Decision Engine

The panels above say what the Decision Engine returns. This long read goes underneath them: where the idea of judging a question from one house and its lord comes from in the classical texts, exactly how the two terms of the score are computed, how to read every block the signed-in page can show, one sample chart run through the real engine twice - one minute of birth time apart - and what the engine cannot know. If dashas are new to you, what a Mahadasha actually is is the six-minute introduction.

Every figure below was produced by running the code that serves the Decision page on a sample birth - 15 June 1990, 10:30 IST, Mumbai - evaluated at noon IST on 5 October 2026 with nothing logged. It is not a real person. The distribution figures come from a sweep of 2,000 randomly generated sample births through the same functions on the same date; they describe how the formula behaves, not the people who use this app.

A yes-or-no question in the classical texts

Classical Jyotisha has two different ways of answering "should I?". The first is prashna, horary astrology: a chart is cast for the moment the question is put, and the answer is read from that chart alone. Daivajna Vallabha, the horary text attributed to Varahamihira, opens (Ch. 1 Sl. 2) by saying that the astrologer who weighs the states of the planets and the results of the houses does not go wrong. Sl. 3-8 names ten such states, five of them by dignity - Dipta in exaltation, Swastha in its own sign, Mudita in a friendly sign, Supta in an enemy's, Dina in debilitation - and attaches outcomes: the object is fully gained under a Dipta planet and accomplished under a Swastha one, while Supta brings danger and Dina want. Ch. 2 Sl. 18 tells the astrologer to treat the house that stands for the matter asked about as the ascendant, Sl. 2 says a house joined or aspected by its own lord flourishes, and Sl. 20 allows three answers rather than two - auspicious, the reverse, and mixed when the indications are mixed. This app keeps that discipline on its own page, Prashna, with a no-account Prashna calculator.

The second way reads the birth chart and asks what the period now running is good for. BPHS Ch. 47 Sl. 2 separates the general effects of a dasha, which come from the nature of the graha, from the distinctive ones, which come from its placement, and Sl. 3-4 says the effects follow the graha's strength. Saravali Ch. 41 Sl. 1 puts it in a line: planets give pleasant or unpleasant results in their dasas according to their natal dispositions. The Decision Engine takes this second route. It never looks at the moment you ask; it reads your nativity and the Mahadasha you are in. That is why asking twice in an afternoon returns the same answer. Daivajna Vallabha asks that a query be put once only (Ch. 1 Sl. 9-10); here that is a property of the arithmetic, not a rule of etiquette. What the engine keeps from both routes is the shape of the reasoning - a house for the matter, a lord graded by dignity and strength, a middle answer allowed - reduced to two numbers that can be checked.

Which house a decision belongs to

You pick one of six areas, and that pick - not your sentence - decides which house is read. The map is a single table in the code (TOPIC_HOUSE), shared with the AI Life Chat, so a topic means the same house on both pages.

Two of these choices deserve a note. Both texts place trade itself in the 7th house, so a business question could classically be argued there; the engine reads the 11th, the house both give to income and gains from enterprise. And BPHS judges bodily well-being first from the Lagna and its lord (Ch. 12 Sl. 1-2), keeping the 6th for disease (Ch. 17 Sl. 1); the engine's Health area reads the 6th. These are one-house simplifications, and the health reading is about timing, not medical advice. Each house has its own page: the 10th, 2nd, 7th, 6th, 11th and 1st.

Only the house's ruler is read - the lord of the sign that falls on that house - not the planets sitting in it. The texts support starting there. BPHS Ch. 21 Sl. 2-3 says a strong lord of the 10th brings fame and good deeds, and that when the same lord is without strength the person meets obstructions in work. Ch. 13 Sl. 1-2 has the lord of the 2nd promoting wealth from an angle or a trine and finances declining when it sits in the 6th, 8th or 12th; Ch. 18 Sl. 4-5 and Ch. 22 Sl. 2 treat the 7th and 11th lords the same way. Occupants and aspects matter in those chapters too, and the score leaves them out; the Kundali page is where they are read.

