# source: https://raw.githubusercontent.com/IvPalmer/Master-Trader/c7955a15d4bfe82e567bc2bd315a372000a442fd/ft_userdata/user_data/strategies/MasterTraderAI.py # directory_url: https://github.com/IvPalmer/Master-Trader/blob/main/ft_userdata/user_data/strategies/ # User: IvPalmer # Repository: Master-Trader # --------------------""" MasterTrader AI - FreqAI Adaptive Trading Strategy LightGBM regression with regime detection for spot crypto trading. Uses FreqAI to predict mean price change over the next N candles, combined with ATR/ADX regime filters to avoid unfavorable conditions. Designed for Freqtrade 2026.2 with FreqAI enabled. """ from datetime import datetime from freqtrade.strategy import IStrategy, DecimalParameter from pandas import DataFrame import talib.abstract as ta import numpy as np class Github_IvPalmer_Master_Trader__MasterTraderAI__20260923_012354(IStrategy): INTERFACE_VERSION = 3 # --- Timeframe & general settings --- timeframe = "5m" can_short = False process_only_new_candles = True use_exit_signal = True startup_candle_count = 100 # --- ROI table (conservative) --- minimal_roi = { "0": 0.03, "30": 0.02, "60": 0.01, "120": 0.005, } # --- Stoploss --- stoploss = -0.10 # Widened from -0.05: FreqAI adapts, but 5% too tight for alt volatility # --- Trailing stop --- trailing_stop = True trailing_stop_positive = 0.01 trailing_stop_positive_offset = 0.02 trailing_only_offset_is_reached = True @property def protections(self): return [ {"method": "CooldownPeriod", "stop_duration_candles": 5}, {"method": "StoplossGuard", "lookback_period_candles": 48, "trade_limit": 4, "stop_duration_candles": 12, "only_per_pair": True}, {"method": "LowProfitPairs", "lookback_period_candles": 288, "trade_limit": 4, "stop_duration_candles": 48, "required_profit": -0.05}, {"method": "MaxDrawdown", "lookback_period_candles": 576, "max_allowed_drawdown": 0.20, "stop_duration_candles": 72, "trade_limit": 1}, ] # --- Max open trades --- max_open_trades = 0 # WIND DOWN: no new entries, let open trades resolve # --- Wind-down stoploss: tighten to just below 24h lows --- use_custom_stoploss = True def custom_stoploss(self, pair: str, trade: 'Trade', current_time: datetime, current_rate: float, current_profit: float, **kwargs) -> float: """ Wind-down mode: set stops just below 24h low for each pair. If price breaks these levels, cut the loss. Otherwise give room for a bounce. These are absolute price levels translated to ratio from current_rate. """ # 24h lows as of 2026-03-13 ~18:00 UTC, with ~0.3% buffer below wind_down_stops = { "ADA/USDT": 0.2595, # 24h low 0.2603, stop -0.3% below "SOL/USDT": 85.42, # 24h low 85.68, stop -0.3% below "AVAX/USDT": 9.49, # 24h low 9.52, stop -0.3% below "ETH/USDT": 2039.0, # 24h low 2045.42, stop -0.3% below "BNB/USDT": 644.5, # 24h low 646.42, stop -0.3% below } if pair in wind_down_stops: stop_price = wind_down_stops[pair] # Convert absolute price to ratio from open_rate stoploss_ratio = (stop_price / trade.open_rate) - 1.0 # Never widen beyond the base stoploss return max(stoploss_ratio, self.stoploss) return self.stoploss # --- FreqAI signal thresholds (hyperoptable) --- entry_threshold = DecimalParameter( 0.0005, 0.01, default=0.003, decimals=4, space="buy", optimize=True ) exit_threshold = DecimalParameter( -0.01, 0.0, default=-0.001, decimals=4, space="sell", optimize=True ) # ------------------------------------------------------------------ # # FreqAI Feature Engineering # # ------------------------------------------------------------------ # def feature_engineering_expand_all( self, dataframe: DataFrame, period: int, metadata: dict, **kwargs ) -> DataFrame: """ Features expanded across ALL config dimensions: timeframes, indicator_periods_candles, corr pairs, shifted candles. The `-period` suffix is auto-replaced with each value from indicator_periods_candles in the FreqAI config. """ # Momentum / trend strength dataframe["%-rsi-period"] = ta.RSI(dataframe, timeperiod=period) dataframe["%-mfi-period"] = ta.MFI(dataframe, timeperiod=period) dataframe["%-adx-period"] = ta.ADX(dataframe, timeperiod=period) # Trend