# Tail risk

Tail risk is the possibility of an unusually large gain or loss in the extreme ends of a return distribution, with downside tail risk referring specifically to severe losses.

## What it means

Most daily stock returns fall within a relatively ordinary range. Tail outcomes sit at the extremes of the return distribution: unusually large gains on the right and unusually large losses on the left. In a risk discussion, *tail risk* usually means the downside, or left, tail.

A sharp overnight price gap is one possible source of a tail loss. An earnings announcement, trading halt or broad market shock can also make an extreme outcome more relevant. These are possibilities, not predictions about any particular stock.

The size of a loss and its likelihood are different questions. A rare loss can matter greatly even when the average daily move looks modest. Estimates are uncertain because extreme events are scarce in historical data, and future conditions may differ from the period measured.

## How it is calculated

Tail risk has no single universal formula. One way to describe downside severity is **expected shortfall** at a chosen confidence level: the average loss in the worst fraction of modeled outcomes. Its associated **value at risk** (VaR) is a cutoff, not the average loss beyond that cutoff.

For a hypothetical set of equally likely outcomes, a simple calculation is `ES_c = (L_1 + L_2 + ... + L_k) / k`, where:

- `ES_c` is expected shortfall at confidence level `c`, measured in US$ when the losses are in US$.
- `c` is the confidence level, expressed as a decimal; at 95%, `c = 0.95`.
- `L_1` through `L_k` are the losses in the worst `k` outcomes, measured in US$.
- `k` is the number of outcomes in the worst `1 − c` fraction of the sample. With 100 equally likely outcomes and `c = 0.95`, `k = 5`.

State the time horizon alongside the result. A one-day estimate and a one-month estimate answer different questions. With unequal outcome probabilities, use probability-weighted tail losses rather than this simple average.

## Worked example

Suppose a hypothetical US$10,000 stock position has 100 equally likely, modeled **one-day** outcomes. For simplicity, assume its five worst outcomes are losses of US$400, US$500, US$600, US$800 and US$1,200, with the remaining outcomes all less severe.

1. Choose 95% confidence. The worst 5% of 100 outcomes contains five outcomes.
2. Add their losses: US$400 + US$500 + US$600 + US$800 + US$1,200 = US$3,500.
3. Divide by five: the estimated one-day 95% expected shortfall is **US$700**, or **7%** of the US$10,000 position.

This does not mean a US$700 loss is expected every day, or that US$1,200 is the largest possible loss. The calculation describes this hypothetical set of modeled outcomes only. A 95% VaR cutoff would answer a different question about the boundary of the worst 5%, not their average severity.

## Why it matters before you trade

A pre-trade check can consider whether an unusually large loss would be manageable for the proposed position or basket. Position size, concentration and exposure to the same event across several stocks can affect the US$ impact. Diversification may reduce some stock-specific exposure but does not eliminate market-wide tail events.

Ordinary volatility and a one-day 95% VaR can help frame typical variation and a loss threshold under their assumptions. Neither guarantees a cap on losses beyond that threshold. Thin liquidity may also make an extreme move harder to trade through at an expected price.

## How PreTrAIde uses it

PreTrAIde does not list a dedicated tail-risk or expected-shortfall calculation among its stated quantitative measures. Its pre-trade analysis does calculate one-day 95% VaR, daily volatility, liquidity measures and basket concentration, which can provide context for discussing downside exposure. VaR does not measure how severe losses may be beyond its cutoff.

For a proposed stock basket, PreTrAIde returns GO, NO-GO or MODIFY verdicts, risk assessments and plain-language reasoning. These are decision support, not investment recommendations or guarantees about extreme outcomes.

## Common mistakes

- **Treating tail risk as only a large standard deviation:** Two return distributions can have similar everyday volatility but different extreme losses.
- **Reading 95% VaR as a maximum loss:** Losses beyond the cutoff can be much larger.
- **Assuming the past contains every possible shock:** Historical samples can miss rare or unprecedented events.
- **Adding individual stock tail losses without considering dependence:** Stocks may move together during stress, changing basket-level losses.

---
Source: https://pretraide.com/glossary/tail-risk