動(dòng)量策略 python

Most traders use Bollinger Bands. However, price is not normally distributed. That’s why only 42% of prices will close within one standard deviation. Please go ahead and read this article. However, I have some good news.

大多數(shù)交易者使用布林帶。 但是，價(jià)格不是正態(tài)分布的。 這就是為什么只有42％的價(jià)格會(huì)在一個(gè)標(biāo)準(zhǔn)偏差之內(nèi)收盤的原因。 請(qǐng)繼續(xù)閱讀本文 。 但是，我有一些好消息。

Price is not normally distributed. Returns are!

價(jià)格不是正態(tài)分布。 退貨都是！

Yes price is not normally distributed. Because price is nothing but sum of returns. And returns are normally distributed. So let’s jump in coding and hopefully you will have an “Aha!” moment.

是的，價(jià)格不是正態(tài)分布。 因?yàn)閮r(jià)格不過(guò)是回報(bào)之和。 收益是正態(tài)分布的。 因此，讓我們開始編碼，希望您會(huì)得到一個(gè)“ 啊哈 ！” 時(shí)刻。

I’m going to use EURUSD daily chart in my sample. However, it’s going to work with all assets and all timeframes.

我將在示例中使用EURUSD每日?qǐng)D表。 但是，它將適用于所有資產(chǎn)和所有時(shí)間范圍。

Add required libraries

添加所需的庫(kù)

# we only need these 3 libraries
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

Not let’s write the code to load our time-series dataset

不讓我們編寫代碼來(lái)加載時(shí)間序列數(shù)據(jù)集

def load_file(f):
fixprice = lambda x: float(x.replace(',', '.'))
df = pd.read_csv(f)
if "Gmt time" in df.columns:
df['Date'] = pd.to_datetime(df['Gmt time'], format="%d.%m.%Y %H:%M:%S.%f")
elif "time" in df.columns:
df['Date'] = pd.to_datetime(df['time'], unit="s")
df['Date'] = df['Date'] + np.timedelta64(3 * 60, "m")
df[['Date', 'Open', 'High', 'Low', 'Close']] = df[['Date', 'open', 'high', 'low', 'close']]
df = df[['Date', 'Open', 'High', 'Low', 'Close']]
elif "Tarih" in df.columns:
df['Date'] = pd.to_datetime(df['Tarih'], format="%d.%m.%Y")
df['Open'] = df['A??l??'].apply(fixprice)
df['High'] = df['Yüksek'].apply(fixprice)
df['Low'] = df['Dü?ük'].apply(fixprice)
df['Close'] = df['?imdi'].apply(fixprice)
else:
df["Date"] = pd.to_datetime(df["Date"])
# we need to shift or we will have lookahead bias in code
df["Returns"] = (df["Close"].shift(1) - df["Close"].shift(2)) / df["Close"].shift(2)
return df

Now if you look at the code, I have added a column “Returns” and it’s shifted. We don’t want our indicator to repaint and i don’t want to have lookahead bias. I am basically calculating the change from yesterday’s close to today’s close and shifting.

現(xiàn)在，如果您看一下代碼，我添加了“ Returns”列，它已轉(zhuǎn)移。 我們不希望我們的指標(biāo)重新粉刷，我也不想有前瞻性偏見。 我基本上是在計(jì)算從昨天的收盤價(jià)到今天的收盤價(jià)的變化。

Load your dataset and plot histogram of Returns

加載數(shù)據(jù)集并繪制退貨的直方圖

sym = "EURUSD"
period = "1d"
fl = "./{YOUR_PATH}/{} {}.csv".format(period, sym)
df = load_file(fl)
df["Returns"].hist(bins=1000, grid=False)EURUSD daily returns histogramEURUSD每日收益柱狀圖

Perfect bell shape curve. Now we know that we can get some real probabilities right? Empirical Rule (a.k.a. 68–95–99.7 rule) states that 68% of data will fall within one standard deviation, 95% of data will fall within two standard deviation and 99.7% of data will fall within three standard deviation. Ok let’s write the code to calculate it for us.

完美的鐘形曲線。 現(xiàn)在我們知道可以得到一些真實(shí)的概率了嗎？ 經(jīng)驗(yàn)法則 (也稱為68–95–99.7規(guī)則)指出，68％的數(shù)據(jù)將落在一個(gè)標(biāo)準(zhǔn)偏差內(nèi)，95％的數(shù)據(jù)將落在兩個(gè)標(biāo)準(zhǔn)偏差內(nèi)，而99.7％的數(shù)據(jù)將落在三個(gè)標(biāo)準(zhǔn)偏差內(nèi)。 好的，讓我們編寫代碼為我們計(jì)算一下。

def add_momentum(df, lb=20, std=2):
df["MA"] = df["Returns"].rolling(lb).mean()
df["STD"] = df["Returns"].rolling(lb).std()
df["OVB"] = df["Close"].shift(1) * (1 + (df["MA"] + df["STD"] * std))
df["OVS"] = df["Close"].shift(1) * (1 + (df["MA"] - df["STD"] * std))
return df

Now as we already have previous Bar’s close, it’s easy and safe to use and calculate the standard deviation. Also it’s easy for us to have overbought and oversold levels in advance. And they will stay there from the beginning of the current period. I also want to be sure if my data going to follow the empirical rule. So i want to get the statistics. Let’s code that block as well.

