beast/beast-trader-strategies
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

v0.2: 增加ATR动态止损 + 多时间框架结构分析

Browse Source
This commit is contained in:
Beast Trader
2026-06-07 22:35:00 +08:00
parent fd6afcfb38
commit bd293d2f5a
1 changed files with 64 additions and 345 deletions
Download Patch File Download Diff File Expand all files Collapse all files

409
strategy.py
View File

@ -2,7 +2,7 @@
多时间框架价格行为策略 — ETH/USDT 中低频交易
==============================================
设计理念 (v0.3)：
设计理念 (v0.2)：
1. 反转大多会失败 → 不做反转预测，只做趋势延续。
在 S/R 位入场不是赌反弹，是赌"回调结束、趋势恢复"。
@ -16,7 +16,17 @@
核心原则：只在大趋势方向上，在关键位置，等确认信号入场。
版本：v0.3.0 — v0.2 回测后优化
版本：v0.2.0 — 多时间框架重构
回测日期：2026-06-07
回测结果：1253笔 / 胜率17.4% / -0.36% / 平均持仓24min
已知问题（诊断见 docs/backtest-pitfalls.md）：
1. 成交量 surge 计算了但未用于入场过滤 → 信号过多
2. 1H 只要求"非反向"而非"同向" → 过滤太弱
3. 止损太紧（保本0.5ATR/追踪1.0ATR） → 持仓仅24min
4. 缺少最低波动率过滤
注意：以下属性在首次回测时缺失，后补（stoploss/use_custom_stoploss/minimal_roi/NaN清理）
"""
from functools import reduce
@ -30,17 +40,7 @@ from freqtrade.strategy import IStrategy, merge_informative_pair
from freqtrade.strategy import IntParameter, DecimalParameter
# ── 工具函数：Swing Point 检测 ──────────────────────────────────
def detect_swing_points(df: DataFrame, window: int, col_high="high", col_low="low"):
"""
在给定 DataFrame 上检测 Swing High / Swing Low。
返回添加了以下列的 DataFrame：
- is_swing_high / is_swing_low : bool
- last_swing_high / last_swing_low : float (前向填充)
"""
w = int(window)
roll_max = df[col_high].rolling(window=w, center=True).max()
roll_min = df[col_low].rolling(window=w, center=True).min()
@ -63,24 +63,18 @@ def detect_swing_points(df: DataFrame, window: int, col_high="high", col_low="lo
def detect_candle_patterns(df: DataFrame, pin_body_ratio=0.3, engulf_ratio=1.5):
"""
K线形态检测。返回添加了形态布尔列的 DataFrame。
"""
body = abs(df["close"] - df["open"])
c_range = df["high"] - df["low"]
upper_wick = df["high"] - df[["open", "close"]].max(axis=1)
lower_wick = df[["open", "close"]].min(axis=1) - df["low"]
safe_range = c_range.replace(0, np.nan)
# 看涨 Pin Bar（锤子线）
df["bullish_pinbar"] = (
(body < pin_body_ratio * safe_range)
& (lower_wick > 2 * body)
& (lower_wick > upper_wick)
& (df["close"] > df["open"])
)
# 看跌 Pin Bar（射击之星）
df["bearish_pinbar"] = (
(body < pin_body_ratio * safe_range)
& (upper_wick > 2 * body)
@ -88,7 +82,6 @@ def detect_candle_patterns(df: DataFrame, pin_body_ratio=0.3, engulf_ratio=1.5):
& (df["close"] < df["open"])
)
# 看涨吞没
prev_open = df["open"].shift(1)
prev_close = df["close"].shift(1)
prev_body = abs(prev_close - prev_open)
@ -100,8 +93,6 @@ def detect_candle_patterns(df: DataFrame, pin_body_ratio=0.3, engulf_ratio=1.5):
& (df["close"] > prev_open)
& (body > engulf_ratio * prev_body)
)
# 看跌吞没
df["bearish_engulfing"] = (
(prev_close > prev_open)
& (df["close"] < df["open"])
@ -113,28 +104,9 @@ def detect_candle_patterns(df: DataFrame, pin_body_ratio=0.3, engulf_ratio=1.5):
return df
# ── 策略类 ──────────────────────────────────────────────────────
class PriceActionStrategyV03(IStrategy):