The first 60 points: the Mahadasha you are in

BPHS Ch. 47 Sl. 5-6 gives the classical test in two sentences. A dasha is favourable when its lord is in the Lagna, in its exaltation, in its own sign or in a friendly sign; it is unfavourable when the lord is in the 6th, 8th or 12th house, in debilitation or in an inimical sign. Saravali Ch. 41 Sl. 6-7 lists much the same conditions and Sl. 8 adds a middle grade, a period whose effects are medium, "neither good nor bad". The engine's function for this, dashaQuality, returns one of three words, checked in this order:

1. 1.Challenging - the Mahadasha lord is debilitated, in an enemy's sign, or placed in the 6th, 8th or 12th house. Worth 0 of the 60 points.
2. 2.Good - none of the above, and the lord is exalted, in its moolatrikona degrees, or in its own sign. Worth 60.
3. 3.Neutral - everything else: a friendly or neutral sign outside those three houses. Worth 30.

That is close to the texts but not identical, and the differences are worth knowing.

• House beats dignity. The challenging test runs first, so Saturn in its own sign in the 6th reads challenging. BPHS is gentler in places: Ch. 48 Sl. 1 pairs debilitation with an inauspicious house for the adverse case, and Ch. 47 Sl. 23-26 gives a strong Moon in those houses "troubles and good times off and on". The engine has no such mixed case.
• A friendly sign earns the middle grade, and sitting in the Lagna earns nothing by itself; BPHS Sl. 5-6 counts both as favourable. Friend and enemy combine the natural table with the temporary friendship of BPHS Ch. 3 Sl. 56-58, read from where the sign's lord stands in your birth chart.
• Only the birth chart is read. BPHS Ch. 47 Sl. 5-6 speaks of the lord's position at the commencement of the dasha, and Ch. 48 Sl. 5-8 says birth and commencement should both be weighed. The engine casts no commencement chart.
• Rahu and Ketu have no dignity table here. BPHS Ch. 47 Sl. 34-39 names exaltation and own signs for the nodes while recording that sages differ; the engine uses none, so a node's period is challenging in the 6th, 8th or 12th and neutral anywhere else. None of the 404 running Rahu or Ketu Mahadashas in the sweep read good. The note closing Ch. 48 Sl. 18-20, that the nodes do well in the 3rd, 6th and 11th, is not applied either.

Only the Mahadasha lord enters the score. The Antardasha is read elsewhere on the page - in the risk register and in the growth axis of the two paths - but not in these 60 points; Antardasha and Pratyantar, explained covers how the sub-periods are sized, and the Vimshottari Dasha calculator dates every level without an account. For what a particular period is classically about, see the Saturn, Mercury, Rahu and Ketu Mahadasha pages.

The other 40 points: the house ruler's Shad Bala

BPHS Ch. 27 defines Shad Bala as six sources of strength computed for the seven grahas from the Sun to Saturn, and says plainly that the nodes are not considered. Sl. 26-29 then builds the strength of a house, and the last step of that sum is to add the strength acquired by the house's lord. Sl. 37-38 closes by saying that whatever has been stated about a house comes to pass through the strongest graha. Measuring what a house can deliver by its lord's Shad Bala is therefore Parashara's own method, not a shortcut invented for this page.

What the engine does not do is compute the full scheme, which BPHS measures in virupas and tests against a required minimum for each graha (Sl. 32-33). It uses a simplified twenty-point Shad Bala - the figure the Kundali page shows and the Shad Bala calculator returns - in four parts:

• Sthana, position, up to 10: exalted or moolatrikona 10, own sign 7.5, friendly sign 6.25, neutral 5, an enemy's sign or debilitation 0 - and 0 whenever the graha is combust.
• Dik, direction, up to 4: full in the graha's peak house (Mercury and Jupiter the 1st, Moon and Venus the 4th, Saturn the 7th, Sun and Mars the 10th), falling evenly to 0 in the house opposite.
• Drig, aspect, up to 3: 3 when only benefics aspect the house it sits in, 0 when only malefics do, 1.5 otherwise.
• One point each for natural rank (Sun 0.875 down to Saturn 0.125), motion and time of birth.

The ruler's total is divided by 20, multiplied by 40 and rounded. A ruler holding 52% of the maximum earns 21 points. The page words the result in three tiers: 60% and above strong, 40% to 59% workable, below 40% thin - "a statement about leverage, not about you", in its own sentence. Across the sweep the share ran from 3% to 99%, and about two rows in three fell between 20% and 60%, so this term usually contributes between 8 and 24 points.

