ep-curve
EvidenceCanvasSeries
An exceedance-probability curve: the loss to expect once every N years, with return periods across on a log scale and up to four curves (OEP and AEP, gross and net).
Use when the question is how big a loss to plan for: a reinsurance program's 1-in-100 and 1-in-250 losses, a cat model's output, a capital model's tail. Each point is {return period, loss}; a curve is a name with its points nested under it.
riskmetricassessmentboard-deck
When to use
- How big a loss to plan for. An EP curve earns its slide when the question is the size of a rare loss: the 1-in-100 a program must cover, how much a reinsurance tower takes off the gross curve. For how often each loss band happens, use
histogram. - Two to four curves. OEP and AEP, or gross and net, read best together: the gap between the curves is the story. Past four, split the slide.
When not to use
- Return period written as a label.
- 1-in-100$410M`is a label and a value, which the chart cannot place. Write the point record, return period first:; the axis prints1-in-100` for you. - A probability instead of a return period. Write the return period, not the annual probability: 1% is
{100, …}, 0.4% is{250, …}.
Common mistakes
- Writing the loss first (
{$410M, 100}). The record is{return period, loss}: the return period runs across the bottom, the way the chart is drawn.
Authoring
<!-- _class: ep-curve -->
## A 1-in-100 year costs $410M.
- `{10, $120M}`
- `{50, $300M}`
- `{100, $410M}`
- `{250, $560M}`Slots
| Slot | Selector | Required | Description |
|---|---|---|---|
title | h2 | yes | Slide heading — name the loss the audience should remember ('A 1-in-100 year costs $410M'), not 'EP curve'. |
axes | p > code | no | Optional axis line, as ONE bracketed list in its own paragraph ABOVE the heading: the return-period axis, then the loss axis — [{Return period, 10..250}, {Loss}]. A return-period domain of at least 1 widens the axis past the points (a minimum below 1 is ignored); a loss domain raises the ceiling, while the floor stays zero or the lowest loss. The return-period axis is always log, so log changes nothing. Without the line the x axis is captioned 'Return period'. Same idiom as scatter and histogram. |
curves | ul > li | yes | A flat list of point records draws one curve: ` {100, $410M} is a 1-in-100 year loss of $410M, return period first. For several curves, write each curve's name and nest its points under it: - OEP then - {100, $410M}`. Up to four curves; a return period below 1 is not a point. |
Data shape
- Each point is
{return period, loss}: the return period a number of years of at least 1, the loss any number Segno reads ($410M,410). - A curve's points are drawn in return-period order whatever the order written; a return period given twice in one curve keeps the first.
lint:decknames a point it cannot read (a thousands separator,{1,000, $900M}, is the commonest), a return period given twice in one curve, and a fifth curve.
Anatomy
┌─────────────────────────────────────────┐
│ header │
│ Where the pipeline leaks. │
│ │
│ Visitors [============] 12,000 │
│ 62% [========] 4,800 │
│ Activated [=====] 2,160 │
│ 40% [==] 864 │
│ Paid [=] 670 │
│ │
│ footer 1/19 │
└─────────────────────────────────────────┘