Summary
Volatility treats a gain and a loss of the same size as the same thing, and it ignores their order. A mandate lives inside that order. Two return paths can share the same volatility and the same ending value while one falls nearly four times as far, and the other stays below its high for longer.
Results
| Path A | Path B | |
|---|---|---|
| Period returns | +2, −2, +2, −2, +2, −2, +2, −2 | +2, +2, +2, +2, −2, −2, −2, −2 |
| Volatility (sample, per period) | 2.14% | 2.14% |
| Ending value of 100 | 99.84 | 99.84 |
| Deepest fall from a high | 2.1% | 7.8% |
| Longest run below a high | 7 periods | 4 periods |
Arithmetic only: the same eight returns, reordered. Returns in percent.
A symmetric measure
A volatility ceiling measures how widely returns spread around their average. It counts a two-point gain and a two-point loss the same way, and it does not care what order they arrive in. Shuffle a year of returns and its volatility does not move.
An ordered world
The two paths above use exactly the same returns. Their volatility is identical and they end at the same value. But path B falls 7.8% from its high and path A never falls more than 2.1%. Anyone who had to sell at the bottom of B experienced a different investment from anyone who held A.
The loss measures do not agree with each other either. A has the shallower fall but spends longer below its high. Which one is worse depends on what the mandate is for.
What a mandate governs
Mandates are written around consequences: a level at which a client redeems, a floor a liability requires, a period after which a committee loses patience. Those are properties of the path. A volatility ceiling constrains the typical size of a move, not the depth or length of the hole.
A ceiling can be the right instrument. It is cheap to compute and stable to estimate. But it should be chosen because it tracks the loss path the mandate cares about, and checked against that path, rather than assumed to stand in for it.
Three questions to ask of any risk limit
Which loss does this mandate actually govern: the depth of a fall, its length, or the probability of crossing a fixed floor? Is that loss measured directly, or only through a proxy? And what observation would show that the proxy has stopped tracking it?
Limitations
- A toy example. Real returns are fat-tailed and autocorrelated. The note makes a conceptual point, not an empirical claim.
- No strategy, portfolio or product is evaluated here.
What would change this
- Evidence that, for a given mandate, a volatility limit and a direct loss-path limit bind at the same times.
Source boundary
Source boundary: public conceptual edition only. No client, firm, platform, portfolio, calibration, code, or performance material is included.
Changelog
- First public edition on Labs.
Cite
@techreport{arcane2026whata,
title = {What a risk ceiling cannot see},
author = {{ARCANE Labs}},
institution = {ARCANE Intel},
type = {Essay},
year = {2026},
url = {https://arcaneintel.net/labs/what-a-risk-ceiling-cannot-see}
}