Skip to content

Commit ef90b61

Browse files
englianhuenglianhu
committed
refresh
1 parent 599ca24 commit ef90b61

13 files changed

+356
-37669
lines changed

‎ARIMA-Adjustment-in-GARCH-Model.Rmd

-532
This file was deleted.

‎ARIMA-Adjustment-in-GARCH-Model.html

-1,905
This file was deleted.

‎README.html

+1-1
Original file line numberDiff line numberDiff line change
@@ -145,7 +145,7 @@ <h3>Question I</h3>
145145
<li><s>Bayesian Time Series</s></li>
146146
<li><s>Midas</s></li>
147147
</ul>
148-
<p>Kindly refer to <a href="http://rpubs.com/englianhu/binary-Q1">Binary.com Interview Q1</a> (<a href="https://englianhu.github.io/2017/09/binary-forex-trading-Q1.html">Old link</a> or <a href="http://rpubs.com/englianhu/binary-forex-trading-Q1">Alternate link</a>) for more information.</p>
148+
<p>Kindly refer to <a href="http://rpubs.com/englianhu/binary-Q1">Binary.com Interview Q1</a> (<a href="https://englianhu.github.io/2017/09/binary-forex-trading-Q1.html">Old link</a> or <a href="http://rpubs.com/englianhu/binary-forex-trading-Q1">Alternate link</a> or <a href="http://rpubs.com/englianhu/binary-Q1-Added">Alternate link 2 (Added MSE comparison)</a>) for more information.</p>
149149
<p>Here I wrote another extention page for Q1 which is analyse the multiple currencies and also models <s>from minutes to</s> daily. You are feel free to browse over <a href="http://rpubs.com/englianhu/binary-Q1E">Binary.com Interview Q1 (Extention)</a> or (<a href="http://rpubs.com/englianhu/316133">Alternate link</a>).</p>
150150
<p>Here I also find the optimal arma order for GARCH models as you can refer to <a href="http://rpubs.com/englianhu/binary-Q1FiGJRGARCH">GARCH模型中的<code>ARMA(p,d,q)</code>参数最优化</a>.</p>
151151
<p>Besides, I wrote a shinyApp which display the real-time price through API. Kindly refer to <a href="https://beta.rstudioconnect.com/content/3073/">Q1App</a> where <a href="https://beta.rstudioconnect.com/content/3138/">Q1App2</a> is another app for financial value betting.</p>

‎README.md

+1-1
Original file line numberDiff line numberDiff line change
@@ -19,7 +19,7 @@ I use daily OHLCV USDJPY data (from 2014-01-01 to 2017-01-20) and application of
1919
- <s>Bayesian Time Series</s>
2020
- <s>Midas</s>
2121

22-
Kindly refer to [Binary.com Interview Q1](http://rpubs.com/englianhu/binary-Q1) ([Old link](https://englianhu.github.io/2017/09/binary-forex-trading-Q1.html) or [Alternate link](http://rpubs.com/englianhu/binary-forex-trading-Q1)) for more information.
22+
Kindly refer to [Binary.com Interview Q1](http://rpubs.com/englianhu/binary-Q1) ([Old link](https://englianhu.github.io/2017/09/binary-forex-trading-Q1.html) or [Alternate link](http://rpubs.com/englianhu/binary-forex-trading-Q1) or [Alternate link 2 (Added MSE comparison)](http://rpubs.com/englianhu/binary-Q1-Added)) for more information.
2323

2424
Here I wrote another extention page for Q1 which is analyse the multiple currencies and also models <s>from minutes to</s> daily. You are feel free to browse over [Binary.com Interview Q1 (Extention)](http://rpubs.com/englianhu/binary-Q1E) or ([Alternate link](http://rpubs.com/englianhu/316133)).
2525

‎binary-Q1.Rmd

+263-4
Original file line numberDiff line numberDiff line change
@@ -11,6 +11,7 @@ output:
1111
toc_float:
1212
collapsed: no
1313
smooth_scroll: no
14+
self_contained: no
1415
tufte::tufte_html:
1516
toc: yes
1617
toc_depth: 4
@@ -2718,19 +2719,277 @@ pl
27182719