現(xiàn)在，由于我們已經(jīng)關(guān)閉了先前的Bar，因此可以輕松安全地使用和計(jì)算標(biāo)準(zhǔn)偏差。 同樣，我們很容易提前超買和超賣。 他們將從當(dāng)前階段開始一直呆在那里。 我還想確定我的數(shù)據(jù)是否遵循經(jīng)驗(yàn)法則。 所以我想獲得統(tǒng)計(jì)數(shù)據(jù)。 讓我們也對(duì)該塊進(jìn)行編碼。

def stats(df):
total = len(df)
ins1 = df[(df["Close"] > df["OVS"]) & (df["Close"] < df["OVB"])]
ins2 = df[(df["Close"] > df["OVS"])]
ins3 = df[(df["Close"] < df["OVB"])]
il1 = len(ins1)
il2 = len(ins2)
il3 = len(ins3)
r1 = np.round(il1 / total * 100, 2)
r2 = np.round(il2 / total * 100, 2)
r3 = np.round(il3 / total * 100, 2)
return r1, r2, r3

Now let’s call these function…

現(xiàn)在讓我們稱這些功能為…

df = add_momentum(df, lb=20, std=1)
stats(df)

Output is (67.36, 83.3, 83.77). So close price falls 67.36% within one standard deviation. Closing price is above OVS with 83.3% and below OVB with 83.77%. Amazing results… Now time to plot our bands to see how they look in the chart.

輸出為(67.36、83.3、83.77)。 因此，收盤價(jià)在一個(gè)標(biāo)準(zhǔn)偏差之內(nèi)下跌67.36％。 收盤價(jià)高于OVS，為83.3％，低于OVB，為83.77％。 驚人的結(jié)果...現(xiàn)在該繪制我們的樂(lè)隊(duì)，看看它們?cè)趫D表中的樣子。

I love candles. So let’s code it in a quick and dirty way and plot how our levels look.

我愛蠟燭。 因此，讓我們以一種快速而骯臟的方式對(duì)其進(jìn)行編碼，并繪制出關(guān)卡的外觀。

def plot_candles(df, l=0):
"""
Plots candles
l: plot last n candles. If set zero, draw all
"""
db = df.copy()
if l > 0:
db = db[-l:]
db = db.reset_index(drop=True).reset_index()
db["Up"] = db["Close"] > db["Open"]
db["Bottom"] = np.where(db["Up"], db["Open"], db["Close"])
db["Bar"] = db["High"] - db["Low"]
db["Body"] = abs(db["Close"] - db["Open"])
db["Color"] = np.where(db["Up"], "g", "r")
fig, ax = plt.subplots(1, 1, figsize=(16, 9))
ax.yaxis.tick_right()
ax.bar(db["index"], bottom=db["Low"], height=db["Bar"], width=0.25, color="#000000")
ax.bar(db["index"], bottom=db["Bottom"], height=db["Body"], width=0.5, color=db["Color"])
ax.plot(db["OVB"], color="r", linewidth=0.25)
ax.plot(db["OVS"], color="r", linewidth=0.25)
plt.show()

I want to see last 100 candles. Now let’s call this function

我想看最后100支蠟燭。 現(xiàn)在我們叫這個(gè)功能

plot_candles(df, l=100)EURUSD daily momentum channel bands歐元兌美元每日動(dòng)能通道帶

接下來(lái)做什么？ (What to do next?)

Well to be honest, if i would be making money using this strategy, i wouldn’t share it here with you (no offense). I wouldn’t even sell it. However, you can use your own imagination and add some strategies on this. You can thank me later if you decide to use this code and make money. I will send you my IBAN later :)

老實(shí)說(shuō)，如果我要使用這種策略來(lái)賺錢，我不會(huì)在這里與您分享(無(wú)罪)。 我什至不賣。 但是，您可以發(fā)揮自己的想象力，并為此添加一些策略。 如果您決定使用此代碼并賺錢，稍后可以感謝我。 稍后我將把您的IBAN發(fā)送給您：)

Disclaimer

免責(zé)聲明

I’m not a professional financial advisor. This article and codes, shared for educational purposes only and not financial advice. You are responsible your own losses or wins.

我不是專業(yè)的財(cái)務(wù)顧問(wèn)。 本文和代碼僅用于教育目的，不用于財(cái)務(wù)建議。 您應(yīng)對(duì)自己的損失或勝利負(fù)責(zé)。

The whole code of this article can be found on this repository:

可以在此存儲(chǔ)庫(kù)中找到本文的完整代碼：

Ah also; remember to follow me on the following social channels:

也啊 記得在以下社交渠道關(guān)注我：

MediumTwitterTradingViewYouTube!

中級(jí) Twitter TradingViewYouTube ！

Until next time; stay safe, trade safe!!!

直到下一次; 保持安全，交易安全！！！

Atilla Yurtseven

阿蒂拉·尤爾特斯文(Atilla Yurtseven)

翻譯自: https://medium.com/swlh/trading-with-momentum-channels-in-python-f58a0f3ebd37

動(dòng)量策略 python

總結(jié)

以上是生活随笔為你收集整理的动量策略 python_在Python中使用动量通道进行交易的全部?jī)?nèi)容，希望文章能夠幫你解決所遇到的問(wèn)題。

如果覺得生活随笔網(wǎng)站內(nèi)容還不錯(cuò)，歡迎將生活随笔推薦給好友。