DownFriday-DownMonday.R

#Script to assess effect size of "down Friday, down Monday" indicator

library(tidyverse)
# tidy financial analysis
library(tidyquant)
# tidy data cleaning functions

#Define function to get y years of daily close price data for SPY from Yahoo Finance
#And then, using this data, run a "robustness_check" function to be defined below
sample_sizes <- function(y){
get_prices <- tq_get("SPY",from=(Sys.Date()-1)-365*y,to=(Sys.Date()-1)) %>%
  mutate(weekday = weekdays(date)) 
#Create change-from-previous-close variable and select columns we want
get_prices <- get_prices %>%
  mutate(change = c(NA,adjusted[-1] - adjusted[-nrow(get_prices)])) %>%
  select(symbol, date, adjusted, weekday, change)
#Create data frame of day-pairs with logical variable for DF_DM indicator
day_pairs <- get_prices %>%
  mutate(date2 = c(date[-1],NA), adjusted2 = c(adjusted[-1],NA), weekday2 = c(weekday[-1],NA), change2 = c(change[-1],NA)) %>%
  filter(!is.na(change) & !is.na(change2)) %>%
  mutate(DF_DM = weekday == "Friday" & weekday2 == "Monday" & change < 0 & change2 < 0)

#Define function to calculate mean momentum change, n days pre vs. n days post each day-pair
robustness_check <- function(n,day_pairs){
 #Create variables for n-day change preceding and succeeding the day-pair
 day_pairs <- day_pairs %>%
   mutate(pre = c(rep(NA,times=n),(adjusted[-c(1:(n-1),nrow(day_pairs))] - adjusted[-c((nrow(day_pairs)-(n-1)):nrow(day_pairs))])/adjusted[-c((nrow(day_pairs)-(n-1)):nrow(day_pairs))]),post = c((adjusted2[-c(1:n)] - adjusted2[-c(1,(nrow(day_pairs)-(n-2)):nrow(day_pairs))])/adjusted2[-c(1,(nrow(day_pairs)-(n-2)):nrow(day_pairs))],rep(NA,times=n))) %>%
   filter(!is.na(pre) & !is.na(post))
 #Create variable for pre-post change of direction for all day pairs
 day_pairs <- day_pairs %>%
   mutate(mom_change = abs(post - pre))
 #Summarize average change of direction for DF_DM and non-DF_DM day pairs
 day_pairs %>%
   group_by(DF_DM) %>%
   summarize(n,mean = mean(mom_change))
}

#Create sequence of different values of n and run robustness check function
robustness_results <- map_dfr(.x = seq(from = 3,to = 30,by = 5),.f = robustness_check,day_pairs = day_pairs)

#For each value of n, subtract DF_DM mean momentum change from non-DF_DM
results_summary <- robustness_results %>%
  mutate(sample_size = y) %>%
  group_by(sample_size,n) %>%
  summarize(effect_size_pct_index = (mean[2] - mean[1]),effect_size_pct_inflection = (mean[2] - mean[1])/mean[1])
}

#Create sequence of different values of y and run sample_size function
robustness_results <- map_dfr(.x = seq(from = 5,to = 20,by = 5),.f = sample_sizes)

#Print results table
robustness_results

#Plot DF+DM "excess inflection" as percent of index price
robustness_results %>%
  ggplot(aes(x = n,y = effect_size_pct_index)) +
  geom_line(aes(color=as.factor(sample_size)))

#Plot DF+DM "excess inflection" as percent of average non-DF+DM inflection
robustness_results %>%
  ggplot(aes(x = n,y = effect_size_pct_inflection)) +
  geom_line(aes(color=relevel(as.factor(paste("Last",sample_size,"years",sep=" ")),"Last 5 years"))) +
  labs(title = "Measuring the inflection effect of the 'Down Friday, Down Monday' stock market signal",
       x = "Number of days after DF+DM signal",
       y = "Excess DF+DM inflection as proportion of average inflection at all other day-pairs",
       color = "Sample size") +
  theme_classic()

#Export chart as image file
ggsave(
  filename = "DFDM.jpg",
  plot = last_plot(),
  path = "", #Insert your preferred file path here
  scale = 1,
  width = 1920/175,
  height = 1080/175,
  units = "in"
)