233 lines
16 KiB
Markdown
233 lines
16 KiB
Markdown
|
|
> Name
|
|
|
|
基于Nadaraya-Watson的多维度整合交易策略-Multi-Dimensional-Integration-Trading-Strategy-Based-on-Nadaraya-Watson
|
|
|
|
> Author
|
|
|
|
ianzeng123
|
|
|
|
> Strategy Description
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
[trans]
|
|
#### 概述
|
|
该策略是一个基于Nadaraya-Watson核回归的多维度交易系统,通过整合技术、情绪、超感和意向四个维度的市场信息,形成综合信号来指导交易决策。策略采用权重优化方法,对不同维度的信号进行加权处理,并结合趋势和动量过滤器以提高信号质量。系统还包含完整的风险管理模块,通过止损和止盈来保护资金安全。
|
|
|
|
#### 策略原理
|
|
策略的核心是通过Nadaraya-Watson核回归方法对多个维度的市场数据进行平滑处理。具体来说:
|
|
1. 技术维度使用收盘价
|
|
2. 情绪维度使用RSI指标
|
|
3. 超感维度使用ATR波动率
|
|
4. 意向维度使用价格与均线偏离度
|
|
这些维度经过核回归平滑后,通过预设权重(技术0.4、情绪0.2、超感0.2、意向0.2)进行加权整合,形成最终的交易信号。当整合信号与其移动平均线发生交叉时,结合趋势和动量过滤器确认后发出交易指令。
|
|
|
|
#### 策略优势
|
|
1. 多维度分析提供了更全面的市场视角,避免单一指标的局限性
|
|
2. Nadaraya-Watson核回归能有效降低市场噪音,提供更平滑的信号
|
|
3. 权重优化机制允许根据市场特征调整各维度的重要性
|
|
4. 趋势和动量过滤器的加入显著提高了信号质量
|
|
5. 完善的风险管理系统确保资金安全
|
|
|
|
#### 策略风险
|
|
1. 参数优化过度可能导致过拟合
|
|
2. 多重过滤条件可能错过部分有效信号
|
|
3. 核回归计算复杂度较高,可能影响实时性能
|
|
4. 权重分配不当可能削弱某些重要市场信号
|
|
缓解措施包括:使用样本外测试验证参数,动态调整过滤条件,优化计算效率,定期评估和调整权重分配。
|
|
|
|
#### 策略优化方向
|
|
1. 引入自适应权重系统,根据市场状态动态调整各维度权重
|
|
2. 开发更智能的过滤机制,平衡信号质量和数量
|
|
3. 优化Nadaraya-Watson算法实现,提高计算效率
|
|
4. 加入市场周期识别模块,在不同市场阶段采用不同的参数设置
|
|
5. 扩展风险管理系统,增加动态止损和仓位管理功能
|
|
|
|
#### 总结
|
|
这是一个将数学方法与交易智慧相结合的创新策略。通过多维度分析和先进的数学工具,策略能够捕捉市场的多个层面,提供相对可靠的交易信号。虽然存在一些优化空间,但策略的整体框架是稳健的,具有实际应用价值。||
|
|
|
|
#### Overview
|
|
This strategy is a multi-dimensional trading system based on Nadaraya-Watson kernel regression, which integrates market information from technical, emotional, extrasensory, and intentional dimensions to form comprehensive signals for trading decisions. The strategy employs weight optimization methods, applies weighted processing to signals from different dimensions, and combines trend and momentum filters to improve signal quality. The system also includes a complete risk management module to protect capital through stop-loss and take-profit mechanisms.
|
|
|
|
#### Strategy Principles
|
|
The core of the strategy lies in using Nadaraya-Watson kernel regression to smooth multi-dimensional market data. Specifically:
|
|
1. Technical dimension uses closing price
|
|
2. Emotional dimension uses RSI indicator
|
|
3. Extrasensory dimension uses ATR volatility
|
|
4. Intentional dimension uses price deviation from moving average
|
|
These dimensions are smoothed through kernel regression and then integrated using preset weights (Technical 0.4, Emotional 0.2, Extrasensory 0.2, Intentional 0.2) to form the final trading signal. Trading orders are issued when the integrated signal crosses its moving average, confirmed by trend and momentum filters.
