Lottery Windfall Model Update: Apophis, Lilith, and Calibration
The latest production handoff adds esoteric asteroid features, publishes the backtest lift, and keeps the proof boundary intact.
This update records the current lottery / windfall research handoff that now feeds the production Astral Edge scorer.
The short version: the engine now resolves Lucifer (1930), Apophis (99942), Mean Black Moon Lilith, and Asteroid Lilith (1181). Two Jupiter-to-natal-body features were promoted into the windfall evidence registry, and the active model backtest improved after the scoring layer stopped double-shrinking validated effect sizes.
This is still a ranking signal, not a lottery promise. The strongest individual windfall findings did not survive Benjamini-Yekutieli correction, so they remain evidence-weighted timing hypotheses rather than proof claims.
What changed in the engine
The backend ephemeris now computes four esoteric bodies used by the windfall research lane:
| Body | Runtime handling |
|---|---|
| Lucifer (1930) | Minor-body solver with Swiss Ephemeris compatible ID fallback |
| Apophis (99942) | Minor-body solver with Swiss Ephemeris compatible ID fallback |
| Mean Black Moon Lilith | Mean lunar apogee calculation |
| Asteroid Lilith (1181) | Minor-body solver with Swiss Ephemeris compatible ID fallback |
The mobile client also has glyph mappings for these bodies, so live forecast reasons can render compactly in the app.
New windfall hypotheses
Two features were added to the production evidence layer:
| Feature | Meaning |
|---|---|
H108_transit_jupiter_to_apophis_partile |
Transiting Jupiter makes a tight Ptolemaic aspect to natal Apophis |
H109_transit_jupiter_to_lilith_partile |
Transiting Jupiter makes a tight Ptolemaic aspect to natal Mean Black Moon Lilith |
The symbolic read is straightforward: Jupiter marks expansion; Apophis and Lilith describe disruptive, shadowed, or taboo material. In a windfall context, the research question is whether exact Jupiter activations of those natal points cluster around jackpot events more often than a shuffled event baseline.
Event-weighted permutation results
The latest timed-birth lottery cohort used 2,000 empirical permutation resamples. The leading event-weighted findings were:
| Feature | Observed | Null mean | Cohen's h | p-value |
|---|---|---|---|---|
H108_transit_jupiter_to_apophis_partile |
14.4% | 4.5% | +0.35 | 0.0035 |
win_transit_venus_to_H2_ruler_hard |
14.4% | 4.8% | +0.34 | 0.0060 |
win_profection_H5 |
18.9% | 7.5% | +0.35 | 0.0090 |
win_progressed_moon_to_H2_ruler_partile |
8.9% | 2.2% | +0.31 | 0.0100 |
H109_transit_jupiter_to_lilith_partile |
13.3% | 5.3% | +0.28 | 0.0185 |
The important boundary: these are strong exploratory and evidence-weighted signals, not settled proof. The report found no BY survivors at q=0.10.
Active model backtest
The active production model is research-anchored-v4-N593. After calibration,
the same-day ranking check across lottery-winner rows produced:
| Metric | Result | Random baseline |
|---|---|---|
| Mean same-day percentile | 0.700 | 0.500 |
| Median same-day percentile | 0.708 | 0.500 |
| Top-quartile rate | 30.4% | 25.0% |
| Top-decile rate | 16.1% | 10.0% |
That is why the app can use these factors as part of a timing rank. It is not a claim that a high-ranked window predicts a win or should change someone's budget.
What failed
The broader sweep is just as important as the hits. Traditional moon-phase claims and lunar mansion / nakshatra claims showed no statistical edge in this run. The strongest extended phenomena were more specific: transiting Moon to natal Jupiter and transiting Moon to the natal eighth-house ruler.
Product translation: Astral Edge can show these windfall factors when they are present, but it must keep the copy reflective, budget-safe, and non-predictive. The right user action is journaling and comparison against reality, not chasing a jackpot.