Standard astrology has no calibration. A configuration is interpreted as significant or not by the practitioner; whether it actually fires above what random chance would produce is rarely tested. Tempora publishes per-chart, per-signature lift figures that are reproducible from public code and a public event set. This piece walks through the math, reads the figures honestly, and is open about which figures are under recalibration after the 2026-05-09 audit.
In conventional astrology, a configuration is treated as predictive on the strength of textual tradition. A practitioner reads that Mars conjunct Rahu signifies sudden violent escalation, and treats any occurrence of the configuration as carrying that meaning. There is no standard procedure for asking how often the configuration actually fires alongside the predicted kind of event versus how often it fires on a random date in the same period.
Calibration is that procedure. For a given chart and a given signature, the calibration engine computes the signature score on a labelled event set drawn from that chart's history, then computes the same score on a baseline of random dates of equivalent period. The lift is the ratio. A lift of 1.0 means the signature is no more predictive than random. A lift of 5.46x means the signature scores 5.46 times more frequently on confirmed events than on random dates. The numerical value is the calibration. The interpretation is what the article does with the value.
This is the difference between asserting and measuring. Tempora measures.
The calibration engine is at engine/calibrate.py. The pipeline is the same for every chart.
Step 1 - assemble the event set. For each chart, a labelled list of historically significant events is compiled. Each event has a date, a category (political, economic, military, pandemic, infrastructure), and a source citation. Event sets range in size from four to fifteen events per chart in the current calibration. Event-set construction is the most subjective step; events are selected by historical consensus and survivorship bias is acknowledged.
Step 2 - score each event. For every event date, the engine evaluates each of nine signatures and produces a numerical score. The signatures used in the current calibration are: Saturn near natal Moon, Saturn opposite natal Moon, Saturn transit of natal Sun, Rahu over natal stellium, Rahu return, malefic opposition to stellium, Jupiter vedha, dasha-lord in dusthana, and Mars-Rahu conjunction. Scores combine an orb component (closer to exact equals higher score) with a dasha-state component (signature-relevant dasha lord active equals higher score).
Step 3 - run the Monte Carlo baseline. The engine generates 300 random dates within the calendar range of the event set and scores each one using the same nine signatures. This produces a baseline distribution against which the event scores are compared.
Step 4 - compute lift. For each signature, the lift is the ratio of the event-set hit rate at a fixed activation threshold to the Monte Carlo hit rate at the same threshold. The activation threshold for the current calibration is 1.0. The raw lift is the unadjusted ratio; the calibrated weight is the lift after a small-sample adjustment that protects against single-event over-credit.
The output of this pipeline is a per-chart, per-signature lift table. The table is stored as a JSON file in the public data folder. Re-running the pipeline regenerates the table.
Nine signatures are computed for every chart. Each draws on a documented Vedic technique and is operationalised as a numerical score the engine can compute.
| Signature | Mechanism | Orb |
|---|---|---|
| Saturn near natal Moon | Transit Saturn within orb of natal Moon (Sade Sati ingress phase) | 6° applying |
| Saturn opposite natal Moon | Transit Saturn within orb of the natal Moon's opposition point | 6° applying |
| Saturn transit natal Sun | Transit Saturn within orb of the natal Sun | 6° applying |
| Rahu over natal stellium | Transit Rahu (true node) within orb of any natal-stellium planet | 5° applying |
| Rahu return | Transit Rahu within orb of natal Rahu (occurs every 18.6 years) | 5° applying |
| Malefic opposition to stellium | Transit Saturn or Rahu opposite a stellium-sign planet | 6° applying |
| Jupiter vedha | Transit Jupiter within orb of a natal Saturn, Rahu or Ketu (vedha contact) | 4° applying |
| Dasha lord in dusthana | Active mahadasha or antardasha lord placed in 6th, 8th or 12th house from natal lagna | House placement |
| Mars-Rahu conjunction | Transit Mars within orb of transit Rahu (Angarak Yoga) | 8° applying |
Each signature score combines orb tightness with dasha relevance. A Saturn-near-Moon contact carries higher score when Saturn is also the active mahadasha or antardasha lord. The combined score is what is checked against the activation threshold.
