated as luck until proven otherwise.

Gate three: keep proving it. Shipping is not tenure. Live results are re-measured against the original evidence, and systems that decay get demoted or retired — publicly, since the track record page updates either way.

Every signal this site publishes is timestamped and scored against what the market actually did next. The PERFORMANCE page tracks the top tiers of each engine against the S&P 500 from launch, updated daily, with losses displayed at the same font size as wins.

The AI FUNDS page runs the same principle as a live experiment: two AI portfolio managers receive identical market data each evening and independently run rival $100,000 paper portfolios. Their trades, reasoning, and running scores against the index are published in full — including when they disagree with our own engines, and including when they're wrong.

We publish receipts because it imposes the right incentive on us: every rule we ship knows it will be graded in public.

They cannot see the future. Every edge here is a historical frequency. Markets change, and a pattern can stop working precisely because enough people found it.

They are most confident in rough markets. By design, the strongest historical edges sit in stressed climates — which are also the highest-variance moments to act in. Strong evidence and a comfortable ride rarely arrive together.

They will be wrong, on schedule. A tier with a strong historical win rate still loses regularly — that is what a win rate means. The systems are built to be honest about uncertainty, not to eliminate it.

And they don't know you. No engine here knows your horizon, your risk tolerance, or your tax situation. They grade setups; they do not give advice.