27192720
#### MIDAS
27202721

2722+
### Comparison of Accuracy
27212723

2722-
### Conclusion
2724+
In order to validate the accuracy of prediction, here I added this section to compare among the models. I added a mean mse for Hi and Lo in order to know the long term accuracy between Hi-Lo since an order stand upon the accuracy of both Hi and Lo. **binary-Q1 Multivariate GARCH Models** will be another study for both punter and banker.
2725+
2726+
```{r mse1}
2727+
## read files
2728+
fls <- list.files('data', patter = '.snorm.hybrid.rds$')
2729+
mds <- llply(fls, function(txt) {
2730+
tryCatch(readRDS(paste0('data/', txt)),
2731+
error = function(e) cat(paste0('data/', txt, ' error!\n')))
2732+
})
2733+
names(mds) <- str_replace_all(fls, '.rds', '')
2734+
2735+
## measure mse.
2736+
mds <- llply(names(mds), function(x) {
2737+
HiLo <- x %>% str_extract_all('Hi.Lo|Lo.Hi|Hi.Cl|Lo.Cl|Op.Hi|
2738+
Op.Lo|Op.Cl') %>% unlist
2739+
if (length(HiLo) > 0) {
2740+
dfm <- mds[[x]]
2741+
2742+
if (HiLo == 'Hi.Lo') {
2743+
dfm %<>% mutate(mse.Hi = mean((Point.Forecast - USDJPY.High)^2),
2744+
mse.Lo = mean((forClose - USDJPY.Low)^2)) %>% unique
2745+
2746+
} else if (HiLo == 'Lo.Hi') {
2747+
dfm %<>% mutate(mse.Lo = mean((Point.Forecast - USDJPY.Low)^2),
2748+
mse.Hi = mean((forClose - USDJPY.High)^2)) %>% unique
2749+
2750+
} else if (HiLo == 'Hi.Cl') {
2751+
dfm %<>% mutate(mse.Hi = mean((Point.Forecast - USDJPY.High)^2),
2752+
mse.Cl = mean((forClose - USDJPY.Close)^2)) %>% unique
2753+
2754+
} else if (HiLo == 'Lo.Cl') {
2755+
dfm %<>% mutate(mse.Lo = mean((Point.Forecast - USDJPY.Low)^2),
2756+
mse.Cl = mean((forClose - USDJPY.Close)^2)) %>% unique
2757+
2758+
} else if (HiLo == 'Op.Hi') {
2759+
dfm %<>% mutate(mse.Op = mean((Point.Forecast - USDJPY.Open)^2),
2760+
mse.Hi = mean((forClose - USDJPY.High)^2)) %>% unique
2761+
2762+
} else if (HiLo == 'Op.Lo') {
2763+
dfm %<>% mutate(mse.Op = mean((Point.Forecast - USDJPY.Open)^2),
2764+
mse.Lo = mean((forClose - USDJPY.Low)^2)) %>% unique
2765+
2766+
} else if (HiLo == 'Op.Cl') {
2767+
dfm %<>% mutate(mse.Op = mean((Point.Forecast - USDJPY.Open)^2),
2768+
mse.Cl = mean((forClose - USDJPY.Close)^2)) %>% unique
2769+
2770+
} else {
2771+
stop('No such option.')
2772+
}
2773+
dfm %<>% .[, -c(1:26)] %>% unique
2774+
}
2775+
})
2776+
names(mds) <- str_replace_all(fls, '.rds', '')
2777+
2778+
mds %<>% ldply
2779+
2780+
mds %>% mutate(mse.HiLo = (mse.Hi + mse.Lo)/2) %>%
2781+
kable(caption = 'Summary') %>%
2782+
kable_styling(bootstrap_options = c('striped', 'hover', 'condensed', 'responsive')) %>%