"""
多时间框架价格行为策略 — D1 定方向 → 1H 找结构 → 5M 抓时机。
v0.3 相比 v0.2 的核心改进：
- 成交量确认由"计算但未使用"→ 成为入场必要条件
- 1H 趋势要求从"非反向"→ 必须同向
- S/R 接近阈值从 3.0% → 1.5%
- 移动止损更宽：初始 2.0 ATR / 保本 1.5 ATR / 追踪 2.0 ATR
- 新增最低 ATR 波动率过滤
- 出场增加 1H 趋势反转条件
适用：ETH/USDT 永续合约，Binance，5M 主时间框架。
"""
class PriceActionStrategy(IStrategy):
INTERFACE_VERSION = 3
# ── 基础设置 ──────────────────────────────────────────────
timeframe = "5m"
can_short = True
max_open_trades = 1
@ -142,42 +114,25 @@ class PriceActionStrategyV03(IStrategy):
process_only_new_candles = True
use_exit_signal = True
# ── 运行时强制属性（回测配置补齐） ─────────────────────────
stoploss = -0.15
use_custom_stoploss = True
minimal_roi = {"0": 100}
stoploss = -0.10 # [回测补] 首次缺失
use_custom_stoploss = True # [回测补] 首次缺失
minimal_roi = {"0": 100} # [回测补] 首次缺失
# ── 可优化参数 ────────────────────────────────────────────
# -- 日线（宏观）--
ema_fast_daily = IntParameter(10, 30, default=20, space="buy")
ema_slow_daily = IntParameter(40, 80, default=50, space="buy")
swing_window_daily = IntParameter(3, 10, default=5, space="buy")
# -- 1H（中期结构）--
ema_fast_h1 = IntParameter(10, 30, default=20, space="buy")
ema_slow_h1 = IntParameter(40, 80, default=50, space="buy")
swing_window_h1 = IntParameter(3, 10, default=5, space="buy")
# -- ATR 止损 --
atr_period = IntParameter(10, 28, default=14, space="buy")
atr_stop_multiplier = DecimalParameter(1.5, 3.0, default=2.0, space="sell")
atr_stop_multiplier = DecimalParameter(1.0, 3.0, default=1.5, space="sell")
# -- K线形态 --
pin_bar_body_ratio = DecimalParameter(0.15, 0.40, default=0.30, space="buy")
engulfing_body_ratio = DecimalParameter(1.2, 3.0, default=1.5, space="buy")
# -- 成交量 --
volume_surge_multiplier = DecimalParameter(1.2, 3.0, default=1.5, space="buy")
# -- S/R 接近阈值 --
sr_proximity_pct = DecimalParameter(0.5, 3.0, default=1.5, space="buy")
# -- ATR 最低波动率 --
min_atr_ratio = DecimalParameter(0.3, 1.0, default=0.5, space="buy")
# ── 多时间框架声明 ────────────────────────────────────────
def informative_pairs(self):
pairs = self.dp.current_whitelist()
informative_pairs = []
@ -186,29 +141,13 @@ class PriceActionStrategyV03(IStrategy):
informative_pairs.append((pair, "1d"))
return informative_pairs
# ── 指标计算 ──────────────────────────────────────────────
def populate_indicators(self, dataframe: DataFrame, metadata: dict) -> DataFrame:
"""
三层时间框架的指标计算流水线：
Layer 1 — D1：宏观趋势方向
Layer 2 — 1H：S/R 区域 + 中期结构
Layer 3 — 5M：入场信号 + K线形态
"""
# ============================================================
# Layer 1: 日线 —— 宏观方向
# ============================================================
# Layer 1: D1
daily = self.dp.get_pair_dataframe(pair=metadata["pair"], timeframe="1d")
if not daily.empty:
daily["ema_fast"] = ta.EMA(daily, timeperiod=self.ema_fast_daily.value)
daily["ema_slow"] = ta.EMA(daily, timeperiod=self.ema_slow_daily.value)
daily = detect_swing_points(daily, self.swing_window_daily.value)
daily["trend_up"] = (
(daily["ema_fast"] > daily["ema_slow"])
& (daily["close"] > daily["ema_fast"])
