In recent times, retail options trading has surged, and according to a (Barclays Report), the influx of retail traders has significantly affected market volatility, particularly in large-cap tech stocks. Retail traders have increasingly influenced implied volatility levels, creating opportunities for advanced strategies like short volatility trades. One way to take advantage of this is through a VolScore-based strategy, which measures the difference between a stock option's implied volatility and its sector's realized volatility.
In this article, we explore how to use VolScore for trading NVIDIA (NVDA) options by comparing its implied volatility to that of its sector, composed of stocks like Microsoft (MSFT), Apple (AAPL), and AMD. Let’s dive into the strategy and its backtest results. I also reccomend watching this video on youtube, explaining the Barclays report in a simpler way (Link to video)
We will:
- Gather implied volatilities and realized volatilities for the selected stocks.
- Calculate the sector's realized volatility.
- Compare NVDA's implied volatility to the sector's realized volatility to generate a VolScore.
- Develop a trading strategy based on VolScore thresholds.
- Backtest the strategy and plot the results.
We have collected implied volatilities and realized volatilities for MSFT, AAPL, and AMD. The data is stored in JSON format.
knitr::opts_chunk$set(warning = FALSE, message = FALSE)
# Load libraries
library(jsonlite)
library(ggplot2)
# Simulated time-series data for implied and realized volatilities from JSON files
msft_data <- fromJSON("MSFT.json")
aapl_data <- fromJSON("AAPL.json")
amd_data <- fromJSON("AMD.json")
# Convert to data frames
msft_df <- as.data.frame(msft_data$data)
aapl_df <- as.data.frame(aapl_data$data)
amd_df <- as.data.frame(amd_data$data)We calculate the sector realized volatility by averaging the realized volatilities of MSFT, AAPL, and AMD.
# Calculate sector-wide realized volatility
# Convert realized volatility columns to numeric (if not already)
msft_df = na.omit(msft_df)
aapl_df = na.omit(aapl_df)
amd_df = na.omit(amd_df)
msft_df$realized_volatility <- as.numeric(msft_df$realized_volatility)
aapl_df$realized_volatility <- as.numeric(aapl_df$realized_volatility)
amd_df$realized_volatility <- as.numeric(amd_df$realized_volatility)
msft_df$implied_volatility <- as.numeric(msft_df$implied_volatility)
aapl_df$implied_volatility <- as.numeric(aapl_df$implied_volatility)
amd_df$implied_volatility <- as.numeric(amd_df$implied_volatility)
msft_df$price <- as.numeric(msft_df$price)
aapl_df$price <- as.numeric(aapl_df$price)
amd_df$price <- as.numeric(amd_df$price)
msft_df$date <- as.Date(msft_df$date)
aapl_df$date <- as.Date(aapl_df$date)
amd_df$date <- as.Date(amd_df$date)
summary(msft_df)## date implied_volatility price realized_volatility
## Min. :2023-10-10 Min. :0.1613 Min. :326.7 Min. :0.1250
## 1st Qu.:2024-01-03 1st Qu.:0.1938 1st Qu.:378.0 1st Qu.:0.1705
## Median :2024-03-27 Median :0.2165 Median :409.4 Median :0.1892
## Mean :2024-03-27 Mean :0.2279 Mean :403.4 Mean :0.1918
## 3rd Qu.:2024-06-20 3rd Qu.:0.2636 3rd Qu.:423.0 3rd Qu.:0.2141
## Max. :2024-09-13 Max. :0.3457 Max. :467.6 Max. :0.2713
# Calculate sector-wide realized volatility
sector_realized_vol <- rowMeans(cbind(msft_df$realized_volatility, aapl_df$realized_volatility, amd_df$realized_volatility), na.rm = TRUE)
sector_realized_vol = na.omit(sector_realized_vol)
summary(sector_realized_vol)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.2058 0.2734 0.2956 0.2976 0.3165 0.3936
# Plot the sector realized volatility
plot(msft_df$date, sector_realized_vol, type = "l", col = "blue", main = "Sector Realized Volatility", xlab = "Date", ylab = "Realized Volatility")
Now, we calculate the VolScore for NVDA, which is the difference between NVDA's implied volatility and the sector's realized volatility, normalized by the sector realized volatility.
