How the SPVS research process works

Sentiment inputs

Each market day, SPVS scores the most-discussed US equities. It measures how much a name is being discussed, how one-sided that discussion is, which direction the conversation is moving, and how credible the sources behind it are — grading analyst-style commentary differently from hype. Discussion data comes from public social platforms and public news sources.

Market activity

Alongside sentiment, SPVS records market activity for each scored name: price momentum, trading-volume trends, distance from the 52-week high, recorded short interest, recent earnings surprises, and the gap between price and analyst targets. Recorded factors are activated into the score only after their relationship to subsequent outcomes has been evaluated against the platform's own research record; until then they are published as context.

Fundamental context and confirmation

SPVS uses AI-assisted processing of significant volumes of publicly available information related to market sentiment — public social-media discussion, discussion-volume trends, recorded short-interest figures, and trading-volume data drawn from public sources. To corroborate and contextualize those sentiment signals, SPVS also draws on publicly available financial statements and other public company information: financial-health checks and fundamental context are recorded and published alongside every signal, so a sentiment reading is never presented in isolation from the company behind it.

A dedicated fundamental-confirmation scoring layer — balance-sheet health, income and cash-flow quality, and peer comparison within sector and size bands — is in development and validation. Its outputs are not part of the published score until that validation is complete and documented.

Catalysts

Each signal records the catalyst behind the conversation where one is identifiable — earnings, M&A, guidance, product, legal or regulatory, or macro — because a sentiment shift with a dated, verifiable event behind it reads differently from one without.

Current research signals

Scores are ranked into a daily research ladder: the strongest positive readings, the strongest negative readings, and a watch list of names that nearly qualified. Every published signal is then graded against what the market actually did over defined short-term horizons of one to fifteen trading days — the misses stay on the record alongside the hits, and per-source accuracy is tracked so the process itself is continuously reviewed and revised.

Limitations

Public data can be wrong, stale, or incomplete; AI-assisted classification can mislabel; and any given input can be unavailable for any given company on any given day. The research record exists precisely so those failure modes are measured rather than assumed away.

More about the platform and the people behind it on the About page.