Explainers / EVENT
News events
Material corporate events, classified from the issuer's own filings as they arrive. This family is descriptive: it tells you an event of a certain kind occurred, and links you to the release.
Engine family slug news_event · methodology news_event/0.2.0
What has to be true
- A filing arrives for an issuer under coverage and is classified into an event category by our own extraction.
- The event is scored for novelty against that issuer's preceding twelve months of disclosures — an event type that issuer files every quarter scores low however dramatic it reads.
- A materiality score is computed, and the signal is published when it clears the threshold published on the row.
- The original release is LINKED. It is never reproduced, quoted at length, or summarised from a third-party article body — see the note on data below.
Thresholds
The settings this family applies. They are published because they are choices: a different window or a different count produces a different set of signals from the same data.
- Materiality score
- must clear the threshold published on the row
- Threshold in force
- recorded per signal
- Source
- the issuer's own filing
- Novelty
- scored against the issuer's trailing twelve months of disclosures
Carried on the row rather than held in a constant, so a signal published under an older threshold still says which one it met.
Primary documents only. No news vendor's article body is read, stored or displayed.
The published work it draws on
Named, not reproduced and not linked. Each entry says what that work reported, in its own terms. None of it was conducted by us, and none of it is a statement about the current market.
Ball, R. and Brown, P. (1968), ‘An Empirical Evaluation of Accounting Income Numbers’, Journal of Accounting Research 6(2).
The founding demonstration that disclosed corporate information moves prices — the premise this family rests on, and nothing more specific than that.
Tetlock, P. (2007), ‘Giving Content to Investor Sentiment: The Role of Media in the Stock Market’, Journal of Finance 62(3).
Found measurable association between the language of financial media and short-horizon market behaviour, with much of it reversing.
Where this does not work
Every detection method has conditions it was measured under and conditions it fails in. These are this one’s, written down rather than left out.
- This family has NO published backtest. It is a classification surface, not an evaluated strategy: we can tell you an event of a given type was filed and how novel it was for that issuer, and we cannot tell you what followed events like it, because we have not measured that yet.
- The classification is automatic and it mislabels. An event category is our extraction's reading of a document, and a reader who acts on the label rather than opening the linked filing is acting on a machine's summary of something they could have read.
- Novelty is scored against twelve months of one issuer's filings. An issuer that has been quiet scores everything as novel, and a serial filer scores nothing as novel, regardless of what the documents say.
- The timestamp we publish is when the filing reached us, not when it became public. Those differ, and the difference is not constant.
- The second citation above measures news media, which this family deliberately does not read. It is offered as context for why disclosed information is worth classifying at all, not as evidence for this detector.
What this page cannot tell you
- What the event means for the business. Read the filing we link to.
- Whether events of this kind have been followed by anything in particular. We have not measured it.
- Anything personal to a reader's circumstances.