2783+
scroll_box(width = '100%', height = '400px')
2784+
```
2785+
2786+
Here I filtered the `sd < 20%` dataset as below.
2787+
2788+
```{r mse2}
2789+
## read files
2790+
fls <- list.files('data', patter = '.snorm.hybrid.rds$')
2791+
mds <- llply(fls, function(txt) {
2792+
tryCatch(readRDS(paste0('data/', txt)),
2793+
error = function(e) cat(paste0('data/', txt, ' error!\n')))
2794+
})
2795+
names(mds) <- str_replace_all(fls, '.rds', '')
2796+
2797+
## measure mse.
2798+
mds <- llply(names(mds), function(x) {
2799+
HiLo <- x %>% str_extract_all('Hi.Lo|Lo.Hi|Hi.Cl|Lo.Cl|Op.Hi|
2800+
Op.Lo|Op.Cl') %>% unlist
2801+
if (length(HiLo) > 0) {
2802+
dfm <- mds[[x]]
2803+
2804+
if (HiLo == 'Hi.Lo') {
2805+
dfm %<>% mutate(diff1 = abs(Point.Forecast/USDJPY.High),
2806+
se1 = ifelse(diff1 <= 0.8 | diff1 >= 1.25, 1, 0),
2807+
diff2 = abs(forClose/USDJPY.Low),
2808+
se2 = ifelse(diff2 <= 0.8 | diff2 >= 1.25, 1, 0)
2809+
) %>%
2810+
dplyr::filter(se1 != 1 & se2 != 1) %>%
2811+
mutate(mse.Hi = mean((Point.Forecast - USDJPY.High)^2),
2812+
mse.Lo = mean((forClose - USDJPY.Low)^2)) %>% unique
2813+
2814+
} else if (HiLo == 'Lo.Hi') {
2815+
dfm %<>% mutate(diff1 = abs(Point.Forecast/USDJPY.Low),
2816+
se1 = ifelse(diff1 <= 0.8 | diff1 >= 1.25, 1, 0),
2817+
diff2 = abs(forClose/USDJPY.High),
2818+
se2 = ifelse(diff2 <= 0.8 | diff2 >= 1.25, 1, 0)
2819+
) %>%
2820+
dplyr::filter(se1 != 1 & se2 != 1) %>%
2821+
mutate(mse.Lo = mean((Point.Forecast - USDJPY.Low)^2),
2822+
mse.Hi = mean((forClose - USDJPY.High)^2)) %>% unique
2823+
2824+
} else if (HiLo == 'Hi.Cl') {
2825+
dfm %<>% mutate(diff1 = abs(Point.Forecast/USDJPY.High),
2826+
se1 = ifelse(diff1 <= 0.8 | diff1 >= 1.25, 1, 0),
2827+
diff2 = abs(forClose/USDJPY.Close),
2828+
se2 = ifelse(diff2 <= 0.8 | diff2 >= 1.25, 1, 0)
2829+
) %>%
2830+
dplyr::filter(se1 != 1 & se2 != 1) %>%
2831+
mutate(mse.Hi = mean((Point.Forecast - USDJPY.High)^2),
2832+
mse.Cl = mean((forClose - USDJPY.Close)^2)) %>% unique
2833+
2834+
} else if (HiLo == 'Lo.Cl') {
2835+
dfm %<>% mutate(diff1 = abs(Point.Forecast/USDJPY.Low),
2836+
se1 = ifelse(diff1 <= 0.8 | diff1 >= 1.25, 1, 0),
2837+
diff2 = abs(forClose/USDJPY.Close),
2838+
se2 = ifelse(diff2 <= 0.8 | diff2 >= 1.25, 1, 0)
2839+
) %>%
2840+
dplyr::filter(se1 != 1 & se2 != 1) %>%
2841+
mutate(mse.Lo = mean((Point.Forecast - USDJPY.Low)^2),
2842+
mse.Cl = mean((forClose - USDJPY.Close)^2)) %>% unique
2843+
2844+
} else if (HiLo == 'Op.Hi') {
2845+
dfm %<>% mutate(diff1 = abs(Point.Forecast/USDJPY.Open),
2846+
