Job Application - Quantitative Analyst

0) Interview Sample Question

The sample question for Interview a job in Binary.com. Here I try to write a web application which is automatically gather data, calculate, forecast, place orders, settlement and also P&L report from tip-to-toe. Here I also conducting few research tasks to test the efficiency of some statistical models, and also refer to a Master Degree level quantitave assignment as my studies. Hope that I can be shortlisted to be a member of Binary.com.

1) Question I

1.1) Answer

I use daily OHLCV USDJPY data (from 2014-01-01 to 2017-01-20) and application of some models to forecast the highest and lowest price :

• Auto Arima models
• Exponential Time Series
• Univariate Garch models
• Exponential Weighted Moving Average
• Monte Carlo Markov Chain
• Bayesian Time Series
• Midas

Kindly refer to Binary.com Interview Q1 (Old link or Alternate link or Alternate link 2 (Added MSE comparison)) for more information.

Well, dataset for below papers daily OHLCV of 7 currencies from 2013-01-01 to 2017-08-31:

• AUDUSD
• EURUSD
• GBPUSD
• USDCAD
• USDCHF
• USDCNY
• USDJPY

1. Here I wrote another extention page for Q1 which is analyse the multiple currencies and also models from minutes to daily. You are feel free to browse over Binary.com Interview Q1 (Extention) or (Alternate link).

2. Here I also find the optimal arma order for GARCH models as you can refer to GARCH模型中的ARIMA(p,d,q)参数最优化. binary.com 面试试题 I - GARCH模型中的ARCH in Mean compares the ARCHM with previous Non-ARCHM models.

3. You can also refer to binary.com Interview Question I - Comparison of Univariate GARCH Models which compares the prediction accuracy of 14 GARCH models (not completed) and 9 models (mostly completed from 2013-01-01 to 2017-08-30).

• sGARCH
• fGARCH.GARCH
• fGARCH.TGARCH
• fGARCH.NGARCH
• fGARCH.NAGARCH
• fGARCH.GJRGARCH
• gjrGARCH
• iGARCH
• csGARCH

Besides, I wrote a shinyApp which display the real-time price through API. Kindly refer to Q1App where Q1App2 is another app for financial value betting.

binary.com Interview Question I - Multivariate GARCH Models introduce few multi-variate GARCH models.

• symmetric DCC
• asymmetric DCC
• Flexible DCC
• GO-GARCH
• Copula-GARCH

In order to started the high-frequency-trading statistical modelling, I inspect the dataset via binary.com面试试题 I - 单变量数据缺失值管理 and also binary.com 面试试题 I - 多变量数据缺失值管理 II but the univariate modelling caused some statistical error. The papers compares multi-methods like interpolatan, kalman, locf and ma. The binary.com Interview Question I - Interday High Frequency Trading Models Comparison compares ts, msts, SARIMA, mcsGARCH, midasr, midas-garch, Levy process models.

1.2) Blooper

Initially, I wrote a shiny app (as showing in below gif file) but it is heavily budden for loading. Kindly browse over ShinyApp (Kindly refer to binary.com Interview Question I - Lasso, Elastic-Net and Ridge Regression for more information) which contain the questions and answers of 3 questions. For the staking model, I simply forecast the highest and lowest price, and then :

• Kelly criterion and using highest or lowest price for closing transaction, otherwise using closing price if the forecasted lowest/highest price is not occur.
• Placed $100 an each of the forecasted variance value and do the settlement based on the real variance value.

Secondly, I wrote another app testRealTimeTransc trial version to test the real time trading, and a completed version is Q1App2.

Due to the paper Binary.com Interview Q1 - Tick-Data-HiLo For Daily Trading (Blooper) simulated the data and then only noticed I not yet updated the new function, then I wrote GARCH模型中的ARIMA(p,d,q)参数最优化 to compare the accuracy. However my later paper simulated dataset doesn't save the $fit$ in order to retrieve the $\sigma^2$ and VaR values for stop-loss pips when I got the idea. Here I put it as blooper and start binary-Q1 Multivariate GARCH Models and later on will write another FOREX Day Trade Simulation which will simulate all tick-data but not only HiLo data.

1.3) Shiny Application

• shinyApp : shiny::runGitHub('englianhu/binary.com-interview-question') - Application which compare the accuracy of multiple lasso, ridge and elastic net models (blooper).
• Q1App : shiny::runGitHub('englianhu/binary.com-interview-question', subdir = 'Q1') - the application gather, calculate and forecast price. Once the user select currency and the forecast day, the system will auto calculate and plot the graph.
• testRealTimeTransc : shiny::runGitHub('englianhu/binary.com-interview-question', subdir = 'testRealTimeTransc') - real time trading system which auto gather, calculate the forecast price, and also place orders, as well as settlement and plot P&L everyday.
• Q1App2 : shiny::runGitHub('englianhu/binary.com-interview-question', subdir = 'Q1App2') - The application contain the Banker and Punter section which applied aboved statistical modelling.

