## Market Analysis

# Clearing the workspace
rm(list=ls(all=TRUE)) # Removes all variables from current workspace if any
gc(reset=TRUE)        # Collects garbage
set.seed(12345)       # Initializes random genarator [actually, not needed here, but it's good style]

# Loading libraries
# If you get error messages here run 'install.packages("package.name")' berofe loading library
# i.e 'install.packages("plyr")' then 'library(plyr)'
library(plyr)      # Data manipulation
library(reshape2)  # Data manipulation
library(lubridate) # Working with dates

# Reading the data
# In .txt file values are separated by tabs (sep='\t'), first row is header
prices <- read.table("data/ADSK_D.txt", sep="\t", header=T)
prices$Date <- mdy(prices$Date) # It's good style to have all dates in YYYY-MM-DD format
head(prices)                    # Demonstrates first rows of the data (just for information)

## Imputing new columns to the data

# Up/Down: 'c(..)' creates array of values. In this case first value is 'NA' (Not Available - it means missing data)
# Then goes logical value, indicating price change
# Feel free to run 'head(prices)' to inspect results at any stage
prices$updown <- c(NA, prices$Close[-1] > prices$Close[-nrow(prices)])

# 10 Day Sum
# Initially we fill column with NAs
prices$sum.10 <- NA
# Then, starting from 11th row we calculate sum of 'UpDown's in previous 10 rows
prices$sum.10[11:nrow(prices)] <- laply(11:nrow(prices), function(i) sum(prices$updown[(i-9):i], na.rm=T))

# Trigger 9 of 10: logical value, it's simple
prices$trigger.9 <- prices$sum.10>=9

# Managing overlaps
# If 'overlap' is zero or less than zero then we don't care about overlaps (first .xlsx file)
# Otherwise, this variable holds minimum number of periods between signals (second .xlsx file)
overlap <- 5

# We run this loop for 'overlap' higher than zero
if(overlap>0){
  # In the loop we screen all overlapped triggers
  for(i in 11:nrow(prices)){
    if(prices$trigger.9[i] & sum(prices$trigger.9[(i-overlap):(i-1)])>0)
      prices$trigger.9[i] <- FALSE  
  }
}

# Creates filter to select only triggered days
# Function 'which()' returns indices for which logical expression is TRUE
# In this case 'filter' holds triggered rows' numbers 
filter <- which(prices$trigger.9==T & !is.na(prices$trigger.9))

# 5-,10- and 15-days returns: initially filled with NAs
prices$ret.15 <- prices$ret.10 <- prices$ret.5 <- NA
# Calculates returns for triggered rows for 5-, 10- and 15-days
# 'laply(filter, function(i) ...)' means we run function with one parameter i
# for each element in 'filter' (i.e. triggered row number).
# Inside function we calculate returns. Also we additionally check if there enough rows ahead
prices[filter,]$ret.5 <- 
  laply(filter, function(i) ifelse((i+5)<=nrow(prices), prices[i+5,]$Close/prices[i,]$Close-1, NA))
prices[filter,]$ret.10 <- 
  laply(filter, function(i) ifelse((i+10)<=nrow(prices), prices[i+10,]$Close/prices[i,]$Close-1, NA))
prices[filter,]$ret.15 <- 
  laply(filter, function(i) ifelse((i+15)<=nrow(prices), prices[i+15,]$Close/prices[i,]$Close-1, NA))

# Constructs final table with occurences
# Here we just filter 'prices' with created 'filter' and columns' names
occurences <- prices[filter, c("Date","Close","ret.5","ret.10","ret.15")]
occurences

# This long command creates summary table
# You are free to add rows with specified statistics here
final.summary <- 
summarise(occurences,
          average.ret.5 = mean(ret.5),
          average.ret.10 = mean(ret.10),
          average.ret.15 = mean(ret.15),
          stdev.ret.5 = sd(ret.5),
          stdev.ret.10 = sd(ret.10),
          stdev.ret.15 = sd(ret.15),
          max.gain.5 = max(ret.5),
          max.loss.5 = min(ret.5),
          max.gain.10 = max(ret.10),
          max.loss.10 = min(ret.10),
          max.gain.15 = max(ret.15),
          max.loss.15 = min(ret.15),
          days.up.5 = sum(ret.5>0)/length(ret.5),
          days.up.10 = sum(ret.10>0)/length(ret.10),
          days.up.15 = sum(ret.15>0)/length(ret.15))

# Transposing 'final.summary' for presentation
final.summary <- melt(final.summary)
final.summary

# Saving results to files
# For each 'overlap' separate file is created
write.table(occurences, paste0("output/occurences", ifelse(overlap>0, overlap, ""), ".csv"), row.names=F, sep=",")
write.table(final.summary, paste0("output/summary", ifelse(overlap>0, overlap, ""), ".csv"), row.names=F, sep=",")