# Simulated NVDA data
nvda_data <- fromJSON("NVDA.json")
nvda_df <- as.data.frame(nvda_data$data)
nvda_df = na.omit(nvda_df)
# Convert 'implied_volatility' in nvda_df to numeric
nvda_df$implied_volatility <- as.numeric(nvda_df$implied_volatility)
nvda_df$date = as.Date(nvda_df$date)
# Calculate VolScore
nvda_df$VolScore <- (nvda_df$implied_volatility - sector_realized_vol) / sector_realized_vol
nvda_df$VolScore## [1] 0.25911963 0.24504680 0.27318219 0.26107159 0.26278263 0.28118220
## [7] 0.34886623 0.33593256 0.34127584 0.76256463 0.73199414 0.78765807
## [13] 0.97564071 1.11525125 1.04042688 1.04853343 0.99657040 1.45401088
## [19] 1.41490680 1.47648142 1.45487798 1.02317504 1.05205103 0.95054729
## [25] 1.05352060 1.01963509 1.05566304 1.03568629 1.05141537 0.98461807
## [31] 0.94412456 0.45162581 0.46271193 0.39720986 0.33369372 0.32078467
## [37] 0.33954072 0.16542307 0.24318803 0.24705368 0.28063789 0.15534638
## [43] 0.26861060 0.35771020 0.37837995 0.38427552 0.41062987 0.28644594
## [49] 0.26531081 0.15942551 0.18901705 0.16151819 0.09542627 0.11707560
## [55] 0.09871148 0.05473755 0.04107436 0.07646779 0.15489835 0.08594149
## [61] 0.04904109 0.08313478 0.16956800 0.17067045 0.13844621 0.11056301
## [67] 0.14665291 0.25650633 0.24357957 0.30900106 0.71759019 0.71563575
## [73] 0.48911870 0.57594573 0.55837573 0.54210138 0.59169983 0.48337951
## [79] 0.49112118 0.53264642 0.66650770 0.71113312 0.71211364 0.73905051
## [85] 0.72278606 0.82880733 0.89589122 0.85089259 0.86759201 0.86038403
## [91] 0.83601003 0.77394150 0.27638053 0.53914451 0.50288660 0.41789834
## [97] 0.38233647 0.35148243 0.52365390 0.69726660 0.81596612 0.74512524
## [103] 0.94180590 1.13042874 1.01517746 1.25817015 1.21089441 1.05943739
## [109] 1.16200514 1.07461448 0.68073042 0.60923031 0.54575803 0.69056673
## [115] 0.67187887 0.60074808 0.54595290 0.46385076 0.48148611 0.38653602
## [121] 0.24793204 0.28277498 0.18420868 0.16923096 0.15026325 0.14939445
## [127] 0.17323228 0.23878445 0.30069113 0.29320013 0.31210423 0.34182155
## [133] 0.57186305 1.01464614 0.98194395 1.07731686 1.02464993 1.07135584
## [139] 0.99732048 1.03257675 1.02564633 1.35298677 1.33364921 1.69856380
## [145] 1.68229643 1.70759727 1.67174272 1.43212727 0.97639679 0.89588246
## [151] 0.92118817 1.04542447 1.07495809 0.88227668 0.77332507 0.79431022
## [157] 0.34460093 0.32383871 0.59663490 0.65580694 0.69731607 0.66886025
## [163] 0.54402475 0.51911009 0.51316670 0.60696639 0.50724968 0.38413569
## [169] 0.35789948 0.47990985 0.63499468 0.76233731 0.44426269 0.39413723
## [175] 0.63249499 0.69288195 0.76949863 0.47328771 0.42750643 0.39479708
## [181] 0.33946630 0.31718777 0.29722599 0.26638302 0.27130230 0.22400789
## [187] 0.22240636 0.21488268 0.29816856 0.28236949 0.28213788 0.23192519
## [193] 0.34257781 0.43365531 0.49605384 0.35441602 0.35796749 0.51500839
## [199] 0.65722966 0.67276395 1.10943970 1.26494029 1.09354005 1.45503032
## [205] 1.93834872 1.79599780 1.72349303 1.86578453 1.56181180 1.64304399
## [211] 1.47567185 1.29201088 1.30572533 1.19959602 1.26997706 1.28550226
## [217] 1.44257524 1.14791984 1.37398478 1.32609067 1.42891003 1.33630557
## [223] 1.33786176 0.76328163 0.57354718 0.77472253 1.03262317 1.02113252
## [229] 1.06715379 0.99654573 0.94611287 0.98236017 1.03777682 0.92219172
class(nvda_df$VolScore)## [1] "numeric"
ggplot(nvda_df) +
aes(x = date) +
geom_line(aes(y = VolScore, colour = "Nvda VolScore"), size = 1) + # Line for Nvda VolScore
geom_line(aes(y = sector_realized_vol, colour = "Sector Realized"), size = 1) + # Line for Sector VolScore
labs(x = "Date", y = "Volscore", title = "Nvda vs Sector Realized") +
scale_color_manual(values = c("Nvda VolScore" = "#9941AC", "Sector Realized" = "#1C84C6")) + # Define colors
theme_minimal()
# Plot NVDA VolScore
plot(nvda_df$date, nvda_df$VolScore, type = "l", col = "red", main = "NVDA VolScore", xlab = "Date", ylab = "VolScore")
The trading strategy involves:
- Entering a short volatility (straddle) position when the VolScore exceeds a certain threshold (1.2).
