VWAP

Large institutional buyers use the VWAP ratio to be able to move into FX and stocks in a way that will not disturb the natural market dynamics of price. If these buyers were to move into a position all at once, it would unnaturally elevate the instrument price. By buying under the intraday VWAP moving average, these buyers can move into a position without too much price disruption.

⊕The volume weighted average price (VWAP) is a trading benchmark. VWAP is calculated by adding up the dollars traded for every transaction (price multiplied by number of shares traded) and then dividing by the total shares traded for the day. \[VWAP = \sum_{i=1}^n \frac{NumShares x Price_i}{Total Shares}\]

To manage the large file sizes generated by Forex exchanges it is useful to save tick data files in the more efficient binary (internal) data formats. This enables highes throughput of data, and faster input/output disk operations.

The easiest way to access Bitcoin transaction data is from the cache of zipped .csv files at http://api.bitcoincharts.com/v1/csv/. I downloaded the data and then saved it in binary format using the save() function. Now 25 million rows of raw tick data can be loaded into R memory using the load() function. Subsequently I process the tick data to convert the Unix timestamp into a date.

load("/home/hduser/zdata/tick_raw_file.RData")
head(tick_raw,5)

##     unixtime price  amount       date
## 1 1315922016  5.80  1.0000 2011-09-13
## 2 1315922024  5.83  3.0000 2011-09-13
## 3 1315922029  5.90  1.0000 2011-09-13
## 4 1315922034  6.00 20.0000 2011-09-13
## 5 1315924373  5.95 12.4521 2011-09-13

knitr::kable(tick_raw[1:6,], caption="Subset of raw tick data")

Subset of raw tick data

unixtime	price	amount	date
1315922016	5.80	1.0000	2011-09-13
1315922024	5.83	3.0000	2011-09-13
1315922029	5.90	1.0000	2011-09-13
1315922034	6.00	20.0000	2011-09-13
1315924373	5.95	12.4521	2011-09-13
1315924504	5.88	7.4580	2011-09-13

The above chunk of code loaded the bitstamp transaction data in US dollars, reads 25 million rows of raw tick data into a data frame.

library(dplyr)

## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

library(xts)

## Loading required package: zoo

## 
## Attaching package: 'zoo'

## The following objects are masked from 'package:base':
## 
##     as.Date, as.Date.numeric

## 
## Attaching package: 'xts'

## The following objects are masked from 'package:dplyr':
## 
##     first, last

library(dygraphs)

Now for the payoff, we use the standard dplyr functions and the efficient xts() timeseries function to aggregate the transactions into a time series and the dygraphs library to visualize the results.

data1 <- select(tick_raw, -unixtime)
data1 <- mutate(data1, value = price * amount)
by_date <- group_by(data1, date)
daily <- summarise(by_date, count = n(),
                   m_price  <-  mean(price, na.rm = TRUE),
                   m_amount <-  mean(amount, na.rm = TRUE),
                   m_value  <-  mean(value, na.rm = TRUE))
 
names(daily) <- c("date","count","m_value","m_price","m_amount")

Daily Table

Now we have the daily data frame with summaries of price, amount and value, we can now convert this to an R xts timeseries eXtended Time Series, which is required for input to the dygraphs routine. xts also provides faster timeseries processing for subsequent analyses.

load("/home/hduser/zdata/daily_file.RData")
knitr::kable(daily[1:6,])

date	count	m_value	m_price	m_amount
2011-09-13	12	5.874167	4.864282	28.84145
2011-09-14	14	5.582143	4.367570	24.41820
2011-09-15	6	5.120000	13.356799	68.04317
2011-09-16	4	4.835000	9.978502	48.44079
2011-09-17	1	4.870000	0.300000	1.46100
2011-09-18	8	4.840000	14.976600	72.48039

daily_ts <- xts(daily$m_value,order.by=daily$date)

The easiest way to access Bitcoin transaction data is from the cache of zipped .csv files at http://api.bitcoincharts.com/v1/csv/. I downloaded the data and then saved it in binary format using the save() function. Now 25 million rows of raw tick data can be loaded into R memory using the load() function. Subsequently I process the tick data to convert the Unix timestamp into a date.

Graph with dygraphs package

# Plot with htmlwidget dygraph
dygraph(daily_ts,ylab="Price (US Dollars)", 
        main="Average Value of bitcoin US$ Buys \n Kraken Exchange") %>%
  dySeries("V1",label="Avg_Buy") %>%
  dyRangeSelector(dateWindow = c("2016-05-01","2018-04-30"))