@ -219,28 +158,18 @@ class PriceActionStrategyV03(IStrategy):
)
else:
daily = dataframe.copy()
for col in [
"ema_fast", "ema_slow", "is_swing_high", "is_swing_low",
"last_swing_high", "last_swing_low", "trend_up", "trend_down",
]:
for col in ["ema_fast", "ema_slow", "is_swing_high", "is_swing_low",
"last_swing_high", "last_swing_low", "trend_up", "trend_down"]:
daily[col] = np.nan
dataframe = merge_informative_pair(
dataframe, daily, self.timeframe, "1d", ffill=True,
)
# ============================================================
# Layer 2: 1H —— 中期结构 + S/R 区域
# ============================================================
dataframe = merge_informative_pair(dataframe, daily, self.timeframe, "1d", ffill=True)
# Layer 2: 1H
hourly = self.dp.get_pair_dataframe(pair=metadata["pair"], timeframe="1h")
if not hourly.empty:
hourly["ema_fast"] = ta.EMA(hourly, timeperiod=self.ema_fast_h1.value)
hourly["ema_slow"] = ta.EMA(hourly, timeperiod=self.ema_slow_h1.value)
hourly = detect_swing_points(hourly, self.swing_window_h1.value)
hourly["trend_up"] = (
(hourly["ema_fast"] > hourly["ema_slow"])
& (hourly["close"] > hourly["ema_fast"])
@ -251,53 +180,33 @@ class PriceActionStrategyV03(IStrategy):
)
else:
hourly = dataframe.copy()
for col in [
"ema_fast", "ema_slow", "is_swing_high", "is_swing_low",
"last_swing_high", "last_swing_low", "trend_up", "trend_down",
]:
for col in ["ema_fast", "ema_slow", "is_swing_high", "is_swing_low",
"last_swing_high", "last_swing_low", "trend_up", "trend_down"]:
hourly[col] = np.nan
dataframe = merge_informative_pair(
dataframe, hourly, self.timeframe, "1h", ffill=True,
)
dataframe = merge_informative_pair(dataframe, hourly, self.timeframe, "1h", ffill=True)
# ============================================================
# Layer 3: 5M —— 入场执行信号
# ============================================================
# ATR
# Layer 3: 5M
dataframe["atr"] = ta.ATR(dataframe, timeperiod=self.atr_period.value)
dataframe["atr_ratio"] = (
dataframe["atr"] / dataframe["atr"].rolling(20).mean()
)
# 5M EMA
dataframe["atr_ratio"] = dataframe["atr"] / dataframe["atr"].rolling(20).mean()
dataframe["ema_20_5m"] = ta.EMA(dataframe, timeperiod=20)
# K线形态
dataframe = detect_candle_patterns(
dataframe,
pin_body_ratio=self.pin_bar_body_ratio.value,
engulf_ratio=self.engulfing_body_ratio.value,
)
# 成交量确认
dataframe["volume_ma20"] = dataframe["volume"].rolling(20).mean()
dataframe["volume_surge"] = (
dataframe["volume"]
> self.volume_surge_multiplier.value * dataframe["volume_ma20"]
dataframe["volume"] > self.volume_surge_multiplier.value * dataframe["volume_ma20"]
)
# ============================================================
# S/R 距离
# ============================================================
support = dataframe["last_swing_low_1h"]
resistance = dataframe["last_swing_high_1h"]
dataframe["dist_to_support_pct"] = np.where(
support > 0,
(dataframe["close"] - support) / support * 100,
(dataframe["close"] - support) / dataframe["close"] * 100,
