Steel Profit Spread Mean Reversion

Code TqSdk

Summary

This code builds a synthetic steel mill profit spread from daily futures prices for rebar, iron ore, and coke. It calculates the spread as rebar minus weighted quantities of the two inputs, smooths it with a 15-day moving average, and estimates a standard deviation. The strategy watches for the spread to cross back inside bands set half a standard deviation above or below the moving average. A downward cross of the upper band triggers a short rebar position and long positions in both inputs; an upward cross of the lower band reverses those directions.

The example uses target-position tasks to submit fixed contract volumes and tracks a local position state to avoid repeating the same signal. It updates calculations when a new daily bar appears and also monitors price changes. The code provides no backtest, transaction cost analysis, or evidence that the weighted spread is stable or cointegrated. Its standard deviation is computed across the available series rather than a stated rolling window, and position sizing, contract specifications, execution risk, and risk limits are not addressed.

Key ideas

The strategy represents steel mill economics with a weighted spread between rebar and its raw material inputs.
A 15-day moving average and half-standard-deviation bands define the spread's reference range.
Crossing back below the upper band opens short rebar and long input positions, while crossing back above the lower band does the reverse.
Fixed target volumes and a local state variable are used to manage positions and reduce duplicate signals.
The example does not establish profitability or address costs, spread stability, contract sizing, or risk controls.

Tags

Futures Commodities Mean reversion Pairs trading Risk management

Full text

# rb-i-jt-spread.py


```py
from tqsdk import TqApi, TargetPosTask
from tqsdk.tafunc import ma
import numpy as np

api = TqApi()

SYMBOL_rb = "SHFE.rb2001"
SYMBOL_i = "DCE.i2001"
SYMBOL_j = "DCE.j2001"

klines_rb = api.get_kline_serial(SYMBOL_rb, 86400)
klines_i = api.get_kline_serial(SYMBOL_i, 86400)
klines_j = api.get_kline_serial(SYMBOL_j, 86400)

target_pos_rb = TargetPosTask(api, SYMBOL_rb)
target_pos_i = TargetPosTask(api, SYMBOL_i)
target_pos_j = TargetPosTask(api, SYMBOL_j)


# 计算钢厂利润线，并将利润线画到副图
def cal_spread(klines_rb, klines_i, klines_j):
    index_spread = klines_rb.close - 1.6 * klines_i.close - 0.5 * klines_j.close
    # 使用15日均值，与注释保持一致
    ma_spread = ma(index_spread, 15)
    # 计算标准差
    spread_std = np.std(index_spread)
    klines_rb["index_spread"] = index_spread
    klines_rb["index_spread.board"] = "index_spread"

    return index_spread, ma_spread, spread_std


# 初始计算利润线
index_spread, ma_spread, spread_std = cal_spread(klines_rb, klines_i, klines_j)

print("ma_spread是%.2f,index_spread是%.2f,spread_std是%.2f" % (ma_spread.iloc[-1], index_spread.iloc[-1], spread_std))

# 记录当前持仓状态，避免重复发出信号
current_position = 0  # 0表示空仓，1表示多螺纹空焦炭焦煤，-1表示空螺纹多焦炭焦煤

while True:
    api.wait_update()

    # 每次有新日线生成时重新计算利润线
    if api.is_changing(klines_j.iloc[-1], "datetime"):
        index_spread, ma_spread, spread_std = cal_spread(klines_rb, klines_i, klines_j)

        # 计算上下轨
        upper_band = ma_spread.iloc[-1] + 0.5 * spread_std
        lower_band = ma_spread.iloc[-1] - 0.5 * spread_std

        print("ma_spread是%.2f,index_spread是%.2f,spread_std是%.2f" % (
            ma_spread.iloc[-1], index_spread.iloc[-1], spread_std))
        print("上轨是%.2f,下轨是%.2f" % (upper_band, lower_band))

        # 确保有足够的历史数据
        if len(index_spread) >= 2:
            # 1. 检测下穿上轨：前一个K线在上轨之上，当前K线在上轨之下或等于上轨
            if index_spread.iloc[-2] > upper_band and index_spread.iloc[-1] <= upper_band:
                if current_position != -1:  # 避免重复开仓
                    # 价差序列下穿上轨，利润冲高回落进行回复，策略空螺纹钢、多焦煤焦炭
                    target_pos_rb.set_target_volume(-100)
                    target_pos_i.set_target_volume(100)
                    target_pos_j.set_target_volume(100)
                    current_position = -1
                    print("下穿上轨：空螺纹钢、多焦煤焦炭")

            # 2. 检测上穿下轨：前一个K线在下轨之下，当前K线在下轨之上或等于下轨
            elif index_spread.iloc[-2] < lower_band and index_spread.iloc[-1] >= lower_band:
                if current_position != 1:  # 避免重复开仓
                    # 价差序列上穿下轨，利润过低回复上升，策略多螺纹钢、空焦煤焦炭
                    target_pos_rb.set_target_volume(100)
                    target_pos_i.set_target_volume(-100)
                    target_pos_j.set_target_volume(-100)
                    current_position = 1
                    print("上穿下轨：多螺纹钢、空焦煤焦炭")

    # 实时监控价差变化情况
    if api.is_changing(klines_rb.iloc[-1], "close") or api.is_changing(klines_i.iloc[-1], "close") or api.is_changing(
            klines_j.iloc[-1], "close"):
        # 实时更新价差
        current_spread = klines_rb.close.iloc[-1] - 1.6 * klines_i.close.iloc[-1] - 0.5 * klines_j.close.iloc[-1]
        # 可以添加实时监控输出
        # print("当前价差: %.2f" % current_spread)

```

Shown in full with attribution under the source's licence. Licence: Apache-2.0

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.

Mean-Reversion Trading on a Steel Mill Profit Spread

Summary

Key ideas

Tags