Daśā, gochara, tārā and candra bala, and aṣṭakavarga transit bindus. Nothing is averaged.
This page shows four classical testimonies side by side on their own scales and never averages them: the Vimshottari dasha read by what its lord rules, the gochara reckoned from the natal Moon, the tara and chandra bala of the day, and the ashtakavarga bindus under the transiting grahas. Each lane carries its own citation and its own units. There is no combined score, no percentage and no column that merges the four, and that is a doctrinal constraint rather than a stylistic one.
The reason is simple once the quantities are named. A dasha lane returns a lordship condition; a gochara lane returns a four-state verdict per graha; a tara lane returns one of nine names; an ashtakavarga lane returns a bindu count out of eight. They are not the same kind of quantity, and no text in this corpus weights a dasha against a gochara or a tara against a bindu. Averaging them would produce a number that looks like knowledge and is arithmetic on incommensurables.
Lane one is the dasha, on Phaladeepika ch. XX, which reads a period by what its lord rules, so the same dasha means something different for every ascendant. Both halves of the chapter always ship, with the chapter's own six-clause strength test selecting between them and reporting undetermined where it does not decide.
Lane two is the gochara, on Phaladeepika 26.1, 26.2 and 26.3: 26.1 makes the natal Moon the only licensed reference point for gochara phala, 26.2 gives each graha its favourable and unfavourable houses from that Moon, and 26.3 applies vedha, the obstruction that cancels a result, along with its reversal.
Lane three is tara bala from Kalaprakasika ch. XXXIII, the nine taras counted from the janma nakshatra and named Janma, Sampat, Vipat, Kshema, Pratyari, Sadhaka, Naidhana, Mitra and Ati-Mitra, in which this app's table marks Janma neutral rather than adverse.
Beside it, as a separate rule with its own citation, sits chandra bala, the Moon's house counted from the janma rasi, where 1, 3, 6, 7, 10 and 11 are favourable and 4, 8 and 12 are not. Lane four is ashtakavarga on BPHS ch. 66: the transiting graha's own bindu in the sign it occupies, from 0 to 8, and that sign's sarvashtakavarga total out of 56, where the twelve signs always sum to 337, making the average sign a little over 28.
One quantity does cross the lanes: a count of how many of them point the same way for a given row. A count is not a weighting. Every lane contributes exactly one, in its own units, the lanes remain individually visible after the count is taken, and nothing is scaled, normalised or averaged to produce it. It exists so that rows can be ordered at all, and the page says on screen that it is this application's ordering device rather than doctrine.
Phaladeepika 15.6 does the same thing in a neighbouring context when it says that two or three conditions coinciding tell more clearly, a count with no magnitude attached, and that is the shape being imitated. It is explicitly not a citation for ranking periods, and the page does not claim one. If you want to know why a row is ordered where it is, the answer is always visible: the four lanes are printed beside the count and each says what it measured.
A tara changes about every day; a dasha lord changes about every few years. Sampling all four lanes at one cadence would either fabricate precision for the slow lanes or throw away the fast ones, so the page runs two tables.
The first has one row per Vimshottari antardasha across a ten-year horizon. In those rows the dasha lane is exact for the whole window, while the gochara and ashtakavarga lanes are sampled at the window's midpoint, and every row says so rather than implying that the sample holds across the period.
The second table has one row per day across thirty days from the start date. There the tara and chandra bala are exact for the day, and the transiting Moon's own gochara house is carried with them. Nothing is interpolated between the two cadences and no row of one is merged into a row of the other, so a daily verdict never inherits a dasha row's ordering and a dasha row never borrows a day's tara.
It will not name a lucky day, a lucky number or a lucky colour, because those would have to come from a merged verdict the page does not compute. It will not tell you to act or wait on a particular date. And it will not report a period as good or bad overall: a row can have a favourable dasha lane, an unfavourable gochara lane and a middling bindu count all at once, and that is information rather than a contradiction to be smoothed away.
What it does give is checkable. For every row you can see which chapter produced each verdict, on what scale, at what moment it was sampled, and how many lanes agreed. That is enough to compare two periods honestly and enough to disagree with the page, which a single blended number would not be.