Ladder Expected Value (rt-ladder-ev)

rt-journal-match returns a shortlist and an order. This answers the question that order implies but never states: what does that sequence cost?

Authors compare venues one at a time — is this one worth a shot? — and in isolation the answer is almost always yes. The cost only appears in the sequence, and it is a trade between months and placement probability that nothing else in this repository made visible, so it was settled by optimism.

The size of the trade is whatever your inputs say it is; the point of the tool is that you find out before spending the months rather than after. In the worked example, one reach rung costs about four and a half months and buys about four points of placement probability — and the sensitivity band shows the four points are not distinguishable from zero while the four months are. That shape of answer, rather than a winner, is the usual output.

When to trigger

• Two candidate submission orders and no principled way to choose.
• A clock: job market, tenure case, grant report, a co-author's graduation.
• A paper that has already been rejected twice and needs the remaining ladder costed.
• Someone asks "is it worth trying X first?" — that is a sequence question.

What it needs

The paper-profile.yml (for ambition, constraints.clock, history) plus, for each rung, three numbers:

p_accept is not the published acceptance rate. A venue's 6% is computed over a submission pool that includes everything sent to it. A clean design with a general-interest result is not a random draw from that pool, and neither is a thin one. Start from the published rate, then move it with the venue's own *-topic-selection fit judgement and rt-desk-reject-risk output, and say which way you moved it and why.

What it does

python3 tools/ladder_ev.py \
    --rung "Journal of Finance:0.05:4.5" \
    --rung "Review of Financial Studies:0.08:5.0" \
    --rung "JFQA:0.20:3.5" \
    --rung "Journal of Banking and Finance:0.35:2.5"

Walks the ladder top-down carrying the probability the paper is still unplaced, and returns: time until the ladder resolves, time to print conditional on placing, the probability of placing at all, and — the number that changes minds — the probability of running the ladder out and having nowhere left to go.

Then run the alternative sequence and compare. The comparison is the deliverable, not either number on its own.

Hard rules

1. Turnaround and acceptance figures come from the source map, read at the time of use. They are volatile; this skill stores none of them.
2. Report the band, not the point. p_accept is a judgement, so the tool prints a ±40% sensitivity band by default. If two ladders' bands overlap, say they are indistinguishable — do not rank them anyway on the third decimal place.
3. A ladder with no floor is not a plan. If the probability of exhausting the ladder exceeds ~25%, the shortlist is missing a credible home; go back to rt-journal-match for a safe rung rather than reporting a number.
4. Never present the output as a forecast. It is arithmetic over stated assumptions. State the assumptions next to the answer.

Output format

【Ladder A】V1 → V2 → V3   resolves in N months · places P% · exhausts E%
【Ladder B】V2 → V3        resolves in N months · places P% · exhausts E%
【Difference】what B buys or costs vs A, in months and in placement probability
【Sensitivity】whether the difference survives the ±40% band
【Assumptions】each p_accept, and why it differs from the published rate
【Recommendation】which sequence, and the one fact that would change it

Anti-patterns

• Using published acceptance rates as p_accept — that is the pool's number, not the paper's.
• Costing a ladder whose rungs were never checked for fit; a fast rung that will desk- reject the paper on scope is not a rung. Run rt-journal-match first.
• Optimising time-to-print alone. A worse-placed paper can cost more career-years than the months it saved — ambition in the profile is what balances that, and it belongs in the write-up.
• Reporting one decimal place of expected months as though it were measured.

Follows rt-journal-match (which builds the ladder) and rt-venue-reframe (which prices the rewrite each rung needs). Method: journal-match.md step 6.