My語言怎么實現DMI量化交易策略

# My語言怎么實現DMI量化交易策略

## 一、DMI指標原理概述

DMI（Directional Movement Index）方向移動指標是由J.Welles Wilder開發的技術分析工具，主要用于判斷市場趨勢強度和方向。該指標包含四個核心組件：

1. **+DI（正向指標）**：反映上升趨勢強度
2. **-DI（負向指標）**：反映下降趨勢強度
3. **ADX（平均趨向指數）**：衡量整體趨勢強度
4. **ADXR（平均趨向指數評級）**：ADX的平滑版本

經典交易規則：
- 當+DI上穿-DI且ADX>25時，產生買入信號
- 當-DI上穿+DI且ADX>20時，產生賣出信號

## 二、My語言環境準備

### 2.1 開發環境配置
```my
// 檢查并導入必要庫
if not library_exists("quant") then
    import quant as q
end

// 設置回測參數
start_time = "2020-01-01"
end_time = "2023-12-31"
capital = 100000
commission = 0.0005 // 萬五手續費

2.2 數據獲取模塊

// 獲取滬深300指數數據
function get_data()
    data = q.get_history(
        symbol="000300.SH",
        start=start_time,
        end=end_time,
        frequency="1d",
        fields=["open","high","low","close","volume"]
    )
    return data
end

三、DMI核心算法實現

3.1 計算真實波幅（TR）

function calc_tr(high, low, prev_close)
    tr1 = high - low
    tr2 = abs(high - prev_close)
    tr3 = abs(low - prev_close)
    return max(tr1, tr2, tr3)
end

3.2 計算方向運動（DM）

function calc_dm(high, prev_high, low, prev_low)
    up_move = high - prev_high
    down_move = prev_low - low
    
    +dm = up_move > down_move and up_move > 0 ? up_move : 0
    -dm = down_move > up_move and down_move > 0 ? down_move : 0
    
    return +dm, -dm
end

3.3 完整DMI計算函數

function dmi(data, period=14)
    // 初始化數組
    len = length(data.close)
    +di = array(len)
    -di = array(len)
    adx = array(len)
    
    // 計算初始值
    for i = 1 to period
        +di[i] = 0
        -di[i] = 0
        adx[i] = 0
    end
    
    // 主計算循環
    for i = period+1 to len
        // 計算平滑TR和DM
        tr_sum = 0
        +dm_sum = 0
        -dm_sum = 0
        
        for j = i-period+1 to i
            tr = calc_tr(data.high[j], data.low[j], data.close[j-1])
            +dm, -dm = calc_dm(data.high[j], data.high[j-1], 
                              data.low[j], data.low[j-1])
            tr_sum += tr
            +dm_sum += +dm
            -dm_sum += -dm
        end
        
        // 計算DI值
        +di[i] = 100 * (+dm_sum / tr_sum)
        -di[i] = 100 * (-dm_sum / tr_sum)
        
        // 計算DX和ADX
        dx = 100 * abs(+di[i] - -di[i]) / (+di[i] + -di[i])
        adx[i] = i == period+1 ? dx : (adx[i-1]*(period-1) + dx)/period
    end
    
    return +di, -di, adx
end

四、交易策略實現

4.1 信號生成邏輯

function generate_signals(data)
    +di, -di, adx = dmi(data)
    
    signals = array(length(data.close))
    position = 0 // 0:空倉 1:多頭 -1:空頭
    
    for i = 2 to length(data.close)
        // 多頭信號：+DI上穿-DI且ADX>25
        if +di[i] > -di[i] and +di[i-1] <= -di[i-1] and adx[i] > 25
            signals[i] = 1
            position = 1
        
        // 空頭信號：-DI上穿+DI且ADX>20
        elseif -di[i] > +di[i] and -di[i-1] <= +di[i-1] and adx[i] > 20
            signals[i] = -1
            position = -1
        
        // 持倉延續
        else
            signals[i] = position
        end
    end
    
    return signals
end

4.2 策略回測模塊

function backtest(data, signals)
    returns = array(length(data.close))
    equity = array(length(data.close))
    equity[1] = capital
    