From two terms to YES, WAIT or NO

The two terms are added. A total of 65 or more reads YES, 35 to 64 reads WAIT, and anything below 35 reads NO. Because the first term can only be 0, 30 or 60, the possible totals fall into three overlapping bands:

The verdict is, in practice, the Mahadasha's verdict. In the sweep it matched the grade of the running period in 11,793 of 12,000 reads; the house term moved the number inside its band and changed the word only at the edges. Four things follow.

• All six areas share the same 60-point term, so they usually share a verdict. What differs between them is the house ruler - at most 40 points, typically far fewer.
• The score for an area is constant until your Mahadasha changes, once in six to twenty years. Natal strength does not move, so asking again tomorrow returns the same number.
• NO is a common answer. Any lord in three of the twelve houses, in an enemy's sign or in debilitation grades challenging, and in the sweep 44% of running periods did. The verdicts split 12.6% YES, 42.9% WAIT and 44.5% NO.
• A NO therefore describes where the lord of this period is placed. It says nothing about the particular plan you typed.

The same number is reused elsewhere: the Timing page colours its windows with it and the chat calls the same function. It has two terms by design - the engine's header says a third, transit-based term would need a favourability rule this app does not otherwise use.

Two signals beside the score that do not move it

Under the breakdown the page prints two further readings headed "Read alongside". The first is timing alignment, and it has three possible values, tested in order. Shifting: your running Antardasha ends within 45 days, so the conditions you are deciding under are about to change. Pressured: Saturn is transiting the 12th, 1st or 2nd sign from your natal Moon. Settled: neither. The 45-day line is the engine's own choice; the idea behind it is older. Advanced Predictive Astrology (Vol-II p. 328) treats dasa junctions as the points where visible change arrives, and p. 229 calls dasa results the hour hand of a clock and transit results the minute hand. Of the 2,000 sample charts, 1,333 read settled, 490 pressured and 177 shifting.

The Saturn position is Sade Sati, which the same book (pp. 311-312) defines as Saturn passing through the 12th, 1st and 2nd from the Moon about once in thirty years. The passage is not one-sided: it exempts a well-placed Saturn that is a yogakaraka for the chart, and its examples include people who reached high office during the transit. That fits showing Sade Sati as a signal - the page says it tends to slow things rather than stop them - instead of subtracting points. The Sade Sati calculator dates all three phases, and Gochar shows the other transits from your Moon.

The second signal is numerology alignment, read from your Personal Year: birth day plus birth month plus the current year, reduced to a single digit. The engine sorts the nine values into three tones - 1, 3, 5 and 8 support starting something, 9 is a closing year, and 2, 4, 6 and 7 lean neither way. Numerology Metaphysical Secret's list reads One as beginning anew, Eight as prosper and progress, and Nine as endings, which matches the ends of that grouping. Adding 3 and 5 is the engine's convention, and the same list files "big decisions" under Six, which the engine reads as neutral. Your number is on the Personal Year calculator, the method is in the Personal Year number, explained, and when astrology and numerology disagree covers the case where this signal and the score point different ways.

Neither signal changes the score, and the page says so in a printed note: the same number is compared across dated windows on other pages, and redefining it here would quietly change which dates those pages prefer. The page also checks its own arithmetic. If the two axes ever failed to add up to the headline it would say the breakdown may be out of date. In the sweep they reconciled in all 12,000 reads.

Reading the risk map

Below the score sits "What kind of risk this carries": four named risks, each with a level, a plain sentence and the fact it was read from. They describe the decision; the Risk level in the reading-strength rail measures the whole area.

Three optional questions sit beside the text box: where you are with this (exploring, confused, ready to act), how big a decision it is, and how you feel about it. Each takes a value from a fixed list and can be cleared by tapping it again. None changes the verdict - only how the risks are read - and the engine treats what you say about your own state as better evidence than the chart.

One block follows, "If you go ahead". It takes the highest live risk - the first High in the order above, otherwise the first Medium - and gives one thing to hold, not four. For financial risk it is to name the most you are willing to lose, in a number, before you act; for emotional risk, to make the decision twice with a night between. If you marked yourself ready to act and the verdict is WAIT, the block says you may be here for permission rather than for a reading. These are habits of deciding; the app prescribes no remedies, gemstones or rituals.

