Archive methodology
How to read this archive
- Closed events: the ones that already have their outlook published and their review done.
- Open archive: the review after the fact, so you can see what held up.
- Optimistic, pessimistic and thin-context outlooks: not every event arrives with the same level of context.
- See for yourself: misses stay visible too.
What each event page shows:
- Outcome: whether the outlook turned out accurate, partial or missed.
- Consolidated: how far the price had moved from the starting price by the close of the following session. It carries the market's sign: positive if it rose, negative if it fell.
- Max favorable move (MFE): the furthest the price went in the anticipated direction after the announcement.
- Max adverse move (MAE): the furthest it went the other way.
- Starting price and Change: on the chart, the price everything is measured from and, at each point you select, how far the price has moved from it.
Important: this is context and historical review, not operational guidance.
What is a partial outlook?
An outlook is partial when, after the announcement, the price moves at least 1% in the anticipated direction but closes the following session less than 1% in the outlook's favor. It may still close slightly in its favor, or close against it.
If it closes against the outlook by more than 2.5 times the favorable move, it is no longer partial: it counts as missed.
For part of the session the price went the anticipated way, but that doesn't mean holding a position until the close would have made money.
An example, the most recent in the archive: NiSource Inc. reported earnings on October 28, 2026, and the published outlook was optimistic. After the announcement the price rose as much as 2.3%, and by the close of the following session it was up 0.6%. On the event page, those two figures are “Max favorable move (MFE)” and “Consolidated”.
The figure, as of today
Across the whole archive, what was published before each event shows 73.6% accurate outlooks over a sample of 785 events, with a 95% Wilson confidence interval running from 70.4% to 76.6%. Figures as of .
The misses are inside that percentage: it is worked out with the outlooks that went wrong included, and the archive can be checked event by event.
How is the % of accurate outlooks calculated?
How each outlook is classified:
- Accurate: after the announcement, the price closes the following session at least 1% in the anticipated direction.
- Partial: after the announcement, the price moves at least 1% in the anticipated direction but closes the following session less than 1% in the outlook's favor, or even against it. If it closes against the outlook, it does so by no more than 2.5 times the favorable move.
- Missed: between the announcement and the close of the following session, the price does not reach 1% in the anticipated direction (that does not mean it moved the other way), or reaches it but closes against it by more than 2.5 times the favorable move.
In the percentage, partial outlooks count as accurate. Each outlook is measured from the announcement to the close of the next session.
Accurate means the price did move in the anticipated direction at some point, not that anyone made money. The archive keeps no tally of gains or losses.
Retraining the system only affects future evaluations: it is never used to re-evaluate events that have already closed.
The same count, month by month, is in the archive months table.