    for i = 2 to length(data.close)
        // 計算收益率
        if signals[i-1] == 1  // 多頭
            ret = (data.close[i]/data.close[i-1] - 1) - commission
        elseif signals[i-1] == -1  // 空頭
            ret = (1 - data.close[i]/data.close[i-1]) - commission
        else
            ret = 0
        end
        
        returns[i] = ret
        equity[i] = equity[i-1] * (1 + ret)
    end
    
    return equity, returns
end

五、策略優化與改進

5.1 動態參數優化

function optimize_parameters(data)
    best_sharpe = -inf
    best_period = 0
    best_adx_th = 0
    
    // 參數搜索空間
    for period = 10 to 20 step 2
        for adx_th = 15 to 30 step 5
            +di, -di, adx = dmi(data, period)
            signals = generate_signals_modified(data, +di, -di, adx, adx_th)
            equity, returns = backtest(data, signals)
            sharpe = calculate_sharpe(returns)
            
            if sharpe > best_sharpe
                best_sharpe = sharpe
                best_period = period
                best_adx_th = adx_th
            end
        end
    end
    
    return best_period, best_adx_th
end

5.2 加入止損機制

function add_stoploss(signals, data, stoploss_pct=0.05)
    modified_signals = copy(signals)
    entry_price = 0
    
    for i = 2 to length(data.close)
        // 記錄入場價格
        if signals[i] !=0 and signals[i-1] ==0
            entry_price = data.close[i]
        end
        
        // 檢查止損
        if signals[i] ==1 and (data.close[i] < entry_price*(1-stoploss_pct))
            modified_signals[i] = 0
        elseif signals[i] ==-1 and (data.close[i] > entry_price*(1+stoploss_pct))
            modified_signals[i] = 0
        end
    end
    
    return modified_signals
end

六、策略評價指標

function evaluate(equity, returns)
    // 年化收益率
    annual_return = (equity[-1]/equity[1])^(252/length(returns)) - 1
    
    // 最大回撤
    max_drawdown = 0
    peak = equity[1]
    for i = 2 to length(equity)
        if equity[i] > peak
            peak = equity[i]
        end
        dd = (peak - equity[i])/peak
        if dd > max_drawdown
            max_drawdown = dd
        end
    end
    
    // 夏普比率
    sharpe = mean(returns)/std(returns) * sqrt(252)
    
    return {
        "annual_return": annual_return,
        "max_drawdown": max_drawdown,
        "sharpe_ratio": sharpe,
        "win_rate": calculate_win_rate(returns)
    }
end

七、完整策略執行流程

// 主程序
function main()
    // 1. 數據準備
    data = get_data()
    
    // 2. 參數優化（可選）
    best_period, best_adx_th = optimize_parameters(data)
    
    // 3. 生成信號
    +di, -di, adx = dmi(data, best_period)
    signals = generate_signals_modified(data, +di, -di, adx, best_adx_th)
    signals = add_stoploss(signals, data)
    
    // 4. 回測評估
    equity, returns = backtest(data, signals)
    metrics = evaluate(equity, returns)
    
    // 5. 結果可視化
    plot_results(data, equity, signals)
    
    return metrics
end

// 執行策略
performance = main()
print(performance)

八、注意事項

1. 參數敏感性問題：DMI指標對周期參數(默認14)較為敏感，需進行參數優化
2. 震蕩市表現：ADX<25時市場可能處于震蕩狀態，建議結合其他指標過濾信號
3. 延遲特性：作為趨勢指標，DMI具有天然的滯后性
4. 品種適配：不同交易品種需要調整ADX閾值參數
5. 風險管理：建議單次交易風險控制在總資金的1-2%

”`