direction dataframe["%-sma-period"] = ta.SMA(dataframe, timeperiod=period) dataframe["%-ema-period"] = ta.EMA(dataframe, timeperiod=period) # Rate of change dataframe["%-roc-period"] = ta.ROC(dataframe, timeperiod=period) # Volatility - Bollinger Band width bollinger = ta.BBANDS(dataframe, timeperiod=period, nbdevup=2.0, nbdevdn=2.0) dataframe["%-bb_width-period"] = ( (bollinger["upperband"] - bollinger["lowerband"]) / bollinger["middleband"] ) # Relative volume dataframe["%-relative_volume-period"] = ( dataframe["volume"] / dataframe["volume"].rolling(window=period).mean() ) return dataframe def feature_engineering_expand_basic( self, dataframe: DataFrame, metadata: dict, **kwargs ) -> DataFrame: """ Features expanded across timeframes, corr pairs, and shifted candles, but NOT across indicator_periods_candles. """ dataframe["%-pct-change"] = dataframe["close"].pct_change() dataframe["%-raw_volume"] = dataframe["volume"] return dataframe def feature_engineering_standard( self, dataframe: DataFrame, metadata: dict, **kwargs ) -> DataFrame: """ Standard features with no expansion -- base timeframe only. Calendar features to capture intraday/weekly seasonality. """ dataframe["%-day_of_week"] = dataframe["date"].dt.dayofweek / 6.0 dataframe["%-hour_of_day"] = dataframe["date"].dt.hour / 23.0 return dataframe # ------------------------------------------------------------------ # # FreqAI Targets # # ------------------------------------------------------------------ # def set_freqai_targets( self, dataframe: DataFrame, metadata: dict, **kwargs ) -> DataFrame: """ Target: mean percentage price change over the next `label_period_candles` candles. """ label_period = self.freqai_info["feature_parameters"]["label_period_candles"] dataframe["&-s_close"] = ( dataframe["close"] .shift(-label_period) .rolling(window=label_period) .mean() / dataframe["close"] - 1 ) return dataframe # ------------------------------------------------------------------ # # Indicator Population # # ------------------------------------------------------------------ # def populate_indicators( self, dataframe: DataFrame, metadata: dict ) -> DataFrame: """ Trigger FreqAI training/prediction, then add regime-detection indicators that the strategy (not the model) uses for filtering. """ # --- FreqAI: train models & generate predictions --- dataframe = self.freqai.start(dataframe, metadata, self) # --- Regime detection (used in entry/exit filters, NOT as ML features) --- # ATR regime: detect volatility spikes / crash conditions dataframe["atr"] = ta.ATR( dataframe, timeperiod=14 ) dataframe["atr_sma"] = ( dataframe["atr"].rolling(window=50).mean() ) # ADX: trend strength (trending vs ranging market) dataframe["adx"] = ta.ADX(dataframe, timeperiod=14) return dataframe # ------------------------------------------------------------------ # # Entry Signal # # ------------------------------------------------------------------ # def populate_entry_trend( self, dataframe: DataFrame, metadata: dict ) -> DataFrame: """ Enter long when: 1. FreqAI predicts upside above threshold 2. FreqAI model is confident (do_predict == 1) 3. Market is not in a crash regime (ATR < 2x its SMA) 4. Some directional trend exists (ADX > 20) 5. Volume is non-zero """ dataframe.loc[ ( (dataframe["&-s_close"] > self.entry_threshold.value) & (dataframe["do_predict"] == 1) & (dataframe["atr"] < 2.0 * dataframe["atr_sma"]) & (dataframe["adx"] > 20) & (dataframe["volume"] > 0) ), "enter_long", ] = 1 return dataframe # ------------------------------------------------------------------ # # Exit Signal # # ------------------------------------------------------------------ # def populate_exit_trend( self, dataframe: DataFrame, metadata: dict ) -> DataFrame: """ Exit long when: - FreqAI predicts downside below threshold - OR a volatility spike occurs (ATR > 2.5x its SMA) Always require non-zero volume. """ dataframe.loc[ ( (dataframe["&-s_close"] < self.exit_threshold.value) | (dataframe["atr"] > 2.5 * dataframe["atr_sma"]) ) & (dataframe["volume"] > 0), "exit_long", ] = 1 return dataframe