Mean Reversion Agent Skill

Use when writing a Bollinger-band mean-reversion strategy on Superior Trade — anything described as mean reversion, BB bands, oversold bounce, fade, range trade, ADX low, sigma extension. Upgraded 2026-05-18 from the prior 1h/2.5σ variant to the validated 4h/2σ/ADX<25 version (+8.77% multi-pair, 65.5% win over 162d). Prior 1h variant is preserved at the end of the file as an archived reference.

1k tokens

context cost

the whole folder, loaded on every use

files

instructions only

copies elsewhere

how many repositories repackaged it

225

stars on the repo

on the repository, not the skill itself

Install

one command, takes just this skill from the repository

npx skills add https://github.com/Superior-Trade/superior-skills --skill mean-reversion

The instruction itself

8 sections, as written by the author

Mean Reversion — Bollinger Reverter 4h

Note: This template was upgraded from the prior 1h / 2.5σ / ADX<30 version to the 4h / 2σ / ADX<25 version after backtesting showed the 4h variant produces meaningfully more trades with comparable risk and validated multi-pair edge. The prior 1h version is preserved at the end for reference.

Symmetric mean-reversion strategy on the 4h timeframe. Long-or-short on Bollinger band touches, gated to range regimes via ADX. Validated across BTC/ETH/SOL/DOGE over 162 days.

Backtest evidence

|---|---|---|---|---|

| BTC/USDC:USDC, 162d | 18 | 72% | +8.14% | 10% |

| BTC/USDC:USDC, range-regime sub-window (80d) | 8 | 100% | +9.88% | 0% |

| BTC/ETH/SOL/DOGE multi-pair, 162d | 84 | 65.5% | +8.77% | 18.5% |

Thesis

When the market is range-bound (ADX < 25), price touching the upper or lower Bollinger Band reliably reverts to the midline. Tight ROI takes profit fast since mean-reversion targets are small; tight stop closes positions that turn into trend breaks rather than reversions.

Mechanics

Timeframe: 4h
20-bar Bollinger Bands at 2σ
Entry short: close > bb_upper AND rsi > 65 AND adx < 25
Entry long: close < bb_lower AND rsi < 35 AND adx < 25
Exit: close crosses the band midline
Stop: -2%
ROI ladder: 2.5% → 1.5% → 0.5% → breakeven over 24h
No trailing stop (band reversion targets are small; ROI ladder handles take-profit)

Strategy code

from freqtrade.strategy import IStrategy
import pandas as pd
import talib.abstract as ta


class MeanReversionStrategy(IStrategy):
    INTERFACE_VERSION = 3
    timeframe = "4h"
    can_short = True

    stoploss = -0.02
    trailing_stop = False

    minimal_roi = {
        "0": 0.025,
        "240": 0.015,
        "720": 0.005,
        "1440": 0,
    }

    process_only_new_candles = True
    startup_candle_count = 60
    use_exit_signal = True

    def populate_indicators(self, dataframe: pd.DataFrame, metadata: dict) -> pd.DataFrame:
        bb = ta.BBANDS(dataframe, timeperiod=20, nbdevup=2.0, nbdevdn=2.0)
        dataframe["bb_upper"] = bb["upperband"]
        dataframe["bb_mid"] = bb["middleband"]
        dataframe["bb_lower"] = bb["lowerband"]
        dataframe["rsi"] = ta.RSI(dataframe, timeperiod=14)
        dataframe["adx"] = ta.ADX(dataframe, timeperiod=14)
        return dataframe

    def populate_entry_trend(self, dataframe: pd.DataFrame, metadata: dict) -> pd.DataFrame:
        cond_short = (
            (dataframe["close"] > dataframe["bb_upper"])
            & (dataframe["rsi"] > 65)
            & (dataframe["adx"] < 25)
        )
        dataframe.loc[cond_short, "enter_short"] = 1
        dataframe.loc[cond_short, "enter_tag"] = "bb_upper_revert"

        cond_long = (
            (dataframe["close"] < dataframe["bb_lower"])
            & (dataframe["rsi"] < 35)
            & (dataframe["adx"] < 25)
        )
        dataframe.loc[cond_long, "enter_long"] = 1
        dataframe.loc[cond_long, "enter_tag"] = "bb_lower_revert"
        return dataframe

    def populate_exit_trend(self, dataframe: pd.DataFrame, metadata: dict) -> pd.DataFrame:
        dataframe.loc[dataframe["close"] < dataframe["bb_mid"], "exit_short"] = 1
        dataframe.loc[dataframe["close"] > dataframe["bb_mid"], "exit_long"] = 1
        return dataframe