The reading-strength rail carries three more figures. Risk adds dasha quality (30 points, Mahadasha and Antardasha), Saturn's transit from the Moon (25), the topic house's strength (25) and active doshas (20). Impact measures how central and how active the house is. System confidence shares the 0-100 scale with the decision score but asks how far the chart can be relied on at all. The Timing long read takes that figure apart component by component.

What the engine reads in your sentence

The sentence you type never reaches the score. It is matched against fixed word lists for four things, and every finding quotes the words it was drawn from.

• Tags. Words such as quit, resign, sell, marry or "sign the" mark a decision as hard to undo; try, test, apply or pilot mark it reversible, and hard-to-undo wins when both appear. Other lists pick out a long or short horizon, a named time ("next year"), high stakes (loan, debt, "all my savings") and urgency. A tag the sentence does not support stays empty.
• Clarity. A 0-100 score for how clearly the decision is put: a specific action (30 points, needs six words or more), a timeframe (25), a recognisable life area (25), and being asked from a settled place (20). Under six words that last part is withheld, not awarded. Each unmet part is listed as something to add.
• Drivers. Fear, pressure, wanting to prove something, wanting to grow - named only when marker words are present, and always as a description of the sentence ("your question reads as written under pressure"), never of you.
• Conflict. Fires only when the stage you selected disagrees with the verdict: ready to act against a WAIT or NO, or confused against a YES. The suggested way through depends on whether the decision was tagged hard to undo.

The sample sentence used for this page was: "Should I quit my job and start my own business next year? Everyone says I have to decide by the deadline and I am worried." The engine tagged it hard to undo ("quit"), dated ("next year"), short-range and urgent ("deadline"), and asked under strain ("worried"). Clarity came to 80: action, timeframe and area were all met, and the settled-place part was not. Pressure was the leading driver on three markers - "have to", "deadline", "everyone says" - with fear second on one. A one-word question, "job?", scored 38 and was marked provisional.

Word lists have edges. "I want to learn and grow, should I apply for the new role next month?" scored 75 with the life area unmet, because "role" is not on the career list and "job" is. The lists are mostly English with a few Hinglish words (naukri, shaadi, paisa, agle saal), and a question typed in Devanagari matches none of them. None of this touches your score. The same layer attaches "What this does not mean" to every answer: a YES does not mean it will work out, a WAIT is not a prediction that it will go wrong, and a NO does not mean the thing is impossible.

What-if, dated alternatives and the two paths

The timing card asks two questions about when. In the what-if box you type the move you are weighing. The text is not scored - the verdict is the same result for the area - but the simulation adds three things. First, a signal range: the score with a margin either side equal to a quarter of whatever System confidence falls short of 100, so thin evidence widens the band. The page labels it a chart-signal band, not a probability. Second, dated alternatives: the next eight Antardasha handovers on your timeline are scored by the same formula, and those inside five years are sorted into better dates, dates to avoid and no difference either way, up to three of each. A date is better when the score rises by 5 or more without risk rising by more than 5, or when the score holds and risk falls by 10 or more; avoid is the mirror image. When nothing qualifies the page says so. Third, for career, money, business and love decisions, a 90-minute window in the thirty days after the better date, from the rule sets of the Muhurat feature (new job, investment, business launch, marriage); health and general have none. You can search windows yourself with the Muhurat calculator.

Because the score reads only the Mahadasha lord, every handover inside the same Mahadasha scores the same. The chart titled "This decision, across your timeline" draws each handover as a step - below 45 weak, 45 to 59 neutral, 60 to 74 strong, 75 and above peak - held flat until the next. Run the simulation four times within two hours and a notice appears: the inputs have not changed, so neither has the answer. That count lives in your browser for the sitting and is not stored.

"Act now, or wait?" puts two dates through one function on three axes. Path B is the date you enter, or the first better handover, or failing both the next handover of any kind - labelled the next turn, not a better one. Risk is the rail's risk score at each date. Growth adds the Mahadasha's grade (30), the Antardasha's grade (20), the Mahadasha lord's Shad Bala (25) and the house ruler's (25). Stability adds transit calm (40, reduced during Sade Sati and when Saturn is 4th or 8th from the Moon), the ruling lord's steadiness (30, less 10 each for combustion, debilitation, an enemy sign, a 6th, 8th or 12th placement and retrogression) and the chart's dosha load (30). A gap of five points or fewer is called level, and an axis measured on fewer than 60 of its 100 points is withheld. The call at the bottom counts axes won; it does not weigh them.

