what does lulu lift do?

Measures what one source was worth, between the arm that carried it and the arm that did not: the appearance rate with the list held whole, the rate with the page taken out, and the paired gap between them with its interval.

how do you run it?

In the parser’s own words, what one source was worth, between the arm that carried it and the arm that did not. Every flag this build accepts, on one line:

lulu lift --held HELD --dropped DROPPED --brand BRAND --source SOURCE --sources URL [--live LIVE] [--margin MARGIN] [--gate-margin GATE_MARGIN] [--confidence CONFIDENCE] [--brands BRANDS] [--per-brand] [--ledger LEDGER]

lulu lift --live ROUND --held ROUND --dropped ROUND --brand BRAND --source https://b.example/list --sources https://a.example/guide --sources https://b.example/list --sources https://c.example/review  ± one page out of three, gated

when may the number be called a lift?

The word lift is not the name of that number, it is a name it has to be granted. The contrast is causal either way, since the list is ours and one thing moved; what a passing gate buys is somewhere to carry it.

Without --live there is no gate, the same arithmetic prints as an arm difference, and the exit code is not zero. With one, the levels are restated with the gate's own margin added on top, because a laboratory rate quoted as a customer's rate is uncertain by both.

is the source list taken on trust?

No. A replica round records the digest of the exact material it was shown, so the list given on the command line is checked against the evidence file, and a reordered list, a swapped pair of arms or an edited instruction is refused by name rather than measured.

--sources is given once per page, in the order the model saw them, and --source is the one page taken out of that list in the dropped arm.

which flags does it take?

--held HELD ± the replica round asked with the full list, required

--dropped DROPPED ± the replica round asked without one source, required

--brand BRAND ± which brand the two arms are scored for, required

--source SOURCE ± the one page removed from the list in the dropped arm, required

--sources URL ± one page of the full list, repeated once per page, in the order the model saw them, required, once per value

--live LIVE ± a live round; supplying it runs the gate, and only a gate that passes allows the word lift

--margin MARGIN ± how many points a source may be worth and still count as not worth chasing, default 5

--gate-margin GATE_MARGIN ± margin the replica gate is held to; the same as --margin unless given

--confidence CONFIDENCE ± default 0.95

--brands BRANDS ± how many brands are held at once, default 1

--per-brand

--ledger LEDGER ± directory the rounds live in, default ./ledger

where does it lead?

lulu ablate ± lulu collect

the parts of the method it reaches: intervals, variance, compare, ablation, lift.

every command: setup, init, doctor, collect, draft, usage, ablate, lift, screened, publish, rivals, report, verify, plan, size. back to the docs