2) Question II

2.1) Answer

For question 2, I simply write an app, kindly use Q2App. The bivariate or trivariate poisson model might useful for analyse the probability of fund-in and fund-out by investors in order to manage whole investment pool. Unfortunately there has no such dataset avaiable for fund pool management modelling.

2.2) Shiny Application

• Q2 : shiny::runGitHub('englianhu/binary.com-interview-question', subdir = 'Q2') - An application which applied queuing theory.

3) Question III

For question 3, due to the question doesn't states we only bet on the matches which overcame a certain edge, therefore I just simply list the scenario. Kindly refer to Betting strategy for more informtion.

4) Reference

4.1) Question I

1. Stock Market Forecasting Using LASSO Linear Regression Model by Sanjiban Sekhar Roy, Dishant Mital, Avik Basu, Ajith Abraham (2015)
2. Using LASSO from lars (or glmnet) package in R for variable selection by Juancentro (2014)
3. Difference between glmnet() and cv.glmnet() in R? by Amrita Sawant (2015)
4. Testing Kelly Criterion and Optimal f in R by Roy Wei (2012)
5. Portfolio Optimization and Monte Carlo Simulation by Magnus Erik Hvass Pedersen (2014)
6. Glmnet Vignette by Trevor Hastie and Junyang Qian (2014)
7. lasso怎么用算法实现？ by shuaihuang (2010)
8. The Sparse Matrix and {glmnet} by Manuel Amunategui (2014)
9. Regularization and Variable Selection via the Elastic Net by Hui Zou and Trevor Hastie
10. LASSO, Ridge, and Elastic Net
11. 热门数据挖掘模型应用入门（一）: LASSO回归 by 侯澄钧 (2016)
12. The Lasso Page
13. Call_Valuation.R by Mariano (2016)
14. Lecture 6 – Stochastic Processes and Monte Carlo (http://zorro-trader.com/manual)
15. The caret Package by Max Kuhn (2017)
16. Time Series Cross Validation
17. Character-Code.com
18. Size Matters – Kelly Optimization by Roy Wei (2012)
19. Time Series Cross Validation by William Chiu (2015)
20. Forecasting Volatility by Stephen Figlewski (2004)
21. Successful Algorithmic Trading by Michael Halls Moore (2015)
22. Financial Risk Modelling and Portfolio Optimization with R (2nd Edt) by Bernhard Praff (2016)
23. Analyzing Financial Data and Implementing Financial Models using R by Clifford S.Ang (2015)

4.2) Question II

1. Queueing model 534 in Excel
2. Queueing model macro in Excel
3. Queueing up in R, (continued)
4. Waiting in line, waiting on R
5. Simulating a Queue in R
6. What is the queue data structure in R?
7. Implementing a Queue as a Reference Class
8. queue implementation?
9. Queueing Theory Calculator
10. The Pith of Performance by Neil Gunther (2010)
11. Computationally Efficient Simulation of Queues - The R Package queuecomputer
12. Waiting-Line Models
13. Queues with Breakdowns and Customer Discouragement

4.3) Question III

1. Data APIs/feeds available as packages in R
2. Application of Kelly Criterion model in Sportsbook Investment

Quantitative Trading

1.1) Abstract

There has already 2 years passed by while Binary.com-is-Rebranding-to-Deriv.com, Here I summarized some previous research papers in Binary.com → Deriv.com and continue from this high-frequency-trading. Here I am not apply job but for research purpose.

I moved the data to https://github.com/englianhu/binary.com-interview-question-data.

Below are some seasonal time series models:

• Deriv.com - Interday High Frequency Trading Models Comparison Review (Part I) (in RPubs.com)
• Deriv.com - Interday High Frequency Trading Models Comparison Review (Part I) (in RStudioConnect.com)
• Deriv.com - Interday & Intraday High Frequency Trading Models Comparison Review (Part II) (in RPubs.com)
• Deriv.com - Interday & Intraday High Frequency Trading Models Comparison Review (Part II) (in RStudioConnect.com)

1.2) Blooper

Deriv.com - Interday High Frequency Trading Models Comparison Blooper built seasonal models.

Powered by - Copyright® Intellectual Property Rights of Sςιβrοκεrs Trαdιηg® 個人の経営企業