A worked example: one chart, one minute apart

The sample chart has Cancer rising at 29°52', the Moon in Aquarius in Shatabhisha, and Saturn at 0°16' of Capricorn - its own sign - retrograde in the 7th house. Saturn's Mahadasha runs from 2 July 2009 to 2 July 2028, with the Jupiter Antardasha running to the same end date. Saturn in its own sign, outside the 6th, 8th and 12th, grades good: 60 points for every area.

Six houses, seven points between the highest and lowest score, one verdict. Saturn's 61% is 7.5 for its own sign, the full 4 for standing in the 7th, its peak house, 0.125 for natural rank and 0.5 for time. Jupiter's 54% includes 5 for position: Gemini belongs to Mercury, Jupiter's natural enemy, but Mercury stands in Taurus, the 12th from Jupiter and so a temporary friend, and the two cancel to a neutral sign. Beside the score, timing alignment reads pressured - Saturn is transiting Pisces, the 2nd from the Aquarius Moon, the last phase of Sade Sati - and numerology alignment reads neutral: 15 + 6 + 2026 = 2047, which reduces to 13 and then to 4.

With nothing declared, the risk map reads emotional Low (Moon at 64%), financial Medium (the Sun, ruler of the 2nd, holds 59%, and career counts as a money-bearing area), timing Medium (Sade Sati) and long-term Medium (about 1.7 years of the Saturn period remain). Select ready to act, high impact and under pressure, and emotional becomes Medium while financial and long-term become High; "If you go ahead" leads with the financial one. The verdict and its 81 do not move.

The what-if for "Resign in January" returns a signal range of 66 to 96: System confidence is 40% (a birth time but no logged events, and all seven grahas in the middle strength band), so the margin is 15 either side. Risk is Low at 33, impact Medium at 44. No better date is found. Two handovers fall inside five years, both under Mercury, whose Mahadasha grades neutral here and scores 51. The first, on 2 July 2028, loses 30 points but lowers risk by 7 and is filed under no difference; the second, on 29 November 2030, loses 30 with risk flat and is listed as a date to avoid. The two-path card compares today with 2 July 2028: risk 33 against 26, growth 61 against 53, stability 68 against 93 - "Waiting until 2028-07-02 is better on risk and stability". So one page shows a YES at 81 and a paths call that favours waiting on two axes of three. They answer different questions: the score says the period supports the move, the paths card that the ground is calmer once Sade Sati has passed.

Now move the birth time one minute, to 10:31. The Ascendant crosses into Leo, every house shifts by one sign, and Saturn - still in its own sign - now sits in the 6th.

At 10:31 the engine finds a better date - the day the Mercury Mahadasha begins, which raises the score by 30 and lowers risk by 21 - and offers a window after it: Wednesday 26 July 2028, 3:30 to 5:00 PM IST, scored 90 by the new-job rule set. Selecting ready to act now raises a conflict, and because the sample sentence was tagged hard to undo, the suggestion is to move the reversible parts and hold the one-way step. Nothing about the person changed between the two columns. One minute moved a boundary, and the formula followed it.

Birth-time sensitivity, and no birth time at all

The example is an edge case, but the dependency is general. Both terms need the Ascendant: the first to know whether the Mahadasha lord sits in the 6th, 8th or 12th, the second to know which sign falls on the house and so who rules it. In the sample, one minute moved the Ascendant by about a quarter of a degree, so a time rounded to the quarter-hour can sit several degrees off. If your rising degree on the Lagna calculator is within a few degrees of either end of its sign, treat the verdict as provisional until the time is confirmed. The dasha dates move too, though far less: one minute shifted the sample's Saturn handover by about four and a half days.

With no birth time the engine declines. It returns WAIT at 0 with a sentence asking for the time, and the page draws that as "Score withheld" - printing a 0 there, its note says, would invert what the engine said. The breakdown shows no axes. In the risk map, emotional and financial are named as unreadable (emotional returns at Medium if you declare a state), while timing and long-term still read, because they need only the Moon and the dasha timeline. For the sample, cast at an assumed noon, the rail's risk is provisional on 55 of 100 points, impact is withheld, System confidence is 0%, and the Saturn handover lands on 20 May 2027, more than a year early. What still works without a birth time explains the general case.