The calibration runs on six national charts as the current public corpus. Each chart has a documented birth moment, a labelled event set, and a per-signature lift table.
| Chart | Birth moment | Event set size |
|---|---|---|
| India 1947 | 15 August 1947, 00:00 IST, New Delhi (Independence) | 15 events |
| Russia 1991 | 12 June 1991, 12:45 MSK, Moscow (Russian Federation founding; canonical per the audit reconciliation) | 11 events |
| United States 1776 | 4 July 1776, 17:10 LMT, Philadelphia (Sibly chart) | 14 events |
| United Kingdom 1801 | 1 January 1801, 00:00 GMT, London (Act of Union) | 9 events |
| China 1949 | 1 October 1949, 15:01 CST, Beijing (PRC founding) | 10 events |
| Pakistan 1947 | 14 August 1947, 09:00 PKT, Karachi (Independence) | 4 events |
The Russia chart canon was reconciled in the May 2026 audit cycle. The 12 June 1991 founding moment matches the per-event positional data already used in the Tempora calibration stack; an alternative end-of-1991 founding-moment chart had been cited in earlier corpus articles and has been retired across the published surface. The reconciliation is documented in the canonical-charts surface alongside the per-chart event sets.
A lift of 5.46x means that on the events in the labelled set, the signature scored above the activation threshold 5.46 times more often than on the Monte Carlo baseline of 300 random dates from the same calendar range. The figure is a backward-looking correlation magnitude. It says nothing about any specific future date.
The activation threshold for forward-call publication is 1.5x. A signature with calibrated weight at or above 1.5x can produce a forward call when it is active on a chart in a dated window. A signature with calibrated weight below 1.5x is documented but not used as a publication trigger. This conservative threshold is the calibration filter that controls the public forward-call surface.
Two failure modes are easy to make and worth naming. The first is reading lift as probability. A 5.46x lift is not a 5.46x probability of an event in a given window. The lift is the multiplier on baseline frequency at the calibration threshold; it does not translate into a unit forecast probability without additional assumptions the framework does not make. The second is reading lift as causation. The lift documents a correlation between signature score and labelled events. The framework does not claim the signature causes the events; it claims the configuration coincides with the events more often than random chance.
The calibrated-lift framework is operationally a filter on the forward-call surface. Without per-chart calibration, every transit configuration in every chart would be a potential prediction. Calibration narrows the publication threshold to windows where one or more signatures with weight at or above 1.5x are simultaneously active. This filters noise, restricts the forward-call surface to high-lift windows, and produces a manageable number of falsifiable claims per year. The framework is not an oracle. It is a calibrated forward-call filter.
An internal audit on 9 May 2026 reviewed the published lift table against the underlying event sets. The audit ran every (country, signature) pair through the calibration engine and counted how many events in the labelled set actually scored above the activation threshold for the cited signature. The findings are uncomfortable and they are public.
Of fifty-four pairs audited, six PASS, thirty-six WARN, and twelve FAIL at the level the audit calls event-set depletion. A FAIL-level pair is one where the calibrated weight is published but only one event in the underlying set actually scored above threshold for the cited signature. A weight derived from a single event is not a meaningful lift; it is a coin flip pretending to be a measurement.
The audit also surfaced two systematic failure modes. The first is Saturn-axis over-credit: across multiple charts, Saturn-Moon and Saturn-Sun signatures produced lift figures that did not survive engine re-verification at the orbs and dasha-state thresholds the published narrative implied. The second is calibration narrative-versus-math drift: in several articles the published prose cited a lift figure correctly as a number but then named historical events as supporting cases that the engine showed were not actually scoring above threshold for the cited signature.
