StockToolsPreview

---

Made by DevTae

This is the repository that summarizes about my own project named DevTae/StockTools

• Project Overview

  • StockDatabase, StockDatabaseHelper
    • Constructing File-System Database
    • Downloading Adjusted Price Datas, Theme, Industry datas
    • Calculating Many Price Indicators as like Ichimoku
    • Auto Update Helper Tool
  • StockBacktester
    • Searching stocks using specific functions with conditions
    • Backtesting Simple Profitability Ratio Result Using Custom Signal
    • Calculating Trendlines on Each Stock Chart
    • Calculating Statistical Clusters about Theme and Industry
  • StockSearcher, AutoTradingBot
    • After the validation process of conditions, that conditions are applied on StockSearcher logic
    • The user could find the attractive stocks, and register them on AutoTradingBot
    • AutoTradingBot is used to buy registered stocks in cheap price using indicators

• Technical Contents

  • StockDatabase Project
    • Saving about 2367 stocks and 9,952,847 daily datas in File-System database
    • Saving of 72% previous processing time in calculating indicator named Leading Span of Ichimoku about a data set of 10 million
  • StockBacktester Project
    • Using Delegate Pattern on Searching Stocks using Conditional Functions
    • Using these backtesting datas to make own buying/selling strategy

• Screenshots in 'StockDatabase' and 'StockBacktester' Project

• Information about Source Distribution

---

Technical Contents

StockDatabase Project

Saving about 2367 stocks and 9,952,847 daily datas in File-System database

• I have saved about a lot of stock datas in File-System database from Korea market (KOSPI, KOSDAQ) until 2023/02/17.
  • There are so many informations as like Stock Price, Volume, Adjusted Stock Price, MarketCap in each daily data.
  • Below this, that is the structure of class diagram.
  • Example code is here

📦Stock
 ┣ 📂StockInfo             // 종목에 대한 정보가 담겨 있음.
 ┃ ┣ 📂Market
 ┃ ┃ ┣ 📜Country           // ex. "한국주식(키움)", ...
 ┃ ┃ ┣ 📜Code              // ex. 0, 1, 2, ...
 ┃ ┃ ┣ 📜Name              // ex. "코스피", "코스닥", ...
 ┃ ┃ ┗ 📜Index             // Static Variable 에 있는 Country 배열의 Index
 ┃ ┣ 📜Code                // ex. "005930"
 ┃ ┣ 📜Name                // ex. "삼성전자"
 ┃ ┣ 📜DebutedDate         // ex. "20190301" (상장일)
 ┃ ┗ 📜ActiveSharesRatio   // ex. 60.5 (유통비율)
 ┣ 📂PriceData
 ┃ ┣ 📂DailyData
 ┃ ┃ ┣ 📜Date
 ┃ ┃ ┣ 📜Open              // 시가
 ┃ ┃ ┣ 📜AdjustedOpen      // 시가 수정주가
 ┃ ┃ ┣ 📜High              // 고가
 ┃ ┃ ┣ 📜AdjustedHigh      // 고가 수정주가
 ┃ ┃ ┣ 📜Low               // 저가
 ┃ ┃ ┣ 📜AdjustedLow       // 저가 수정주가
 ┃ ┃ ┣ 📜Close             // 종가
 ┃ ┃ ┣ 📜AdjustedClose     // 종가 수정주가
 ┃ ┃ ┣ 📜Volume            // 거래량
 ┃ ┃ ┣ 📜AdjustedVolume    // 수정 거래량
 ┃ ┃ ┣ 📜Shares            // 상장 주식 수
 ┃ ┃ ┣ 📜MarketCap         // 시가총액
 ┃ ┃ ┣ 📜AdjustedEvent     // 수정주가 이벤트 발생 코드 (in Kiwoom API docs)
 ┃ ┃ ┣ 📜AdjustedRatio     // 수정주가 변동 비율
 ┃ ┃ ┗ 📜CheckAdjusted     // 수정주가 변동 여부 확인 column
 ┃ ┗ 📂WeeklyData
 ┣ 📜LastUpdatedDate       // 마지막 업데이트 날짜
 ┣ 📜LastAdjustedDate      // 마지막 수정주가 변동 날짜
 ┗ 📜LastAdjustedIndex     // 마지막 수정주가 변동 index