The record: logging, follow-through and the accuracy figure

Asking a question answers it and forgets it. Pressing "Log this decision" is what stores it: the server recomputes the verdict from your chart - it does not accept one from the browser - and saves it with the first 500 characters of your question, the area, the score, the reasoning and the dasha lords of that moment. The three framing answers are not sent. Later you record what happened (went well, mixed, went badly), whether you followed the verdict, and an optional note. Nothing is inferred; a field you leave empty stays empty.

• Follow-through appears once three decisions have both an outcome and a followed-or-not answer. It reports how many of the followed ones you marked as having gone well, beside the ones you went against.
• Engine accuracy for you appears at five, and counts only decisions where you followed the verdict and then recorded an outcome - the one case that tests the engine and not your life. Below five the figure is withheld, because one more entry would swing it too far to mean anything.
• "How you decide" also needs five closed entries. It can name one of four patterns, each defined on the record and shown with its count: following through and it working, acting through a WAIT (three or more closed WAITs, 60% or more gone against), deciding and moving on, and weighing one area repeatedly without closing it. Success rates by area need three closed entries in that area.

A poor record is reported as plainly as a good one. The Accuracy feature gathers these figures with the app's other feedback tracks.

What this engine does not do

• It does not read your plan, so it cannot choose between two job offers: both are career decisions by the same person at the same moment, and every figure is identical. The engine's two-option comparison (available through the API, not drawn on the Decision page) declines to pick inside one area; across areas, the sample chart put career 5 points above business and called career and money level.
• It does not score today's sky. Transits appear as a signal and in the risk and stability figures, never in the verdict. For a day or an hour, use Muhurat; for a question of the moment, Prashna.
• It does not use the Antardasha, aspects on the Mahadasha lord, the divisional charts or the yogas in the score, though BPHS weighs several of these when judging a period.
• It does not give a probability, and it does not replace a doctor, a lawyer or a financial adviser. The 0-100 figure is a signal strength with two named parts, the second a simplified Shad Bala; its Health and Money areas are timing readings only.

How positions are computed - the Lahiri ayanamsa, whole-sign houses, the limits of the astronomy - is set out on the methodology page, and the books behind each rule are listed under sources. To see the inputs for yourself before signing up, cast a free birth chart; for the same question seen as a profile and not a verdict, see Career and Money.

Does the Decision Engine actually read my question?

For wording only - whether the decision sounds reversible, how clearly it is put, and what the sentence suggests is driving it. The verdict and its score come from the area you pick and your chart, so two different questions in the same area on the same day get the same score.

Why is my verdict the same for every area, and the same every time I ask?

Sixty of the hundred points come from the Mahadasha you are running, which is the same for every area and lasts for years. The other forty come from the natal strength of a house ruler, which never changes. The score moves only when your Mahadasha does; the dasha calculator shows when that is.

I got a NO. Does that mean my plan will fail?

No. A NO almost always means the lord of your current Mahadasha is in the 6th, 8th or 12th house, in an enemy's sign or debilitated, which the classical texts read as an unfavourable period. In a sweep of 2,000 sample charts, 44.5% of all reads were NO.

What is the difference between the decision score and System confidence?

The decision score says how supportive the timing is for that area. System confidence says how far the chart can be relied on at all - whether a birth time is on file, how many life events you have logged, how decisively the chart sorts its planets. The sample scores 81 with System confidence at 46%.

How is this different from Prashna?

Prashna casts a chart for the moment you ask and reads the answer from it, so the same question on another day is a different chart. The Decision Engine reads your birth chart and current Mahadasha, so the moment of asking plays no part. The Prashna feature is the place for a question of the moment.

What happens if I do not know my birth time?

The engine withholds both the verdict and the score, because forty of the hundred points need the Ascendant and so does the test for the 6th, 8th and 12th houses. The timing and long-term risks still read. This article covers what the rest of the app can still do.

Is my question stored?

Only if you log it. Asking returns an answer and keeps nothing; logging saves the question, the area and the verdict so you can record the outcome later. What-if runs are not stored. Logged decisions are deleted with your account, as described on the privacy page.