se1 = ifelse(diff1 <= 0.8 | diff1 >= 1.25, 1, 0),
2847+
diff2 = abs(forClose/USDJPY.High),
2848+
se2 = ifelse(diff2 <= 0.8 | diff2 >= 1.25, 1, 0)
2849+
) %>%
2850+
dplyr::filter(se1 != 1 & se2 != 1) %>%
2851+
mutate(mse.Op = mean((Point.Forecast - USDJPY.Open)^2),
2852+
mse.Hi = mean((forClose - USDJPY.High)^2)) %>% unique
2853+
2854+
} else if (HiLo == 'Op.Lo') {
2855+
dfm %<>% mutate(diff1 = abs(Point.Forecast/USDJPY.Open),
2856+
se1 = ifelse(diff1 <= 0.8 | diff1 >= 1.25, 1, 0),
2857+
diff2 = abs(forClose/USDJPY.Low),
2858+
se2 = ifelse(diff2 <= 0.8 | diff2 >= 1.25, 1, 0)
2859+
) %>%
2860+
dplyr::filter(se1 != 1 & se2 != 1) %>%
2861+
mutate(mse.Op = mean((Point.Forecast - USDJPY.Open)^2),
2862+
mse.Lo = mean((forClose - USDJPY.Low)^2)) %>% unique
2863+
2864+
} else if (HiLo == 'Op.Cl') {
2865+
dfm %<>% mutate(diff1 = abs(Point.Forecast/USDJPY.Open),
2866+
se1 = ifelse(diff1 <= 0.8 | diff1 >= 1.25, 1, 0),
2867+
diff2 = abs(forClose/USDJPY.Close),
2868+
se2 = ifelse(diff2 <= 0.8 | diff2 >= 1.25, 1, 0)
2869+
) %>%
2870+
dplyr::filter(se1 != 1 & se2 != 1) %>%
2871+
mutate(mse.Op = mean((Point.Forecast - USDJPY.Open)^2),
2872+
mse.Cl = mean((forClose - USDJPY.Close)^2)) %>% unique
2873+
2874+
} else {
2875+
stop('No such option.')
2876+
}
2877+
dfm %<>% .[, -c(1:30)] %>% unique
2878+
}
2879+
})
2880+
names(mds) <- str_replace_all(fls, '.rds', '')
2881+
2882+
mds %<>% ldply
2883+
2884+
mds %>% mutate(mse.HiLo = (mse.Hi + mse.Lo)/2) %>%
2885+
kable(caption = 'Summary') %>%
2886+
kable_styling(bootstrap_options = c('striped', 'hover', 'condensed', 'responsive')) %>%
2887+
scroll_box(width = '100%', height = '400px')
2888+
```
2889+
2890+
Now I filtered the forecast price must be within highest and lowest price in order to made the order stand.
2891+
2892+
```{r mse3}
2893+
## read files
2894+
fls <- list.files('data', patter = '.snorm.hybrid.rds$')
2895+
mds <- llply(fls, function(txt) {
2896+
tryCatch(readRDS(paste0('data/', txt)),
2897+
error = function(e) cat(paste0('data/', txt, ' error!\n')))
2898+
})
2899+
names(mds) <- str_replace_all(fls, '.rds', '')
2900+
2901+
## measure mse.
2902+
mds <- llply(names(mds), function(x) {
2903+
HiLo <- x %>% str_extract_all('Hi.Lo|Lo.Hi|Hi.Cl|Lo.Cl|Op.Hi|
2904+
Op.Lo|Op.Cl') %>% unlist
2905+
if (length(HiLo) > 0) {
2906+
dfm <- mds[[x]]
2907+
2908+
if (HiLo == 'Hi.Lo') {
2909+
dfm %<>% dplyr::filter(Point.Forecast <= USDJPY.High &
2910+
Point.Forecast >= USDJPY.Low &
2911+
forClose <= USDJPY.High &
2912+
forClose >= USDJPY.Low) %>%
2913+
mutate(mse.Hi = mean((Point.Forecast - USDJPY.High)^2),
