Tick Trading Python === Taiwan R User Group / MLDM Monday @2017-05-09 --- 關於我 === - 生魚片@永豐金證券 --- ```flow st=>start: 資料源 e=>end: $$$ op=>operation: 指標(1..*) op2=>operation: 計算 cond=>condition: 下單 st->op->op2->cond cond(yes)->e cond(no)->op ``` --- # Demo-1 均線操作法 ## [黃金交叉]([https://www.google.com.tw/search?q=%E5%85%A9%E6%A2%9D%E5%9D%87%E7%B7%9A&oq=%E5%85%A9%E6%A2%9D%E5%9D%87%E7%B7%9A&aqs=chrome..69i57.3098j0j4&sourceid=chrome&ie=UTF-8) ## 以 5ma 穿越 20ma 為例子 ---- ## 買進 ```python= cross(ma(c,5),ma(c,20)) ``` ## 賣出 ```python= cross(ma(c,20),ma(c,5)) ``` ---- # 技術分析 ## 1. 容易上手 <!-- .element: class="fragment" data-fragment-index="1" --> - Easy Language Like <!-- .element: class="fragment" data-fragment-index="2" --> - 缺少彈性 <!-- .element: class="fragment" data-fragment-index="3" --> ---- ## 2. 指標工具箱 - 提供解決 <!-- .element: class="fragment" data-fragment-index="1" --> - 黑箱:為何你們的跟OO的數值不一樣 <!-- .element: class="fragment" data-fragment-index="2" --> - 解決問題 or 理解問題 <!-- .element: class="fragment" data-fragment-index="3" -->  <!-- .element: class="fragment" data-fragment-index="4" --> ---- 所以改用  === [2004年12月推出](http://www.sinopac.com/news/release01/news_931207.asp) [官方操作文件](http://www.sinotrade.com.tw/ec/eleader1/pro-trade.pdf) ## 有甚麼不一樣 <!-- .element: class="fragment" data-fragment-index="1" -->  ---- # PYTHON  <!-- .element: class="fragment" data-fragment-index="1" --> --- # Demo-2 均線操作法 * 自己的指標自己刻 <!-- .element: class="fragment" data-fragment-index="1" --> * 無法解決，至少理解 <!-- .element: class="fragment" data-fragment-index="2" --> ---- ```python= def pure_running_mean(data, N): """Simply MA using pure python""" if len(data) <= N: return _sum = 0 result = list(0 for x in data) data = data[::-1] for i in range(0, N): _sum = _sum + data[i] result[i] = _sum/(i+1) for i in range(N, len(data)): _sum = _sum - data[i - N] + data[i] result[i] = _sum / N return result[::-1][:-N] ``` * c = 收盤價 * c[0] 當下最新 * c[1] 前一個Tick * c[-1] 最後一筆資料 ---- ```python= def pure_cross(list_a, list_b): two_list = zip(list_a, list_b) result = list(0 for _ in two_list) for idx, item in enumerate(two_list[:-1]): prev_item = two_list[idx + 1] if prev_item[0] < prev_item[1] and item[0] > item[1]: result[idx] = 1 else: result[idx] = 0 return result ``` ## 穿越的定義 前一個Tick LineA < LineB and 現在的Tick LineA > LineB --- # 當 # 數學方程式 <!-- .element: class="fragment" data-fragment-index="1" --> # 遇見 <!-- .element: class="fragment" data-fragment-index="2" --> # 電腦程式 <!-- .element: class="fragment" data-fragment-index="3" --> ---- ## c3h3  <!-- .element: class="fragment" data-fragment-index="1" --> ---- # Demo-3 均線操作法 ## Numpy <!-- .element: class="fragment" data-fragment-index="1" --> ```python= import numpy def np_running_mean(data, N): cumsum = numpy.cumsum(numpy.insert(data, 0, 0)) return (cumsum[N:] - cumsum[:-N]) / N ``` <!-- .element: class="fragment" data-fragment-index="1" --> [StackOverflow](https://stackoverflow.com/questions/13728392/moving-average-or-running-mean/27681394#27681394) <!-- .element: class="fragment" data-fragment-index="1" --> --- # 自己的指標自己刻 # ohlcv <!-- .element: class="fragment" data-fragment-index="1" --> # 自己的資料自己餵 <!-- .element: class="fragment" data-fragment-index="2" --> ---- 我的策略 - Demo4 == ## 只想在沒人賣的時候進場 <!-- .element: class="fragment" data-fragment-index="1" --> * 委賣筆數為10天內最小的時候進場 <!-- .element: class="fragment" data-fragment-index="2" --> ## 在一堆人賣的時候出場 <!-- .element: class="fragment" data-fragment-index="3" --> * 委賣筆數為10天內最大的時候出場 <!-- .element: class="fragment" data-fragment-index="4" --> :secret: 委賣筆數 :secret: <!-- .element: class="fragment" data-fragment-index="5" --> ---- [證券交易所 - 每5秒委託成交統計](http://www.tse.com.tw/ch/trading/exchange/MI_5MINS/MI_5MINS.php) 江波圖  <!-- .element: class="fragment" data-fragment-index="1" --> ---- 自給自足的資料庫 ==  ---- 委賣筆數為N天內最小 進場 == ```python= def at_low_sell_power(N): sell_power = load_csv_list('ask') asklist = map(float, sell_power) ask = array(0) ask.data = asklist[:len(c)] cond = lowest(ask, N) == ask[0] return cond ``` ---- 委賣筆數為N天內最大 出場 == ```python= def at_high_sell_power(N): sell_power = load_csv_list('ask') asklist = map(float, sell_power) ask = array(0) ask.data = asklist[:len(c)] cond = highest(ask, N) == ask[0] return cond ``` ---- 成果分析 ==  --- 資源 == [Pandas/Numpy on eLeader](http://ypochien.sytes.net:8080/share.cgi?ssid=0Rv4g3F) ## 解壓縮至bin目錄之下 ## import 自己的模組，記得reload ```python= import mypy reload(mypy) mypy.at_low_sell_power(10) ``` ---- 資料不多，加減用 == 畫面看到的皆可用.... ## **item.data** is python list(double) --- 下單API == Python版本 - Windows [GitHub - Sinopac Order API](https://github.com/ypochien/Sinopac-Order-API) <!-- .element: class="fragment" data-fragment-index="1" --> ## Live Demo <!-- .element: class="fragment" data-fragment-index="1" --> --- 報價 == ## [Demo-Web](http://alvin.tw/quote) ## Demo-Python <!-- .element: class="fragment" data-fragment-index="1" --> ## MQTT / Web socket <!-- .element: class="fragment" data-fragment-index="2" --> ## 希望未來能開放 Sinopac Quote API <!-- .element: class="fragment" data-fragment-index="3" --> ---  == # 加入我們 ## 永豐金證券 --- :clock9::question: 有問題: [加入LINE群組](http://line.me/ti/g/4b6wkylcz6)  --- END