Specific lift figures retired pending recalibration include: Russia's 5.46x Mars-Rahu conjunction figure, the United States' 4.31x Saturn-Sun figure, the United Kingdom's 4.21x Saturn-Moon-opposition figure, China's 2.07x Saturn-near-Moon figure, and Pakistan's 2.51x Rahu-return figure. Where these figures were cited in published articles, the article pages now carry a Section that supersedes the original framing and documents what survives, what retires, and what stays open for recalibration.
This is the discipline. A research field declares a figure, then audits whether the figure holds. When audits find drift, the finding is documented on the public surface where the figure was originally cited.
Not every figure was caught. Six (country, signature) pairs PASS the audit cleanly with adequate event credit at threshold. The framework's mechanical claims survive: the math is reproducible, the Monte Carlo baseline is well-defined, and the engine returns deterministic output for any (chart, signature, date) triple given the public natal data and the public ephemeris. What the audit affects is which lift figures the framework currently publishes. The pipeline that produces the figures is intact.
The discipline of public reconciliation is itself the brand-defining move. A standard astrology firm cannot publish an audit of its own published numbers because there are no underlying numbers to audit. Tempora has numbers. Tempora audits its numbers. Tempora publishes the audit on the same surface where the numbers were cited. This is unusual in astrology and standard in any empirically grounded research field.
Every Tempora forward call carries an embedded lift figure. The Russia February 2028 window cited the 5.46x Mars-Rahu lift; that figure has now been flagged for recalibration and the article carries a superseding Section that walks through what the engine returns versus what the original narrative claimed. The India December 2027 window cites the Saturn-Moon-opposition lift on the India 1947 chart; that figure is among the pairs the audit flagged at WARN level for stricter credit. The US late 2029 window cited Saturn-near-natal-Moon as a transit signature; the engine showed the signature did not actually fire across the cited window, and the article reframed accordingly.
The framework's response to each finding is the same: document what the engine returns, document what the original article claimed, document the gap, retire what cannot be supported, preserve the falsifier and window structure, and let the call stand or fall on the engine truth rather than on the original prose. This is the falsifier discipline applied recursively to the calibration figures themselves. Open methodology applied to its own output.
The pieces that articulate this in the corpus are Falsifiable astrology: what it means and how Tempora does it, Reproducible Vedic astrology backtesting, and the Calibration research note (Note 005). The forward-call articles named above are Russia February 2028 forecast and Russia Ukraine Resolution Q3 2027.
The framework has known limitations and they are documented on the same surface where the calibration is documented.
Sample size. Event counts per chart are small, four to fifteen events. Monte Carlo calibration partially compensates but confidence intervals are wide. A 5.46x lift derived from four events is suggestive; it is not definitive in any statistical sense the term carries in physics or biology.
Survivorship bias. Events were selected as historically significant by consensus. What counts as significant is itself a judgment. The framework cannot score the significance of absence, and the labelled set inherits whatever bias the selectors imported.
Birth-time uncertainty. National-chart birth times are recorded to varying precision. A two-hour error in birth time shifts the natal Moon by approximately one degree, affecting nakshatra-boundary cases and the small-orb signatures.
Forward overfitting. Calibration on historical data necessarily overfits to some degree. Forward predictions apply additional filters: requiring at least one strong signature with weight at or above 1.5x to be active simultaneously, and deduplicating windows within forty-five day clusters. Even with these filters, the forward window-by-window hit rate is the ongoing falsification test the framework is undergoing in public.
Audit-driven recalibration. Twelve pairs are currently under recalibration. Until the rebuild closes, the public lift table is read alongside the audit findings. This is uncomfortable; it is also honest.
The economic and structural incentives in conventional astrology run against calibration.
A practitioner whose income depends on continuous repeat consultations cannot publish a number that could later be shown wrong. Calibration produces such numbers continuously. The published lift is either supported by the underlying event set or it is not, and any reader with access to the same engine and the same event set can check.
