Research only. Not investment advice. Figures are point-in-time.
DiscerniumConsole

Signal, discerned.

Equity research that shows its work.

Discernium screens the S&P 500 on point-in-time price factors, runs a full multi-agent analysis on the names that earn attention, and publishes each decision with the evidence behind it. Every figure reflects only what was public on the analysis date — no look-ahead.

Latest analyses

TickerAnalysis dateRating

Each rating is settled against the benchmark and scored in the public track record — including when it was wrong.

Screen leaders

#TickerScoreSetup

A score orders names for attention; it is not a forecast of return.

Why point-in-time matters

Most backtests quietly use data that did not exist yet: restated fundamentals, revised estimates, filings that arrived weeks after the date being studied. Discernium dates every figure by when it became public and refuses anything later than the analysis date.

A worked example, from a real run

Apple's most recent reported quarter ends 2026-06-27 and was filed 2026-07-31. The September quarter, not filed until late October, is absent from every figure in the report, because on the analysis date nobody had seen it. That is not a gap in the data; it is the honest cutoff.

Two questions, kept separately

The screen asks what deserves attention. It reads price and volume only — relative strength, trend structure, momentum, liquidity.

The research asks what it is worth. It reads as-filed fundamentals, filings and news, and argues both sides before rating.

The two can disagree. When they do, both answers are published.

How a report is produced

4

Analysis streams

market, sentiment, news, fundamentals

2

Opposing researchers

argue the case

3

Risk reviews

aggressive, neutral, conservative

5

Ratings

Buy through Sell, with sizing and levels

One model sampling per run, so a rerun will not be identical — the report states what it read, and the reasoning is shown in full.

Where the data comes from

SourceUsed forVintage
SharadarSplit/dividend-adjusted prices, as-filed fundamentals and statementsDated by when each figure became public
SEC EDGARCompany filings, as filed, for US filersDated by filing date
Alpha VantageNews headlines and market sentimentServed only for present-day runs

Track record

Decisions logged
Resolved
Awaiting settlement

A decision settles over the configured holding window after its analysis date. Until then it cannot be scored, so no hit rate is shown rather than an unearned one.

See the full record →

What this is — and what it isn't

It is a research aid for people who want to see the working. Fundamentals as filed and dated by when they became public. A screen over the index membership that actually existed then. A published record of every rating, including the ones that went against us.

You are meant to disagree with it. The reports state what they read so you can check it.

It isn’t investment advice, a signal service, or a backtested strategy. Nothing here is personalised, and no rating is a recommendation to trade.

Nor is it a data terminal. There are no analyst estimates, no earnings-call transcripts and no segment-level breakdowns — so it cannot see where consensus is wrong, only what companies have actually reported.

One honest limitation, up front

A run is a single model sampling, so rerunning the same name on the same date will not produce an identical report. What is fixed is the data it read — every figure is bounded by the analysis date. The reasoning is shown in full so you can judge it rather than trust it.

Not investment advice

Discernium produces research for your own judgement. Nothing here is a recommendation to buy or sell any security, and nothing is personalised advice. Past decisions do not predict future results. Verify anything before acting on it.