FiClaw
-
Financial Agent Platform
中文

Mean Reversion

A-share mean reversion: defining and testing oversold rebound signals

Mean reversion seeks price moves that may retrace after a measurable deviation from a recent range. 'Buy the dip' is not a strategy: deviation, entry time, exit, maximum holding period and loss control all require zero-discretion rules.

Typical holding period

Often days to weeks, with a mandatory maximum holding period.

Rebalance and trigger

Signals can be checked daily, but trade frequency must survive net cost.

Research universe

Liquid stocks with explicit abnormal-state and event exclusions.

Research hypothesis

Write the hypothesis before reading the backtest

For liquid names without persistent fundamental deterioration, short-term overreaction may partly reverse. In a trend, after a major event or through repeated limit-down days, the deviation can continue to widen.

This guide is for research education and backtest design. It is not investment advice or a return promise; results depend on sample, cost, regime and execution.

Strategy rules

Turn the idea into zero-discretion rules

Standardised deviation

Define an extreme with a z-score, Bollinger band or moving-average distance.

Turns 'oversold' into a repeatable condition.

Confirmation filter

Filter on volume, volatility or market regime.

Reduces mechanical entries into persistent trends.

Exit rule

Set a reversion target, risk exit and time exit.

Prevents indefinite waiting for a rebound.

Implementation

From signal to portfolio

  1. 1State whether signals use close, open or intraday data and only trade after the signal is available.
  2. 2Codify entry, reversion exit, risk exit and time exit exactly.
  3. 3Record trigger reason, holding days, slippage and unfilled quantity for every trade.
  4. 4Distinguish stock, within-sector and index-relative reversion rather than mixing their interpretations.

A-share constraints

Execution belongs in the test

  • A limit-down stock may not be sellable, so a close-based stop can be fictional.
  • Corporate actions, resumptions and restructurings can create price jumps that require event rules.
  • Cash equities are T+1: a share bought today cannot be sold today.

Validation checklist

Try to break the strategy first

Prefer stable regions that survive costs, regime changes and parameter perturbation to a single maximum return.

Grid nearby deviation, holding-period and risk-exit values and reject narrow optima.
Report trade count and expectancy by volatility regime, sector and market trend.
Add conservative trade-level slippage and check whether a few rebound days drive the result.
Validate on a later period that was not reused for parameter selection.

Risk map

How the strategy can fail

Trend continuationConsecutive limit-down daysHigh turnoverStructural event gaps

Research prompt

A first backtest specification

“In a liquid A-share universe, test 20-day Bollinger-band mean reversion: buy at the next open after a close below the lower band, exit at the middle band or after 10 trading days, with a fixed risk exit. Enforce T+1, price limits, suspensions, corporate actions and conservative slippage; report trade samples, P&L distribution, regime splits and parameter sensitivity.”

FAQ

Common questions

Can RSI alone define mean reversion?

RSI can be one signal, but it does not resolve trend, liquidity or execution risk. Test it with fully specified entry, exit and cost rules.

Does a risk exit contradict mean reversion?

No. It bounds tail exposure. The important test is whether results remain stable across a reasonable range of exit settings.

Next research step

Fix research conventions before the backtest

Use the related factors, data boundaries and out-of-sample review to turn a strategy idea into an auditable research record.

Open A-share factor libraryReview A-share data boundaries