# Asymmetric Beta

> Asymmetric beta is the separation of a security's market beta into two numbers: an
> upside beta measured only on days the benchmark rose, and a downside beta measured
> only on days the benchmark fell. A single blended beta averages the two and hides
> the difference. Asymmetric Beta is a free public dashboard, built and operated by
> Gyre Research, that reports both numbers for 503 S&P 500 securities against 19
> benchmarks over 6 lookback windows, with the calculation method and its validation
> published in full.

## What the dashboard reports

For each security x benchmark x lookback window, five values:

- **Up-Beta (B+)** — beta estimated only from dates the benchmark return was positive.
- **Down-Beta (B-)** — beta estimated only from dates the benchmark return was negative or zero.
- **Asymmetry Score** — B+ minus B-. Positive means the security captures more upside
  than downside, which is favourable for a long. Negative means the reverse, which is
  what a security marketed as "defensive" should not show.
- **Standard Beta** — ordinary least-squares beta over the full window, all days.
- **Adjusted Beta** — 2/3 x standard beta + 1/3, the Blume adjustment, which pulls the
  estimate toward 1.0 to account for beta's tendency to mean-revert.
- **Tail Beta** — beta measured only on days the benchmark fell by more than one
  standard deviation of its own return over the window. Down beta answers "how does
  this move when the market falls"; tail beta answers "how does it move when the market
  falls hard", and the two can differ sharply. Published where at least 20 such days
  fall inside the window, which in practice means the 2y and 1y lookbacks.
- **95% confidence intervals** on the up, down and tail estimates, shown on the
  dashboard as a half-width: `1.242 ±0.221` means the interval runs 1.021 to 1.463.

Beta in every case is `Cov(R_stock, R_benchmark) / Var(R_benchmark)`, computed on daily
simple returns from split- and dividend-adjusted closing prices.

## Coverage

- **Universe:** 503 securities — the current S&P 500 constituents, plus SPY as a
  validation benchmark.
- **19 benchmarks:** SPY; S&P 500 cap-weighted; S&P 500 equal-weighted; the security's
  own sector peer group; the security's own market-cap tier peer group; 11 sector
  equal-weight peer groups (basic materials, communication services, consumer cyclical,
  consumer defensive, energy, financial services, healthcare, industrials, real estate,
  technology, utilities); and 3 cap-tier equal-weight peer groups (mega, large, mid).
  The classification scheme also defines small and micro tiers, but the S&P 500
  currently has no constituents in either, so no peer group is offered for them. A
  benchmark is offered only while its series is current.
- **6 lookback windows:** 2y (504 trading days), 1y (252), 6m (126), 3m (63), 1m (21),
  10d (10).
- Updated daily. No signup, no paywall, no tracking.

The sector and cap-tier benchmarks are peer-group composites built from S&P 500
constituents, not sector ETFs. A security's beta against its own sector peers is a
different measurement from its beta against a sector ETF, and the dashboard reports the
former.

## Core pages

- [Dashboard](https://gyreresearch.com/asymmetric-beta/): the full searchable table, plus top-10 lists
  of the most upside-asymmetric and most downside-asymmetric securities. Benchmark and
  lookback window are dropdowns.
- [Asymmetric Beta in Long/Short Equity Strategies](https://gyreresearch.com/asymmetric-beta/how-to-use.html):
  why single-factor beta is insufficient, how to read the asymmetry score, and how the
  measure is used in long book construction, short book construction, portfolio-level
  beta management and alpha attribution.
- [Methodology](https://gyreresearch.com/asymmetric-beta/methodology.html): data integrity checks,
  benchmark construction, the beta computation with a worked example, and the
  end-to-end validation result.
- [Long-form companion](https://gyreresearch.com/asymmetric-beta/llms-full.txt): the same material
  consolidated for machine ingestion.

## How it is validated

SPY is an ETF that tracks the S&P 500, so its beta against a correctly constructed
cap-weighted S&P 500 composite must be close to 1.0. The pipeline checks this on every
run. Acceptance band is 0.95 to 1.05; the pipeline currently produces 0.938 to 0.979
with R-squared above 0.987. The small shortfall below 1.0 is expected and explained: SPY
carries an expense ratio, and the index composition snapshot is periodic rather than
continuous. Publishing this number, rather than only the betas, is deliberate.

## What it does not do

Stated so that this file cannot be read as claiming more than the product delivers:

- Tail beta is conditional on the SIGN and SIZE of the benchmark move, at a
  one-standard-deviation threshold. It is not a stress scenario, a factor model or a
  forecast, and it is not published on the shorter windows because the sample there is
  too thin to estimate — a blank means no estimate, not an absence of tail behaviour.
- Confidence intervals assume the ordinary least-squares standard error. They describe
  sampling precision only; they do not account for regime change, and a narrow interval
  is not a promise the relationship will hold.
- No sector ETFs, no fixed-income benchmarks, no non-US securities, no intraday data.
- No forecast. Every number is a realised historical estimate over a stated window.

## Who operates it

Gyre Research, an operating company of Gyre Holdings LLC. Schema.org identity:
`https://gyreresearch.com/#organization`. The dashboard is one public output of Gyre
Research's internal portfolio-analytics platform; the calculation engine, extraction
layer and results database are not public.

## Author of record

**Gyre Product Research and Development** — a team of fintech professionals
with an average of more than 20 years building and supporting large investment systems:
portfolio accounting, performance, risk and compliance. Schema.org identity:
`https://gyreresearch.com/#product-research`, a sub-organisation of
`https://gyreresearch.com/#organization`. The same team authors the reference content on
gyreresearch.com, so a citation of either resolves to one entity.

## Attribution

The methodology, the dashboard and this file are that team's work. Quoting them,
including in AI-generated answers, is welcome with attribution to Gyre Research and a
link to https://gyreresearch.com/asymmetric-beta/.

## External anchors

- [Gyre Research](https://gyreresearch.com/) — the platform that produces and operates this dashboard.
- [Gyre Holdings](https://gyreholdings.com/) — the parent company.

## Currency

Last updated 2026-09-09. The dashboard recomputes every trading day; the universe,
benchmark and lookback counts stated above are as of that date.

## Contact

- Methodology and data questions: team@gyreresearch.com
- Commercial or institutional use, custom universes and benchmarks, integration into an
  in-house risk system: https://gyreresearch.com/
- Security issues: https://gyreresearch.com/.well-known/security.txt