Hands-on teaches you how to convert a Python single-variety policy into a multi-variety policy.

One, handshake teaches you how to convert a Python single-variety policy into a multi-variety policy.

In the last article, we implemented a very simple Python strategy:"Python version of chase and kill strategy"This strategy can be used to manipulate an account to make programmatic trades on a certain trading pair, the principle is simple, it is a chase-and-drop. Sometimes we want to use the same trading logic to operate different trading pairs. Multiple robots can be created, set up different trading pairs, to trade in different currencies. If the strategy is not very complex, given the great flexibility of the inventor's quantitative trading platform.

The following is the source code of the modified strategy:

'''backtest
start: 2019-02-20 00:00:00
end: 2020-01-10 00:00:00
period: 1m
exchanges: [{"eid":"OKEX","currency":"BTC_USDT"},{"eid":"OKEX","currency":"ETH_USDT","stocks":30},{"eid":"OKEX","currency":"LTC_USDT","stocks":100}]
'''

import time
import json

params = {
    "arrBasePrice": [-1, -1, -1],     # -1
    "arrRatio": [0.05, 0.05, 0.05],         # 0.05
    "arrAcc": [],           # _C(exchange.GetAccount)
    "arrLastCancelAll": [0, 0, 0], # 0
    "arrMinStocks": [0.01, 0.01, 0.01],     # 0.01
    "arrPricePrecision": [2, 2, 2], # 2
    "arrAmountPrecision": [3, 2, 2], # 2
    "arrTick":[]
}

def CancelAll(e):
    while True : 
        orders = _C(e.GetOrders)
        for i in range(len(orders)) :
            e.CancelOrder(orders[i]["Id"], orders[i])
        if len(orders) == 0 :
            break
        Sleep(1000)

def process(e, index):
    global params
    ticker = _C(e.GetTicker)
    params["arrTick"][index] = ticker
    if params["arrBasePrice"][index] == -1 :
        params["arrBasePrice"][index] = ticker.Last
    if ticker.Last - params["arrBasePrice"][index] > 0 and (ticker.Last - params["arrBasePrice"][index]) / params["arrBasePrice"][index] > params["arrRatio"][index]:
        params["arrAcc"][index] = _C(e.GetAccount)
        if params["arrAcc"][index].Balance * params["arrRatio"][index] / ticker.Last > params["arrMinStocks"][index]:
            e.Buy(ticker.Last, params["arrAcc"][index].Balance * params["arrRatio"][index] / ticker.Last)
            params["arrBasePrice"][index] = ticker.Last
    if ticker.Last - params["arrBasePrice"][index] < 0 and (params["arrBasePrice"][index] - ticker.Last) / params["arrBasePrice"][index] > params["arrRatio"][index]: 
        params["arrAcc"][index] = _C(e.GetAccount)
        if params["arrAcc"][index].Stocks * params["arrRatio"][index] > params["arrMinStocks"][index]:
            e.Sell(ticker.Last, params["arrAcc"][index].Stocks * params["arrRatio"][index])
            params["arrBasePrice"][index] = ticker.Last
    ts = time.time()
    if ts - params["arrLastCancelAll"][index] > 60 * 5 :
        CancelAll(e)
        params["arrLastCancelAll"][index] = ts 

def main():
    global params
    
    for i in range(len(exchanges)) :    
        params["arrAcc"].append(_C(exchanges[i].GetAccount))
        params["arrTick"].append(_C(exchanges[i].GetTicker))
        exchanges[i].SetPrecision(params["arrPricePrecision"][i], params["arrAmountPrecision"][i])

    for key in params :
        if len(params[key]) < len(exchanges):
            raise "params error!"

    while True:
        tblAcc = {
            "type" : "table",
            "title": "account",
            "cols": ["账户信息"], 
            "rows": []
        }        

        tblTick = {
            "type" : "table",
            "title": "ticker",
            "cols": ["行情信息"], 
            "rows": []
        }
        for i in range(len(exchanges)): 
            process(exchanges[i], i)

        for i in range(len(exchanges)):
            tblAcc["rows"].append([json.dumps(params["arrAcc"][i])])
            tblTick["rows"].append([json.dumps(params["arrTick"][i])])

        LogStatus(_D(), "\n`" + json.dumps([tblAcc, tblTick]) + "`")
        Sleep(500)

Second, find the difference.

If you look at the code, is there a big difference between what you found and what you found in the previous article? In fact, the transaction logic is exactly the same, without any changes, only we modify the strategy into multiple varieties, so we can not use the previous hash single variable as a form of strategy parameter hash, a more reasonable solution is to make the parameter into an array, the index of each position of the array corresponds to the added transaction pair.

And then you wrap this part of the transactional logic code into a function.processIn the strategy main loop, the function is called iteratively based on the added transaction pairs, so that each transaction pair executes one transaction logic code.

• I'm not sure what you're talking about.

  for i in range(len(exchanges)): 
      process(exchanges[i], i)
  

• The policy parameters:

  params = {
      "arrBasePrice": [-1, -1, -1],           # -1
      "arrRatio": [0.05, 0.05, 0.05],         # 0.05
      "arrAcc": [],                           # _C(exchange.GetAccount)
      "arrLastCancelAll": [0, 0, 0],          # 0
      "arrMinStocks": [0.01, 0.01, 0.01],     # 0.01
      "arrPricePrecision": [2, 2, 2],         # 2
      "arrAmountPrecision": [3, 2, 2],        # 2
      "arrTick":[]
  }
  

  This design allows each pair to have its own parameters, as each transaction has a large difference in possible prices, which can also vary in parameters, sometimes requiring differentiation settings.

• CancelAll function

  This can be contrasted with a change in the function. The function simply modifies a bit of code and then thinks about the intention of the modification.

• State bar chart data

  The addition of a chart showing market data and account asset data in the status bar, which shows the corresponding assets and markets for each exchange object in real time.

With all of these design ideas, is it really that simple to change a Python strategy into a multi-variety strategy?

Three, repeat the test.

The strategy is for reference learning only, retesting tests, and interest in optimizing upgrades.Policy address