2914+
mse.Lo = mean((forClose - USDJPY.Low)^2)) %>% unique
2915+
2916+
} else if (HiLo == 'Lo.Hi') {
2917+
dfm %<>% dplyr::filter(Point.Forecast <= USDJPY.High &
2918+
Point.Forecast >= USDJPY.Low &
2919+
forClose <= USDJPY.High &
2920+
forClose >= USDJPY.Low) %>%
2921+
mutate(mse.Lo = mean((Point.Forecast - USDJPY.Low)^2),
2922+
mse.Hi = mean((forClose - USDJPY.High)^2)) %>% unique
2923+
2924+
} else if (HiLo == 'Hi.Cl') {
2925+
dfm %<>% dplyr::filter(Point.Forecast <= USDJPY.High &
2926+
Point.Forecast >= USDJPY.Low &
2927+
forClose <= USDJPY.High &
2928+
forClose >= USDJPY.Low) %>%
2929+
mutate(mse.Hi = mean((Point.Forecast - USDJPY.High)^2),
2930+
mse.Cl = mean((forClose - USDJPY.Close)^2)) %>% unique
2931+
2932+
} else if (HiLo == 'Lo.Cl') {
2933+
dfm %<>% dplyr::filter(Point.Forecast <= USDJPY.High &
2934+
Point.Forecast >= USDJPY.Low &
2935+
forClose <= USDJPY.High &
2936+
forClose >= USDJPY.Low) %>%
2937+
mutate(mse.Lo = mean((Point.Forecast - USDJPY.Low)^2),
2938+
mse.Cl = mean((forClose - USDJPY.Close)^2)) %>% unique
2939+
2940+
} else if (HiLo == 'Op.Hi') {
2941+
dfm %<>% dplyr::filter(Point.Forecast <= USDJPY.High &
2942+
Point.Forecast >= USDJPY.Low &
2943+
forClose <= USDJPY.High &
2944+
forClose >= USDJPY.Low) %>%
2945+
mutate(mse.Op = mean((Point.Forecast - USDJPY.Open)^2),
2946+
mse.Hi = mean((forClose - USDJPY.High)^2)) %>% unique
2947+
2948+
} else if (HiLo == 'Op.Lo') {
2949+
dfm %<>% dplyr::filter(Point.Forecast <= USDJPY.High &
2950+
Point.Forecast >= USDJPY.Low &
2951+
forClose <= USDJPY.High &
2952+
forClose >= USDJPY.Low) %>%
2953+
mutate(mse.Op = mean((Point.Forecast - USDJPY.Open)^2),
2954+
mse.Lo = mean((forClose - USDJPY.Low)^2)) %>% unique
2955+
2956+
} else if (HiLo == 'Op.Cl') {
2957+
dfm %<>% dplyr::filter(Point.Forecast <= USDJPY.High &
2958+
Point.Forecast >= USDJPY.Low &
2959+
forClose <= USDJPY.High &
2960+
forClose >= USDJPY.Low) %>%
2961+
mutate(mse.Op = mean((Point.Forecast - USDJPY.Open)^2),
2962+
mse.Cl = mean((forClose - USDJPY.Close)^2)) %>% unique
2963+
2964+
} else {
2965+
stop('No such option.')
2966+
}
2967+
dfm %<>% .[, -c(1:26)] %>% unique
2968+
}
2969+
})
2970+
names(mds) <- str_replace_all(fls, '.rds', '')
27232971
2972+
mds %<>% ldply
27242973
2725-
## Question 2
2974+
mds %>% mutate(mse.HiLo = (mse.Hi + mse.Lo)/2) %>%
2975+
kable(caption = 'Summary') %>%
2976+
kable_styling(bootstrap_options = c('striped', 'hover', 'condensed', 'responsive')) %>%
2977+
scroll_box(width = '100%', height = '400px')
2978+
```
2979+
2980+
### Conclusion
27262981