|
|
|
|
#### Strategy Advantages
|
|
1. Multi-dimensional analysis provides a more comprehensive market perspective, avoiding limitations of single indicators
|
|
2. Nadaraya-Watson kernel regression effectively reduces market noise, providing smoother signals
|
|
3. Weight optimization mechanism allows adjustment of dimensional importance based on market characteristics
|
|
4. Addition of trend and momentum filters significantly improves signal quality
|
|
5. Comprehensive risk management system ensures capital safety
|
|
|
|
#### Strategy Risks
|
|
1. Parameter optimization may lead to overfitting
|
|
2. Multiple filtering conditions might miss some valid signals
|
|
3. Kernel regression has high computational complexity, potentially affecting real-time performance
|
|
4. Improper weight distribution may weaken important market signals
|
|
Mitigation measures include: using out-of-sample testing for parameter validation, dynamically adjusting filter conditions, optimizing computational efficiency, and periodically evaluating and adjusting weight distribution.
|
|
|
|
#### Strategy Optimization Directions
|
|
1. Introduce adaptive weight system to dynamically adjust dimensional weights based on market conditions
|
|
2. Develop smarter filtering mechanisms to balance signal quality and quantity
|
|
3. Optimize Nadaraya-Watson algorithm implementation to improve computational efficiency
|
|
4. Add market cycle recognition module to use different parameter settings in different market phases
|
|
5. Expand risk management system to include dynamic stop-loss and position management functions
|
|
|
|
#### Summary
|
|
This is an innovative strategy combining mathematical methods with trading wisdom. Through multi-dimensional analysis and advanced mathematical tools, the strategy can capture multiple aspects of the market, providing relatively reliable trading signals. While there is room for optimization, the overall framework of the strategy is robust and has practical application value.
|
|
|
|
[/trans]
|
|
|
|
|
|
|
|
> Source (PineScript)
|
|
|
|
``` pinescript
|
|
/*backtest
|
|
start: 2025-02-17 00:00:00
|
|
end: 2025-02-19 00:00:00
|
|
period: 1m
|
|
basePeriod: 1m
|
|
exchanges: [{"eid":"Binance","currency":"ETH_USDT"}]
|
|
*/
|
|
|
|
//@version=5
|
|
strategy("Enhanced Multidimensional Integration Strategy with Nadaraya", overlay=true, initial_capital=10000, currency=currency.USD, default_qty_type=strategy.percent_of_equity, default_qty_value=10)
|
|
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
// 1. Configuration and Weight Optimization Parameters
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
// Weights can be optimized to favor dimensions with higher historical correlation.
|
|
// Base values are maintained but can be fine-tuned.
|
|
w_technical = input.float(0.4, "Technical Weight", step=0.05)
|
|
w_emotional = input.float(0.2, "Emotional Weight", step=0.05)
|
|
w_extrasensor = input.float(0.2, "Extrasensory Weight", step=0.05)
|
|
w_intentional = input.float(0.2, "Intentional Weight", step=0.05)
|
|
|
|
// Parameters for Nadaraya-Watson Smoothing Function:
|
|
// Smoothing period and bandwidth affect the "memory" and sensitivity of the signal.
|
|
smooth_length = input.int(20, "Smoothing Period", minval=5)
|
|
bw_param = input.float(20, "Bandwidth", minval=1, step=1)
|
|
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
// 2. Risk Management Parameters
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
// Incorporate stop-loss and take-profit in percentage to protect capital.
|
|
// These parameters can be optimized through historical testing.
|
|
stopLossPerc = input.float(1.5, "Stop Loss (%)", step=0.1) / 100 // 1.5% stop-loss
|
|
takeProfitPerc = input.float(3.0, "Take Profit (%)", step=0.1) / 100 // 3.0% take-profit
|
|
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
// 3. Additional Filters (Trend and Momentum)
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
// A long-term moving average is used to confirm the overall trend direction.
|
|
trend_length = input.int(200, "Trend MA Period", minval=50)
|
|
// RSI is used to confirm momentum. A level of 50 is common to distinguish bullish and bearish phases.
|
|
rsi_filter_level = input.int(50, "RSI Confirmation Level", minval=30, maxval=70)
|
|
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
// 4. Definition of Dimensions
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
tech_series = close
|
|
emotional_series = ta.rsi(close, 14) / 100
|
|
extrasensorial_series = ta.atr(14) / close
|
|
intentional_series = (close - ta.sma(close, 50)) / close
|
|
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
// 5. Nadaraya-Watson Smoothing Function
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
// This function smooths each dimension using a Gaussian kernel.
|
|
// Proper smoothing reduces noise and helps obtain a more robust signal.