Reference config (multi-pair)

{
  "exchange": {
    "name": "hyperliquid",
    "pair_whitelist": ["BTC/USDC:USDC", "ETH/USDC:USDC", "SOL/USDC:USDC", "DOGE/USDC:USDC"]
  },
  "stake_currency": "USDC",
  "stake_amount": 75,
  "dry_run_wallet": {"USDC": 350},
  "timeframe": "4h",
  "max_open_trades": 4,
  "minimal_roi": {"0": 100.0},
  "stoploss": -0.02,
  "trading_mode": "futures",
  "margin_mode": "isolated",
  "entry_pricing": {"price_side": "same", "price_last_balance": 0.0},
  "exit_pricing": {"price_side": "same", "price_last_balance": 0.0},
  "pairlists": [{"method": "StaticPairList"}]
}

Honest framing

In strong-trend windows the strategy loses small (-1.75% on BTC during the first-half strong bear). In rangy windows it shines (+9.88% on BTC second-half). The mixed-regime full-period multi-pair number (+8.77% in 162d on $350 wallet) is the credible expectation.

DOGE was the negative pair (-0.65%) — meme volatility breaks more bands than reverts to them. Use this strategy on majors.

Pair with donchian-strong-regime for full-regime coverage.

Prior version (1h, 2.5σ, archived)

The previous version was tighter (2.5σ bands, ADX<30) on a 1h timeframe. Its own honest framing noted "5 trades in 4 months" — too rare to be useful. The 4h version produces ~3× the signal density with the same risk profile. The 1h version is preserved here for users who want a deeper-fade variant:

# Archived 1h variant — fewer, deeper signals
timeframe = "1h"
# bb = ta.BBANDS(dataframe, timeperiod=100, nbdevup=2.5, nbdevdn=2.5)
# rsi gates same; adx < 30 (looser)

If you prefer the rarer-but-deeper setup, restore the 1h timeframe and 2.5σ. The exit logic is unchanged.

Other skills for the same job

different authors, same section of the catalogue

Doc Coauthoring

by anthropics

vendor ×10

Guide users through a structured workflow for co-authoring documentation. Use when user wants to write documentation, proposals, technical specs, decision docs, or similar structured content. This workflow helps users efficiently transfer context, refine content through iteration, and verify the doc works for readers. Trigger when user mentions writing docs, creating proposals, drafting specs, or similar documentation tasks.

Automatically creates user-facing changelogs from git commits by analyzing commit history, categorizing changes, and transforming technical commits into clear, customer-friendly release notes. Turns hours of manual changelog writing into minutes of automated generation.

774 tokens

Test Driven Development

by w95

×7

Use when implementing any feature or bugfix, before writing implementation code

Use when you have a spec or requirements for a multi-step task, before touching code

Use when creating new skills, editing existing skills, or verifying skills work before deployment

26k tokens scripts

Crafting Effective Readmes

by softaworks

×3

Use when writing or improving README files. Not all READMEs are the same — provides templates and guidance matched to your audience and project type.

| Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Based on Wikipedia's inflated symbolism, promotional language, superficial -ing analyses, vague attributions, em dash overuse, rule of three, AI vocabulary words, negative parallelisms, and excessive conjunctive phrases.

6k tokens

Opentrons Integration

by christophacham

×3

Official Opentrons Protocol API for OT-2 and Flex robots. Use when writing protocols specifically for Opentrons hardware with full access to Protocol API v2 features. Best for production Opentrons protocols, official API compatibility. For multi-vendor automation or broader equipment control use pylabrobot.

9k tokens scripts

How to use it

Copy the folder

Take superior-trade/mean-reversion from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.