How we test
Last Updated: October 16, 2026
We check whether a reading picks out the month of a real life event better than random months do. So far it does not. That is why AlmaMesh readings are for reflection: they show what your chart highlights, not what is coming.
- Baseline: dasha periods and slow-planet transits6,120 events1.03 (0.90–1.16)
- + exact transits (a slow planet within 1° of a birth planet)6,120 events1.03 (0.90–1.18)
- + the obstacles theme (legal matters and accidents only)772 events0.88 (0.56–1.30)
- + Ashtakavarga strength of transits6,120 events0.98 (0.85–1.11)
- + divisional-chart lords (D9, D10, D30)6,120 events0.99 (0.86–1.12)
- + classical dasha rules (relevance, placement from the parent lord, strength), by area5,098 events across the four areascareer 1.04 (0.74–1.41)relationships 1.06 (0.85–1.29)family 1.22 (0.87–1.66)health 0.94 (0.49–1.57)
What we test
Take a dated life event, like a marriage, a job change or a loss in the family. Score every month in the six years around it with the same calculations the app uses: 73 months in all. Then ask: is the real month among the top 3?
Picking 3 of 73 months at random lands on the right one about 4% of the time. We also score each chart on placebo months with no event, so a chart that rates every month highly cannot look good. Lift compares the real months with the placebos: 1.0 means no better than chance.
What we test it on
6,120 dated events from 5,062 people with reliably recorded birth times, from Astro-Databank, a research database we use under licence. Most of these people are public figures.
The licence allows totals only. We never publish names, birth details, single events or examples, and none of that data is in the app.
What we found
No lift over chance. The baseline scores 1.03, and none of the five techniques we added moves it: every range above includes 1.0. The classical dasha rules are how readings are built now, but this test finds no lift from them (best corrected p-value 0.80).
What this means for you
Readings stay specific. They name the dasha lords, transits and dated windows behind each theme, compare periods, and say plainly when hard themes such as legal friction or losses stand out.
But they are for reflection. Use them to think about a period, not as a reason to expect a particular event.
How a technique earns stronger wording
Each test is written down before it runs. A new technique is scored on half of the people first. Only if it beats chance there, after correcting for every test we ran, does it go to the other half, held out and unseen. Only a technique that passes that held-out test can make the app's wording stronger. None has so far.
We also tested one exploratory pattern, relocation timing with sharper planetary transits, on data we had set aside and never looked at. On the held-out half of the sample (71 relocation events) it did not replicate: lift 0.69 (95% interval 0.08 to 1.95, p = 0.72), where 1.0 means no better than chance. The earlier result on the development half (4.56) did not hold up. We pre-registered this test, ran it once, and changed nothing afterwards. Relocation readings stay reflective only.
We keep testing
We re-run this test whenever we change how readings are computed, so a change cannot quietly make readings look better than they are.