One Hundred Thousand Evaluations
Before I let this software speak to a single client, I made it face fifty thousand randomly generated couples in each mode: real ephemeris positions, births from 1970 to 2005, random times, fifteen Indian cities, the complete classical pipeline every time. Here is what an honest engine looks like under that pressure.
Where rejections come from
Every rejection is attributed to a specific classical gate, and every report lists the exact factors behind it. In the traditional mode: divorce-yoga load above 15 points for either partner accounts for 72.1% of all couples, a composed final score below 50 for 6.6%, an individual marriage-dignity below the classical 40 threshold for 3.6%, and compounded pakka indications for the remainder.
The pass rate is also derivable on paper. Each classical factor has a knowable base rate: a given lord falls in a given house once in twelve charts, two planets share a sign once in twelve and stand within fifteen degrees in three of four such cases, and so on. Summing probability times points gives an expected divorce load near 13 with a spread near 9, which puts about 47% of individuals over the 15-point ceiling. Both partners clearing it is 0.53 squared, roughly 28%; both clearing the dignity threshold multiplies in another 0.69; the final-score floor trims the rest. The product lands at about 17 to 18 percent. The simulation measured 17.6. When theory and one hundred thousand trials agree to within half a point, the engine is doing exactly what the shastra it encodes says it should.
Reading these numbers honestly
Random pairings are not proposals. Real couples reach an astrologer already selected by family, community and their own instinct, so pass rates in practice run far higher than this baseline. What the baseline proves is discrimination: this engine does not flatter. It is built source-first on the Brihat Parashara Hora Shastra and Bhrigu Sutram, it applies the classical cancellations exactly as prescribed, and it is willing to say no. When it says yes, that yes has survived every screen the shastra prescribes.
Study conducted on engine v2.0.23, June 2026. Method and full tables available on request.