np.nan,
)
dataframe["dist_to_resistance_pct"] = np.where(
@ -306,24 +215,11 @@ class PriceActionStrategyV03(IStrategy):
np.nan,
)
# ============================================================
# v0.3 新增：连续确认（避免单根假突破）
# ============================================================
dataframe["bullish_pattern_prev"] = dataframe["bullish_pinbar"].shift(1) | dataframe["bullish_engulfing"].shift(1)
dataframe["bearish_pattern_prev"] = dataframe["bearish_pinbar"].shift(1) | dataframe["bearish_engulfing"].shift(1)
# ============================================================
# NaN 清理：多时间框架合并后布尔列前部有 NaN
# ============================================================
# NaN 清理 [回测补]
bool_cols = [
"trend_up_1d", "trend_down_1d",
"trend_up_1h", "trend_down_1h",
"trend_up_1d", "trend_down_1d", "trend_up_1h", "trend_down_1h",
"bullish_pinbar", "bearish_pinbar",
"bullish_engulfing", "bearish_engulfing",
"volume_surge",
"bullish_pattern_prev", "bearish_pattern_prev",
"bullish_engulfing", "bearish_engulfing", "volume_surge",
]
for col in bool_cols:
if col in dataframe.columns:
@ -331,178 +227,46 @@ class PriceActionStrategyV03(IStrategy):
return dataframe
# ── 入场信号 ──────────────────────────────────────────────
def populate_entry_trend(self, dataframe: DataFrame, metadata: dict) -> DataFrame:
"""
入场逻辑 —— 四层确认（v0.3 强化版）：
daily_bullish = dataframe["trend_up_1d"] & (dataframe["close"] > dataframe["ema_fast_1d"])
daily_bearish = dataframe["trend_down_1d"] & (dataframe["close"] < dataframe["ema_fast_1d"])
做多：
D1: 上升趋势
1H: 也必须上升趋势（v0.2 只要求"非下降"→ v0.3 要求同向）
5M: 价格在支撑附近(<1.5%) + 看涨形态 + 成交量放大 + 连续确认
风控: ATR 波动率充足（不在沉闷市场中交易）
"""
h1_not_bearish = ~dataframe["trend_down_1h"]
price_near_support = (dataframe["dist_to_support_pct"] < 3.0) & (dataframe["dist_to_support_pct"] > 0)
# ── 宏观环境 ──
daily_bullish = (
dataframe["trend_up_1d"]
& (dataframe["close"] > dataframe["ema_fast_1d"])
)
daily_bearish = (
dataframe["trend_down_1d"]
& (dataframe["close"] < dataframe["ema_fast_1d"])
)
# ── 1H 中期条件 ──
# v0.3 改动：从 "h1_not_bearish" 升级为 "h1_bullish"（必须同向）
h1_bullish = dataframe["trend_up_1h"] & (dataframe["close"] > dataframe["ema_fast_1h"])
h1_bearish = dataframe["trend_down_1h"] & (dataframe["close"] < dataframe["ema_fast_1h"])
sr_pct = self.sr_proximity_pct.value
price_near_support = (
(dataframe["dist_to_support_pct"] < sr_pct)
& (dataframe["dist_to_support_pct"] > 0)
)
price_near_resistance = (
(dataframe["dist_to_resistance_pct"] < sr_pct)
& (dataframe["dist_to_resistance_pct"] > 0)
)
# ── 5M 入场形态 ──
h1_not_bullish = ~dataframe["trend_up_1h"]
price_near_resistance = (dataframe["dist_to_resistance_pct"] < 3.0) & (dataframe["dist_to_resistance_pct"] > 0)
bullish_pattern = dataframe["bullish_pinbar"] | dataframe["bullish_engulfing"]
bearish_pattern = dataframe["bearish_pinbar"] | dataframe["bearish_engulfing"]
# ── v0.3 新增过滤 ──
# 成交量必选
volume_ok = dataframe["volume_surge"]
# 最低波动率：ATR 不能太小（市场太沉闷不做）
sufficient_volatility = dataframe["atr_ratio"] >= self.min_atr_ratio.value
