# source: https://raw.githubusercontent.com/AtanasIvanov07/AiTradingBot/86b0c906a6b14f62d11368a8ee10056c4de2f189/user_data/strategies/h5_trend_vol.py # directory_url: https://github.com/AtanasIvanov07/AiTradingBot/blob/main/user_data/strategies/ # User: AtanasIvanov07 # Repository: AiTradingBot # --------------------"""H5: H4's trend filter with volatility sizing, fixed at N=150 and 30% target vol. Learning-only path (PROJECT_PLAN.md amendment A1): dry-run only, not evidence of an edge. H5 was killed at Gate 1 (research/hypotheses/H5.md). The rule mirrors the research code (``aitradingbot.analysis.strategies.h4_trend_vol`` and the engine's sleeve logic). This file is self-contained, because the Docker image has no ``aitradingbot`` package. ``tests/test_freqtrade_strategy.py`` checks the two stay identical. Each pair is an independent sleeve worth 50% of the starting capital plus that pair's own P&L. Decisions use the last closed daily candle; orders go out at the next candle's open: - target exposure = trend (1/0: SMA(150) with a 2% band) × min(1, 30% / 30-day realised vol); - enter when flat and the target is above the 10-point rebalance band; - resize once per candle, when the target and the current exposure differ by more than 10 points; - exit fully when the target drops to 0 (trend off). Differences from the research engine, accepted for this path: - Freqtrade fills missing candles (previous close, zero volume); the research engine skips them. - The current exposure includes Freqtrade's estimate of the exit fee. - Exchange minimum orders apply (€5 on Bitvavo), so small resizes can be skipped. """ from __future__ import annotations import math from datetime import datetime from typing import Any import numpy as np import pandas as pd from freqtrade.persistence import Trade from freqtrade.strategy import IStrategy from pandas import DataFrame N = 150 BAND = 0.02 TARGET_VOL = 0.30 VOL_LOOKBACK = 30 PERIODS_PER_YEAR = 365 REBALANCE_BAND = 0.10 SLEEVE = 0.5 CHECKED_KEY = "resize_checked_candle" def target_exposure(close: pd.Series) -> DataFrame: """Trend state, volatility size and their product, from closes up to each candle only.""" sma = close.rolling(N, min_periods=N).mean() state = pd.Series(np.nan, index=close.index) state[close > sma * (1 + BAND)] = 1.0 state[close < sma * (1 - BAND)] = 0.0 trend = state.ffill().fillna(0.0) realised = np.log(close).diff().rolling(VOL_LOOKBACK, min_periods=VOL_LOOKBACK).std() size = (TARGET_VOL / (realised * math.sqrt(PERIODS_PER_YEAR))).clip(upper=1.0) size = size.replace([np.inf, -np.inf], 1.0).fillna(0.0) return DataFrame({"trend": trend, "size": size, "target": trend * size}, index=close.index) class Github_AtanasIvanov07_AiTradingBot__h5_trend_vol__20261025_172517(IStrategy): INTERFACE_VERSION = 3 timeframe = "1d" can_short = False minimal_roi = {"0": 1000.0} # no take-profit: only the trend rule exits stoploss = -0.99 # no stop-loss: the card exits on daily closes only trailing_stop = False use_exit_signal = True exit_profit_only = False ignore_roi_if_entry_signal = False # Market orders, as the research engine assumed (taker fill at the open). Freqtrade requires # entry/exit pricing price_side "other" for these (config.json). order_types = { "entry": "market", "exit": "market", "emergency_exit": "market", "force_entry": "market", "force_exit": "market", "stoploss": "market", "stoploss_on_exchange": False, } process_only_new_candles = True position_adjustment_enable = True max_entry_position_adjustment = -1 # SMA(150) plus room for the band state to settle; recursive-analysis checks this. startup_candle_count = 400 def populate_indicators(self, dataframe: DataFrame, metadata: dict[str, Any]) -> DataFrame: exposure = target_exposure(dataframe["close"]) dataframe["trend"] = exposure["trend"] dataframe["size"] = exposure["size"] dataframe["target"] = exposure["target"] return dataframe def populate_entry_trend(self, dataframe: DataFrame, metadata: dict[str, Any]) -> DataFrame: dataframe.loc[dataframe["target"] > REBALANCE_BAND, ["enter_long", "enter_tag"]] = ( 1, "trend_on", ) return dataframe def populate_exit_trend(self, dataframe: DataFrame, metadata: dict[str, Any]) -> DataFrame: dataframe.loc[dataframe["target"] <= 0, ["exit_long", "exit_tag"]] = (1, "trend_off") return dataframe # --- sizing --------------------------------------------------------------------------------- def last_candle(self, pair: str) -> pd.Series | None: """The last closed candle (in backtesting, the one before the candle being traded).""" dataframe, _ = self.dp.get_analyzed_dataframe(pair, self.timeframe) if dataframe.empty: return None return dataframe.iloc[-1] def starting_capital(self) -> float: """Fixed capital from the config. Not ``wallets.get_starting_balance()``: that one already includes the open trades' realised profit, which ``sleeve_equity`` adds itself. """ capital = self.config.get("available_capital", self.config.get("dry_run_wallet", 0.0)) if isinstance(capital, dict): # multi-currency dry-run wallet capital = capital.get(self.config["stake_currency"], 0.0) return float(capital) def sleeve_equity(self, pair: str, rate: float) -> float: """50% of the starting capital plus this pair's realised and unrealised P&L.""" equity = SLEEVE * self.starting_capital() for trade in Trade.get_trades_proxy(pair=pair, is_open=False): equity += trade.close_profit_abs or 0.0 for trade in Trade.get_trades_proxy(pair=pair, is_open=True): equity += trade.realized_profit + trade.calc_profit(rate) return equity def custom_stake_amount( self, pair: str, current_time: datetime, current_rate: float, proposed_stake: float, min_stake: float | None, max_stake: float, leverage: float, entry_tag: str | None, side: str, **kwargs: Any, ) -> float: candle = self.last_candle(pair) if candle is None: return 0.0 stake = float(candle["target"]) * self.sleeve_equity(pair, current_rate) return min(stake, max_stake) def adjust_trade_position( self, trade: Trade, current_time: datetime, current_rate: float, current_profit: float, min_stake: float | None, max_stake: float, current_entry_rate: float, current_exit_rate: float, current_entry_profit: float, current_exit_profit: float, **kwargs: Any, ) -> tuple[float | None, str | None] | None: candle = self.last_candle(trade.pair) if candle is None or trade.has_open_orders: return None candle_date = str(candle["date"]) if trade.get_custom_data(CHECKED_KEY) == candle_date: return None # once per candle, as in research: not on every intraday loop trade.set_custom_data(CHECKED_KEY, candle_date) target = float(candle["target"]) if target <= 0: return None # the exit signal closes the whole position equity = self.sleeve_equity(trade.pair, current_rate) held = trade.amount * current_rate if equity <= 0 or abs(target - held / equity) <= REBALANCE_BAND: return None delta = target * equity - held if delta > 0: return min(delta, max_stake), "resize_up" # Freqtrade sells |stake| × amount / trade.stake_amount, i.e. a share of the cost basis, # not of market value. Convert, so that `delta` euros of market value are sold. return delta * trade.stake_amount / held, "resize_down"