How DeepAstro calculates and scores cities

01

1. Inputs that belong to this chart

The calculation starts with birth date, local birth time, birthplace coordinates and timezone. These determine one UTC instant and the planetary positions used to project ASC, DSC, MC and IC lines across the Earth.

A city contributes its coordinates and country. The selected goal, such as relationships or career, changes how relevant lines are weighted. If a Bazi preference is available, it is kept as a separate directional cross-check rather than blended into the astronomical calculation.

• Birth time and timezone affect angles and can materially move line geography.
• City coordinates come from the selected place result, not from the city name alone.
• Changing the goal can change both the top city and its primary line.

02

2. Line proximity, angle and goal relevance

For each candidate city, the engine finds the closest sampled point on every calculated planetary line. Version 5 considers lines within 600 km and applies a smooth distance decay. A line close to the city carries more weight than the same line near the edge of the evidence radius.

Planet and angle are evaluated together. For example, MC receives the highest angle weight for a career goal, DSC for a relationship goal, IC for home or inward work, and ASC for identity and vitality. These weights are DeepAstro product conventions, not empirical laws.

Support

A nearby line whose goal affinity remains materially positive after its pressure weight is included.

Trade-off

A nearby Mars, Saturn, Uranus, Neptune, Pluto or South Node line whose modeled pressure outweighs its support for the selected goal.

Neutral

A line inside the radius whose contribution is too small to justify a strong positive or negative claim.

03

3. Combining evidence without score inflation

Several close lines do not receive unlimited additive points. Support and trade-off contributions use diminishing returns, so the strongest evidence matters most and later lines add progressively less. This prevents dense crossings from automatically producing a perfect score.

The optional Bazi directional match adds at most 10 raw points. It can confirm a close decision, but it cannot outrank strong contradictory line evidence. Distance from the birthplace is recorded as context only and adds no recommendation points.

Earlier ranking logic rewarded proximity to the birth city and forced familiar relocation hubs into the first ten results. Version 5 removes the home-distance bonus. A practical hub receives only a two-point near-tie adjustment, and geographic diversity can replace a duplicate metro only when the raw score gap is six points or less.

04

4. Score, evidence band and confidence

The combined raw result is converted to a display index from 4 to 96. The curve keeps visible separation at the high and low ends without presenting 100 as certainty. SSS begins at 85 and means the strongest evidence band for this chart and selected goal. It does not mean the city is in a global top one percent.

Confidence is separate from score. High confidence requires at least two material support lines with close evidence. Medium requires at least one material support line within 450 km. A result with only weak, neutral or pressure evidence is labeled low confidence even if it remains in the candidate list.

05

5. Ranking, evidence and monitoring

The global recommender scores a broad city pool, preserves the strongest calculated matches and then applies narrow practical and diversity tie-breaks. The result exposes the primary line, its distance, support count, trade-off count and confidence. When no material line exists, the interface says so instead of inventing a Jupiter, Venus or element label.

Every rendered Top 3 records the algorithm version, goal, city ids, scores, primary lines and confidence. This makes it possible to detect a regression where different charts begin returning the same cities or saturated scores.

• Automated differentiation tests run at least six structurally different charts side by side.
• The complete production candidate dataset is included in the ranking test, not only hand-picked examples.
• Important conclusions must trace back to the current chart line, angle and distance.

06

6. Version and change policy

The current recommendation algorithm is v5, introduced on September 3, 2026. A scoring change must increment the version, pass cross-input differentiation tests and preserve an audit trail in analytics. Production helpfulness is evaluated separately from code correctness.

A rule can be internally consistent and still be unhelpful to users. DeepAstro therefore treats user feedback, response rate and negative reasons as a release gate rather than evidence that astrology has been scientifically validated.

Questions about the score