|
|
nadaraya_smooth(_src, _len, _bw) =>
|
|
if bar_index < _len
|
|
na
|
|
else
|
|
float sumW = 0.0
|
|
float sumWY = 0.0
|
|
for i = 0 to _len - 1
|
|
weight = math.exp(-0.5 * math.pow(((_len - 1 - i) / _bw), 2))
|
|
sumW := sumW + weight
|
|
sumWY := sumWY + weight * _src[i]
|
|
sumWY / sumW
|
|
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
// 6. Apply Smoothing to Each Dimension
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
sm_tech = nadaraya_smooth(tech_series, smooth_length, bw_param)
|
|
sm_emotional = nadaraya_smooth(emotional_series, smooth_length, bw_param)
|
|
sm_extrasens = nadaraya_smooth(extrasensorial_series, smooth_length, bw_param)
|
|
sm_intentional = nadaraya_smooth(intentional_series, smooth_length, bw_param)
|
|
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
// 7. Integration of Dimensions
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
// The integrated signal is composed of the weighted sum of each smoothed dimension.
|
|
// This multidimensional approach seeks to capture different aspects of market behavior.
|
|
integrated_signal = (w_technical * sm_tech) + (w_emotional * sm_emotional) + (w_extrasensor * sm_extrasens) + (w_intentional * sm_intentional)
|
|
// Additional smoothing of the integrated signal to obtain a reference line.
|
|
sma_integrated = ta.sma(integrated_signal, 10)
|
|
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
// 8. Additional Filters to Improve Accuracy and Win Rate
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
// Trend filter: only trade in the direction of the overall trend, determined by a 200-period SMA.
|
|
trendMA = ta.sma(close, trend_length)
|
|
// Momentum filter: RSI is used to confirm the strength of the movement (RSI > 50 for long and RSI < 50 for short).
|
|
rsi_val = ta.rsi(close, 14)
|
|
|
|
longFilter = (close > trendMA) and (rsi_val > rsi_filter_level)
|
|
shortFilter = (close < trendMA) and (rsi_val < rsi_filter_level)
|
|
|
|
// Crossover signals of the integrated signal with its SMA reference.
|
|
rawLongSignal = ta.crossover(integrated_signal, sma_integrated)
|
|
rawShortSignal = ta.crossunder(integrated_signal, sma_integrated)
|
|
// Incorporate trend and momentum filters to filter false signals.
|
|
longSignal = rawLongSignal and longFilter
|
|
shortSignal = rawShortSignal and shortFilter
|
|
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
// 9. Risk Management and Order Generation
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
// Entries are made based on the filtered integrated signal.
|
|
if longSignal
|
|
strategy.entry("Long", strategy.long, comment="Long Entry")
|
|
if shortSignal
|
|
strategy.entry("Short", strategy.short, comment="Short Entry")
|
|
|
|
// Add automatic exits using stop-loss and take-profit to limit losses and secure profits.
|
|
// For long positions: stop-loss below entry price and take-profit above.
|
|
if strategy.position_size > 0
|
|
strategy.exit("Exit Long", "Long", stop = strategy.position_avg_price * (1 - stopLossPerc), limit = strategy.position_avg_price * (1 + takeProfitPerc))
|
|
// For short positions: stop-loss above entry price and take-profit below.
|
|
if strategy.position_size < 0
|
|
strategy.exit("Exit Short", "Short", stop = strategy.position_avg_price * (1 + stopLossPerc), limit = strategy.position_avg_price * (1 - takeProfitPerc))
|
|
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
// 10. Visualization on the Chart
|
|
//────────────────────────────────────────────────────────────────────────────
|
|
plot(integrated_signal, color=color.blue, title="Integrated Signal", linewidth=2)
|
|
plot(sma_integrated, color=color.orange, title="SMA Integrated Signal", linewidth=2)
|
|
plot(trendMA, color=color.purple, title="Trend MA (200)", linewidth=1, style=plot.style_line)
|
|
plotshape(longSignal, title="Long Signal", location=location.belowbar, color=color.green, style=shape.labelup, text="LONG")
|
|
plotshape(shortSignal, title="Short Signal", location=location.abovebar, color=color.red, style=shape.labeldown, text="SHORT")
|
|
|
|
```
|
|
|
|
> Detail
|
|
|
|
https://www.fmz.com/strategy/482917
|
|
|
|
> Last Modified
|
|
|
|
2025-02-27 17:21:26
|