Purpose — claims written · claims resolved · merges—
Three numbers on adjacent lines so that merges rising while claims stand still is impossible to read past.
Last run—
Open claims by resolution date—
What is riding on each future date. A date already passed is the resolver's business and is not shown here.
The house model, er_v3 — ranked both directions—
Bullish
Bearish
Every name, every horizon
One published expected return per name per horizon, and it is er_v3's: the median of what followed the name's own past dates in the same state — trend, valuation, market — closed a year before today. The band is their 25th–75th; the count is how many dates, and how many of them are non-overlapping; a coarsened row names every part of the state it dropped to find six. Within noise is a median smaller than the name's own σ√h, and no claim is fewer than six windows even on the trend alone; neither is written as a claim. Every er_v3 number is measured on survivors: the store holds no delisted name, so the bias is upward and its size is unmeasured. er_v1 no longer appears here; its claims are still in the table below. Not a recommendation: D-022's later stages need quote data this feed does not carry.
Every claim written today—
C2. Every claim for the as-of date, every horizon and every stream — not a ranking of them. The ranked lists above are a summary of this table and not a different measurement. A band appears only where the stream writes one: a momentum claim has an expected return and no band, and an empty cell there is the honest answer rather than a missing one.
Look up a name—
C4. Same JSON, same page — this reaches nothing. A name with no fair value shows the reason it has none, because a blank where a band should be is the one answer that tells the reader nothing.
Universe — fair value v1 against the close—
Too wide to inform — shown, and out of the ranking
Part B. A band whose p75/p25 is above 3 does not rank and is not in the distribution above — its gap would move a median the same panel calls uninformative. It still shows, with its v1 point and its close, because a name held out silently is indistinguishable from one dropped silently.
Carrying no fair value, and why
Cheapest 25 by v1's gap from the point
Fair value v1 is a point you can quote: the name's own ten-year median PE times its TTM EPS on the as-of. The zone is ±10% of the point; cheap below it, expensive above it, fair inside it and at either edge. The gap is measured from the point. Refused means the latest EPS is more than 50% from the median of the four before it with no matching price move — the reason is on the row. Trust is the stability of the name's PE against the Nifty 50's over ten years, times its history length, and is shown, not a gate. Projected 1y is the point grown by its 3-year EPS CAGR, capped at its own earlier best and worst. The old p25–p75 band is still computed and is not shown. The cheapest 25 is ranked by the gap from the point; a refused name or one with no point is held out of it and stays in the full table with its reason. The vendor's median PE is never an input. Basis UNVERIFIED: nothing in this system has established whether these figures are consolidated or standalone. Not a recommendation, and a cheap name is not a tradeable one: D-022's later stages need quote data this feed does not carry.