# 避免极端波动
normal_vol = dataframe["atr_ratio"] < 2.0
# 连续确认：当前和前一根 K 线都有看涨/看跌信号，减少假突破
consecutive_bullish = bullish_pattern & dataframe["bullish_pattern_prev"]
consecutive_bearish = bearish_pattern & dataframe["bearish_pattern_prev"]
# ============================================================
# 做多条件（严格过滤）
# ============================================================
conditions_long = [
daily_bullish,
h1_bullish, # v0.3: 1H 必须同向上升
price_near_support,
bullish_pattern,
volume_ok, # v0.3: 成交量必选
sufficient_volatility, # v0.3: 最低波动率
normal_vol,
]
# ============================================================
# 做空条件（严格过滤）
# ============================================================
conditions_short = [
daily_bearish,
h1_bearish, # v0.3: 1H 必须同向下降
price_near_resistance,
bearish_pattern,
volume_ok,
sufficient_volatility,
normal_vol,
]
# ── 写入信号 ──
conditions_long = [daily_bullish, h1_not_bearish, price_near_support, bullish_pattern, normal_vol]
conditions_short = [daily_bearish, h1_not_bullish, price_near_resistance, bearish_pattern, normal_vol]
if conditions_long:
dataframe.loc[
reduce(lambda a, b: a & b, conditions_long),
"enter_long",
] = 1
dataframe.loc[reduce(lambda a, b: a & b, conditions_long), "enter_long"] = 1
if conditions_short:
dataframe.loc[
reduce(lambda a, b: a & b, conditions_short),
"enter_short",
] = 1
dataframe.loc[reduce(lambda a, b: a & b, conditions_short), "enter_short"] = 1
return dataframe
# ── 出场信号 ──────────────────────────────────────────────
def populate_exit_trend(self, dataframe: DataFrame, metadata: dict) -> DataFrame:
"""
信号出场（v0.3 增强）：
主要出场仍由 custom_stoploss 的移动止损处理。
这里追加结构破坏级别的强制离场。
"""
# ── 多头离场 ──
daily_no_longer_bullish = ~dataframe["trend_up_1d"]
h1_no_longer_bullish = ~dataframe["trend_up_1h"] # v0.3 新增
conditions_exit_long = [
daily_no_longer_bullish,
h1_no_longer_bullish,
]
# ── 空头离场 ──
daily_no_longer_bearish = ~dataframe["trend_down_1d"]
h1_no_longer_bearish = ~dataframe["trend_down_1h"] # v0.3 新增
conditions_exit_short = [
daily_no_longer_bearish,
h1_no_longer_bearish,
]
# ── 写入 ──
conditions_exit_long = [daily_no_longer_bullish]
conditions_exit_short = [daily_no_longer_bearish]
if conditions_exit_long:
dataframe.loc[
reduce(lambda a, b: a | b, conditions_exit_long),
"exit_long",
] = 1
dataframe.loc[reduce(lambda a, b: a | b, conditions_exit_long), "exit_long"] = 1
if conditions_exit_short:
dataframe.loc[
reduce(lambda a, b: a | b, conditions_exit_short),
"exit_short",
] = 1
dataframe.loc[reduce(lambda a, b: a | b, conditions_exit_short), "exit_short"] = 1
return dataframe
# ── 动态移动止损 ──────────────────────────────────────────
def custom_stoploss(
self,
pair: str,
trade,
current_time,
current_rate: float,
current_profit: float,
after_fill: bool,
**kwargs,
) -> Optional[float]:
"""
v0.3 宽止损设计 —— 给趋势呼吸空间：
阶段1（利润 < 1.5 ATR）：初始止损 ATR × 2.0
阶段2（利润 1.5~3.0 ATR）：保本
阶段3（利润 > 3.0 ATR）：追踪止损 ATR × 2.0
v0.2 参考：初始 1.5 / 保本 0.5 / 追踪 1.0
"""