Building the calibration infrastructure is itself a substantial engineering and research investment with no short-term commercial return. Maintaining a labelled event set per chart, running Monte Carlo permutations, and surfacing per-signature lift tables takes more code than a directory of horoscope readings. The work is invisible to a casual reader and does not change the appearance of the published article relative to a non-calibrated one. The investment is structural, not cosmetic.
Public audits of one's own published numbers are commercially costly. They make the practitioner appear fallible and create a paper trail of figures that did not hold. This is the cost Tempora has chosen to pay because the brand position is research-grade publication, not divination. The audit is not a flaw in the framework; it is a feature of any framework that publishes numbers.
This article is a method-defining piece. It documents what calibrated lift means in the Tempora framework, how the figures are computed, what the published figures currently are, and which figures are under recalibration after the May 2026 audit. It is not a defence of the published figures. It is a description of the framework that produces them and the discipline that audits them.
The framework will be tested as the published forward windows close and as the recalibration completes. Some figures will hold. Some will move. Both outcomes will be reconciled on the same public surface. That is the test the framework has chosen to undergo.
A lift figure is the ratio of how often an astrological signature scores above its activation threshold on a set of confirmed historical events compared with how often it scores above the same threshold on a random-date baseline. A lift of 5.46x means the signature fired on confirmed events 5.46 times more frequently than on random dates of comparable period and chart context. The figure is a correlation magnitude, not a forecast probability.
For each chart and each of nine signatures, the calibration engine computes the signature score on every confirmed event date in the chart's event set. It then computes the same score on 300 Monte Carlo random dates drawn from the same calendar range. The lift is the ratio of the event-set hit rate to the baseline hit rate at a fixed activation threshold. The code is at engine.calibrate and the calibrated weights table lives in the public data folder.
Different signatures use different orbs. Conjunction-style signatures such as Mars-Rahu use a 6 to 8 degree orb. Opposition-style signatures such as Saturn opposite natal Moon use a 6 degree applying orb. Eclipse-axis contacts use a 3 degree orb on the Rahu-Ketu axis. Each orb is fixed at calibration time and applied uniformly to events and to Monte Carlo baseline. Changing the orb after calibration would constitute look-ahead leakage.
A calibrated chart is a national or principal chart for which Tempora has built a labelled event set, run the calibration engine, and produced a per-signature lift table. The current six calibrated charts are India 1947 (Independence), Russia 1991 (Russian Federation founding 12 June), United States 1776 (Sibly 17:10 LMT), United Kingdom 1801 (Act of Union), China 1949 (PRC founding) and Pakistan 1947. Calibration is being extended to additional charts post-launch.
An internal audit dated 2026-05-09 found that twelve of fifty-four (country, signature) calibration pairs cited fewer than two events scoring above the activation threshold. The audit also flagged a systematic over-credit on Saturn-axis signatures and several articles where the published narrative cited a lift figure whose underlying event set did not survive engine re-verification. The affected lift figures are flagged on the corresponding article pages. The recalibration is in progress and the public lift table is being rebuilt with stricter event-credit requirements.
A lift figure is a backward-looking correlation magnitude derived from a labelled event set. It says nothing about any specific future date. A forward call combines an active high-lift signature with a dated window, an observable signal, and a falsifier condition. Many high-lift windows produce no event. The lift figure is the filter Tempora uses to decide which windows clear the publication threshold; it is not the prediction itself.
This article documents the calibrated-lift framework Tempora Research operates under and the audit-driven recalibration in progress as of 9 May 2026. Lift figures cited are reproducible against the calibration engine and the calibrated weights table published with Research Note 005. The 2026-05-09 audit findings referenced here are summarised from an internal audit log maintained as part of Tempora's research-publishing standards. This article is a method-defining piece for the Tempora corpus and does not constitute scientific peer-reviewed publication. It does not constitute medical, financial, legal or professional advice. Article first published 2026-05-09 by Tempora Research.