2982+
#### Conclusion Words
27272983

2728-
## Question 3
2984+
Based on the central limit theorem, the better accurate model will produced higher returns in long run. In this paper I compare couples models and concludes that the `gjrGARCH` model generates highest return in USD/JPY.
27292985

2986+
The article [Does anything NOT beat the GARCH(1,1)?](http://www.unstarched.net/2013/01/07/does-anything-not-beat-the-garch11/) compares couple models but also compare the Garch order GARCH(1,1) as well.
27302987

2731-
# 3. Conclusion
2988+
My later paper **GARCH模型中的ARMA(p,d,q)参数最优化** introduced a better accurate Fractional Intergrated GARCH model.
27322989

2990+
#### Future Works
27332991

2992+
GARCH(1,1) and multivariate GARCH models as well as better staking model required in order to generates higher return.
27342993

27352994
```{r stopPar, echo = FALSE, warning = FALSE}
27362995
## Set options back to original options

‎binary-Q1.html

-34,951
This file was deleted.

‎binary-Q1Multi-GARCH.Rmd

+20
Original file line numberDiff line numberDiff line change
@@ -591,6 +591,26 @@ rm(sys1, sys2)
591591
05. [Financial Econometrics Practical - Univariate Volatility Modelling](https://raw.githubusercontent.com/englianhu/binary.com-interview-question/master/reference/Financial%20Econometrics%20Practical%20-%20Univariate%20Volatility%20Modelling.pdf)
592592
06. [The GARCH-DCC Model and 2-Stage DCC(MVT) Estimation](http://www.unstarched.net/2013/01/03/the-garch-dcc-model-and-2-stage-dccmvt-estimation/)
593593
07. [Multivariate Volatility Forecasting, Part 2 – Equicorrelation](https://eranraviv.com/multivariate-volatility-forecasting-2/)
594+
29. [The Kelly Criterion - Implementation, Simulation and Backtest](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/The%20Kelly%20Criterion%20-%20Implementation%2C%20Simulation%20and%20Backtest.pdf)
595+
30. [The Kelly Criterion and Bet Comparison in Spread Betting](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/The%20Kelly%20Criterion%20and%20Bet%20Comparisons%20in%20Spread%20Betting.pdf)
596+
31. [The Kelly Criterion for Spread Bets](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/The%20Kelly%20Criterion%20for%20Spread%20Bets.pdf)
597+
32. [The Market for English Premier League (EPL) Odds](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/The%20market%20for%20English%20Premier%20League%20(EPL)%20Odds.pdf)
598+
33. [Valuation of Soccer Spread Bets](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/Valuation%20of%20Soccer%20Spread%20Bets.pdf)
599+
28. [Stochastic Modelling and Optimization Methods in Investments](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/Stochastic%20Modeling%20and%20Optimization%20Methods%20in%20Investments.pdf)
600+
15. [How Does the Fortune's Formula-Kelly Capital Growth Model Perform](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/How%20does%20the%20Fortune's%20Formula-Kelly%20Capital%20Growth%20Model%20Perform.pdf)
601+
16. [Information Theory and Gambling or Economics](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/Information%20Theory%20and%20Gambling%20or%20Economics.pdf)
602+
17. [Investment Portfolio Optimization with GARCH Models](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/Investment%20Portfolio%20Optimization%20with%20GARCH%20Models.pdf)
603+
18. [Kelly Criterion Revisited - Optimal Bets](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/Kelly%20Criterion%20Revisited%20-%20Optimal%20Bets.pdf)
604+
19. [Markov-Switching Autoregressive Models for Wind Times Series (ppt)](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/Markov-Switching%20Autoregressive%20Models%20for%20Wind%20Time%20Series%20(ppt).pdf)
605+
20. [Markov-Switching Autoregressive Models for Wind Times Series](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/Markov-Switching%20Autoregressive%20Models%20for%20Wind%20Time%20Series.pdf)
606+
21. [Medium Term Simulations of the Full Kelly and Fractional](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/Medium%20Term%20Simulations%20of%20The%20Full%20Kelly%20and%20Fractional.pdf)
607+
22. [Modelling Exchange Rates using Regime Switching Models](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/Modelling%20Exchange%20Rates%20using%20Regime%20Switching%20Models.pdf)
608+
23. [Money Management (V1)](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/Money%20Management%20(V1).pdf)
609+
24. [Money Management (V2)](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/Money%20Management%20(V2).pdf)
610+
25. [Optimal Betting under Parameter Uncertainty - Improving the Kelly Criterion](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/Optimal%20Betting%20under%20Parameter%20Uncertainty%20-%20Improving%20the%20Kelly%20Criterion.pdf)
611+
13. [Enhancing Trading Strategies with Order Book Signals](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/Enhancing%20Trading%20Strategies%20with%20Order%20Book%20Signals.pdf)
612+
10. [Creating Optimal Portfolios of Stocks with Time-Varying Risk](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/Creating%20Optimal%20Portfolios%20of%20Stocks%20with%20Time-Varying%20Risk.pdf)
613+
11. [Dynamic Portfolio Optimization using Generalized Dynamic Conditoinal Heteroskedastic Factor Models](https://github.com/englianhu/binary.com-interview-question/blob/master/reference/Dynamic%20Portfolio%20Optimization%20using%20Generalized%20Dynamic%20Conditional%20Heteroskedastic%20Factor%20Models.pdf)
594614

595615
---
596616

0 commit comments

Comments
 (0)