def custom_stoploss(self, pair, trade, current_time, current_rate, current_profit,
after_fill, **kwargs) -> Optional[float]:
dataframe, _ = self.dp.get_analyzed_dataframe(trade.pair, self.timeframe)
if dataframe.empty:
return None
@ -514,68 +278,37 @@ class PriceActionStrategyV03(IStrategy):
if trade.is_short:
profit_ratio = -current_profit
if profit_ratio > atr_ratio * 3.0:
return -atr_ratio * 2.0
elif profit_ratio > atr_ratio * 1.5:
if profit_ratio > atr_ratio * 2.0:
return -atr_ratio * 1.0
elif profit_ratio > atr_ratio * 0.5:
return 0
else:
return -atr_ratio * self.atr_stop_multiplier.value
else:
if current_profit > atr_ratio * 3.0:
return -atr_ratio * 2.0
elif current_profit > atr_ratio * 1.5:
if current_profit > atr_ratio * 2.0:
return -atr_ratio * 1.0
elif current_profit > atr_ratio * 0.5:
return 0
else:
return -atr_ratio * self.atr_stop_multiplier.value
# ── 自定义出场（结构破坏） ────────────────────────────────
def custom_exit(
self,
pair: str,
trade,
current_time,
current_rate: float,
current_profit: float,
**kwargs,
) -> Optional[str]:
"""
结构层面出场：D1 或 1H 趋势反转 → 立刻离场。
"""
def custom_exit(self, pair, trade, current_time, current_rate, current_profit,
**kwargs) -> Optional[str]:
dataframe, _ = self.dp.get_analyzed_dataframe(trade.pair, self.timeframe)
if dataframe.empty:
return None
last_candle = dataframe.iloc[-1]
if trade.is_short:
if last_candle.get("trend_up_1d", False) or last_candle.get("trend_up_1h", False):
return "trend_reversed"
if last_candle.get("trend_up_1d", False):
return "daily_trend_reversed"
else:
if last_candle.get("trend_down_1d", False) or last_candle.get("trend_down_1h", False):
return "trend_reversed"
if last_candle.get("trend_down_1d", False):
return "daily_trend_reversed"
return None
# ── 仓位管理 ──────────────────────────────────────────────
def custom_stake_amount(
self,
pair: str,
current_time,
current_rate: float,
proposed_stake: float,
min_stake: Optional[float],
max_stake: float,
leverage: float,
entry_tag: Optional[str],
side: str,
**kwargs,
) -> float:
"""
固定风险仓位管理：每次交易风险 = 账户的 1%。
"""
def custom_stake_amount(self, pair, current_time, current_rate, proposed_stake,
min_stake, max_stake, leverage, entry_tag, side, **kwargs) -> float:
dataframe, _ = self.dp.get_analyzed_dataframe(pair, self.timeframe)
if dataframe.empty:
return min_stake or proposed_stake
@ -589,24 +322,10 @@ class PriceActionStrategyV03(IStrategy):
position_size = risk_amount / stop_distance if stop_distance > 0 else proposed_stake
position_size = min(position_size, max_stake or float("inf"))
if min_stake and position_size < min_stake:
return 0
return position_size
# ── 最终入场确认 ──────────────────────────────────────────
def confirm_trade_entry(
self,
pair: str,
order_type: str,
amount: float,
rate: float,
time_in_force: str,
current_time,
entry_tag: Optional[str],
side: str,
**kwargs,
) -> bool:
def confirm_trade_entry(self, pair, order_type, amount, rate, time_in_force,
current_time, entry_tag, side, **kwargs) -> bool:
return True