Saving of 72% previous processing time in calculating indicator named Leading Span of Ichimoku about a data set of 10 million

There are two ways to get the Maximum and Minimum Value in Specific Range to calculate the Leading Span of Ichimoku (n : a number of all daily datas contained every stocks)

• Calculate the Maximum and Minimum value using the Linear way

  • Whenever the index is changed, try to get a maximum and minimum value calculating every 52 elements
  • It would be needed the time Θ(52 * n) to calculate all daily datas on each stock

• Calculate the Maximum and Minimum value using the Segment Tree Algorithm

  • Before calculating, make a Segment Tree (it needs Θ(n * log(n)))
  • Whenever the index is changed, try to get a maximum and minimum value using Segment Tree
  • It would be needed the time Θ(log(n) * n) to calculate all daily datas on each stock

• Θ(52 * n) vs Θ(log(n) * n)

  • I estimated that n is the average of having daily datas on each stock
  • n = 9,952,847 daily datas / 2367 stocks = 4204
  • Θ(52 * n) = 218,608 vs Θ(log(n) * n) = 50,448
  • The winner is the Θ(log(n) * n)

While developing the logic, I applied the Segment Tree Algorithm.

As a result, I made a saving of 72% previous processing time about a data set of 10 million. (You could see the detailed process in here)

---

StockBacktester Project

Using Delegate Pattern on Searching Stocks using Conditional Functions

• Below codes are the examples of searching condition function

  // it will return true, when "20 SMA crosses up 60 SMA" using simple moving average "while remaining convergence"
  public static bool IsCrossUpCandle(ref DailyData[] dailyDatas, ref Indicator[] indicators, int start_index, int index)
  {
    if(index <= 0) return false;
  
    if(index - start_index < 5) return false;
  
    // 1. 20 SMA < 60 SMA * 1.05
    float SMA20, SMA60;
    if(float.TryParse(indicators[index].Values[(int)IndicatorDailyCode.SMA20], out SMA20)
    && float.TryParse(indicators[index].Values[(int)IndicatorDailyCode.SMA60], out SMA60)) {
      if(SMA20 < SMA60 * 1.05) {
        // 2. pivot close CrossUp
        float close_p = float.Parse(dailyDatas[start_index].AdjustedClose);
        float close_0 = float.Parse(dailyDatas[index].AdjustedClose);
        float close_1 = float.Parse(dailyDatas[index - 1].AdjustedClose);
        if(close_0 > close_p && close_1 < close_p) {
          return true;
        }
      }
    }
    return false;
  }

• I used delegate variable for reducing code complexity when I needed to use a lot of conditional functions.

  public delegate bool CondFunc(ref DailyData[] dailyDatas, ref Indicator[] indicators, int point_index2, int index);
  ...
  // select condFunc using switch operation and so on
  CondFunc condFunc = IsCrossUpCandle;
  ...
  // get daily datas and indicator of stocks
  bool result = this.condFunc(ref dailyDatas, ref indicators, 5, 0);
  ...
  // search and analyze results ...
  ...

Using these backtesting datas to make own buying/selling strategy

• It could be totally possible to analyze the price patterns of stock using this program with backtesting results.
  • Until now, I have studied and analyzed so many stocks and price patterns using this program
  • Moreover, I tried to analyze the stock price using Deep Learning Model.
    • You could see on this link

---

Screenshots in 'StockDatabase' and 'StockBacktester' Project

• Update the daily datas automatically from Kiwoom API and Naver Finance in asynchronous way

• Backtesting Simulation with many buying/selling strategy as like SMA Golden-Cross Signal (on the bottom right)

• Showing the chart of searched backtesting results with many indicators like trendlines, theme clusters

---

Information about Source Distribution

• The source code of this project is saved on private repository, because I worried about the misusing some strategies in this project to their buying / selling decision.

• Whenever if you have any questions for this project, contact me for anything using email.

  • e-mail is here. → [email protected]

---

The End

Thank you for reading!