Do planetary positions predict market crashes?

If planetary positions carry information about market stress, then days with a given configuration should be followed by more drawdowns than days without it. That is a testable statement, and it is the one this project set out to test.

12,014 trading sessions from 11 February 1992 to 1 January 2025, with the Nifty 50 close and the nine bodies (Sun, Moon, Mars, Mercury, Jupiter, Venus, Saturn, Rahu, Ketu) placed by Swiss Ephemeris in the Lahiri sidereal zodiac. Two drawdown labels were defined in advance: a minor one (a fall of at least 5% from the recent peak) and a major one (15%).

The full dataset is in the repository and in the site�s database: the daily record, the positions, the aspects, the sign ingresses, the episodes and the base rates.

For each configuration, count the days it occurred, then look at the next 5, 10, 21 and 42 sessions. Compare the observed rate of each drawdown label with the rate over all days. This is a base-rate study: it says what followed, not what will follow.

The prototype shipped with a headline score for spotting crashes. When it was checked against the simplest possible rule — always answer �no crash� — the score was the same. A rule that never raises an alarm scores well on a rare event, and the headline number was measuring the rarity of the event, not any skill.

A second figure, for labelling market regimes, could not be reproduced from the data at all. The reported improvement over the baseline was negative once the split was rebuilt honestly.

So the answer to the question in the title is: not in this test. The configuration-level base rates that remain on this site are shown with their sample sizes, and they are descriptive. Nothing here is a prediction.

A good-looking number is not evidence. Three questions separate a real model from a lucky one: what does the trivial baseline score; how many examples support the number; and does the result survive on data the model has not seen. The first question alone would have caught this.

The crash atlas shows 22 drawdowns since 1906 with the sky recorded around them — a historical record, not a forecast. The base-rate tables are available in the database for anyone who wants to test a different hypothesis, and the method is written down.