- Exiting the position when the VolScore falls below a specified threshold (0.7) or after a fixed holding period.
# Initialize position tracking
# Trading parameters
entry_threshold <- 1.2
exit_threshold <- 0.7
holding_period <- 30 # in days
# Initialize position tracking
positions <- list()
cash <- 100000 # Starting cash
position_size <- 10000 # Trade size
entry_price <- 0
entry_date <- NA
pnl <- 0 # To track profit/loss
# Loop through NVDA data to simulate trades
for (i in 1:nrow(nvda_df)) {
# Entry condition
if (nvda_df$VolScore[i] > entry_threshold && is.na(entry_date)) {
entry_date <- nvda_df$date[i]
entry_price <- nvda_df$implied_volatility[i]
positions[[length(positions) + 1]] <- list(
type = "Entry",
date = entry_date,
price = entry_price,
VolScore = nvda_df$VolScore[i],
pnl = NA # No P/L for entry
)
}
# Exit condition
if (nvda_df$VolScore[i] < exit_threshold && !is.na(entry_date)) {
exit_date <- nvda_df$date[i]
exit_price <- nvda_df$implied_volatility[i]
pnl <- position_size * (entry_price - exit_price) # Calculate P/L
cash <- cash + pnl
positions[[length(positions) + 1]] <- list(
type = "Exit",
date = exit_date,
price = exit_price,
VolScore = nvda_df$VolScore[i],
pnl = pnl # Include P/L for exit
)
# Reset entry after exit
entry_date <- NA
entry_price <- 0
}
}We now plot the entries and exits on the same chart with VolScore and implied volatility to visualize the strategy.
# Ensure the relevant columns are numeric
nvda_df$VolScore <- as.numeric(nvda_df$VolScore)
nvda_df$realized_volatility <- as.numeric(nvda_df$realized_volatility)
positions_df <- do.call(rbind, lapply(positions, function(x) as.data.frame(x, stringsAsFactors = FALSE)))
# Ensure the positions dataframe also has numeric VolScore
positions_df$VolScore <- as.numeric(positions_df$VolScore)
# Plotting VolScore with realized volatility, sector volatility, and vertical lines for entry/exit points
ggplot(nvda_df) +
aes(x = date) +
geom_line(aes(y = VolScore, colour = "VolScore"), size = 0.5) + # Line for VolScore
# geom_line(aes(y = realized_volatility, colour = "Realized Volatility"), size = 0.5) + # Line for Realized Volatility
geom_line(aes(y = sector_realized_vol, colour = "Sector Volatility"), size = 0.5) + # Line for Sector Volatility
geom_line(aes(y = implied_volatility, colour = "IV"), size = 0.5) +
# Add vertical lines for Entry points
geom_vline(data = subset(positions_df, type == "Entry"), aes(xintercept = as.numeric(date)),
color = "green", size = 1, linetype = "dotted") +
# Add vertical lines for Exit points
geom_vline(data = subset(positions_df, type == "Exit"), aes(xintercept = as.numeric(date)),
color = "red", size = 1, linetype = "dotted") +
labs(x = "Date", y = "Volscore", title = "Nvda Volscore, Implived Volatility, and Sector Volatility with Entry/Exit Points") +
scale_color_manual(values = c("VolScore" = "#9941AC", "Realized Volatility" = "#FF5733", "Sector Volatility" = "#1C84C6", "IV" = "black")) + # Custom colors
theme_minimal()
positions_df## type date price VolScore pnl
## 1 Entry 2023-11-02 0.5246691 1.4540109 NA
## 2 Exit 2023-11-22 0.3503089 0.4516258 1743.602
## 3 Entry 2024-03-12 0.6294907 1.2581702 NA
## 4 Exit 2024-03-19 0.5004700 0.6807304 1290.207
## 5 Entry 2024-05-02 0.6229188 1.3529868 NA
## 6 Exit 2024-05-23 0.3948357 0.3446009 2280.832
## 7 Entry 2024-07-30 0.7206185 1.2649403 NA
## 8 Exit 2024-08-30 0.4633520 0.5735472 2572.665
This report demonstrates how to calculate a VolScore and develop a simple short-volatility trading strategy based on implied and realized volatilities. The strategy involves entering positions when the VolScore is high and exiting when the VolScore drops or after a set holding period. The backtest results show the P&L for each trade and the overall performance of the strategy.