Introduction

The idea of this post is to give a high level overview of what’s needed to develop a stock screener with R. Even though there are websites that provide an interface for this even for free in some cases it is useful to develop your own custom screener. Assuming you know what you are looking for, it’s not a complex process.

The main ingredients are:

1. Collect the relevant data
2. Use a criteria to rank stocks

In this example I’ll demonstrate how to screen stocks to implement part of the strategy that’s described in Laurence bensdorp’s book, “The 30 minute trader”. In the book it’s described as “Mean reversion long”, there is only one parameter I removed for simplicity so in general terms this should be fairly close to what’s in the book. I’m not affiliated to him, but I thought the book was simply excellent and will recommend anyone to read it.

1. Collecting price data

Collecting price data is simple in R. In order to keep this post as a demo I’ll use the Nasdaq-100 stocks but you could collect prices of more stocks. For example the Russel-1000 or S&P 500 constituents. In this example I’ll use quantmod that provides a simple way to access to yahoo-finance data and data.table to aggregate the result of the screen and query it.

# Load libraries
library(data.table)
library(magrittr)
library(quantmod)

options("getSymbols.warning4.0"=FALSE)
options("getSymbols.yahoo.warning"=FALSE)

# Read data from ishares site to get stock tickers

# get stock tickers
tickers = ticker_lkp[nchar(ISIN) == 12, Issuer Ticker]

I’ll collect the price data in a list in order to be able to access each ticker by it’s name

# Get prices as xts
data = list()
for (i in seq_along(tickers)) {
Sys.sleep(0.005) # not to get rate limited
ticker = tickers[i]
from = "2015-01-01"
to = as.character(Sys.Date())
data[[ticker]] = try(getSymbols(ticker, from=from, to=to, auto.assign = FALSE, src='yahoo'), silent=TRUE)
}
names(data) = tickers

First I’ll just check if all the stocks were downloaded. It looks like we have 100 stocks plus 2 that have A shares. All objects are of class xts so it seems the data was collected properly.

# Remove empty data
table(sapply(data, function(x) class(x)[1]))
##
## xts
## 102

Below I show a few examples of how to access the price and volume data of each stock.

tail(data[['MSFT']], 3)
##            MSFT.Open MSFT.High MSFT.Low MSFT.Close MSFT.Volume MSFT.Adjusted
## 2021-01-20    217.70    225.79   217.29     224.34    37777300        224.34
## 2021-01-21    224.70    226.30   222.42     224.97    30749600        224.97
## 2021-01-22    227.08    230.07   225.80     225.95    30124900        225.95
tail(data[['COST']], 3)
##            COST.Open COST.High COST.Low COST.Close COST.Volume COST.Adjusted
## 2021-01-20    354.39    361.90   353.41      361.3     2767200         361.3
## 2021-01-21    361.30    363.99   359.94      362.8     2122600         362.8
## 2021-01-22    363.20    364.63   359.85      362.3     1959100         362.3

2. Develop the screener

In this case I’ll show how to find stocks in a long term up-trend that seem have some short term weakness.

First of all I’ll adjust the OHLC values.

data = lapply(data, adjustOHLC, use.Adjusted=TRUE)

In order to make the code reproducible I’ll keep the data up to today. You can remove this like if you want to get the most recent results

data = lapply(data, function(d) d['/20210115'])

Screener parameters

• Long term uptrend: Close > SMA-150 (long term uptrend)
• High volatility: ATR t10 > 4% (high volatility stocks)
• Short term oversold: RSI t3 < 30
• Ranked by lower RSI-3

Some definitions: - ATR: Average true range - RSI: Relative strenght indicator - SMA-150: Simple moving average of the past 150 days

Of course this is an example, you can try different parameters.

Here is the main function I used, the code is commented.

mean_reversion <- function(data, ticker){

df = data[[ticker]]

# Compute indicators
HLC = HLC(df)
sma150 = SMA(Cl(df), n=150)
atr10 = ATR(HLC, n=10, maType='EMA')
rsi3 = RSI(Cl(df), n=3)

# join them to the main data
df = merge(df, sma150)
df = merge(df, atr10$atr) df = merge(df, rsi3) df$atr = df$atr / Cl(df) df$close = Cl(df)

# Select relevant columns
keep_cols = c("SMA", "atr", "rsi", "close")
df = df[, keep_cols]
names(df)[1:3] = c("sma150", "atr10", "rsi3")

# Get the most recent observation
df = tail(df, 1)

# Convert the output as a data table
dt = as.data.table(df)
dt[, ticker:=ticker]
setnames(dt, "index", "date")

dt
}

Execute the screener and evaluate results

indicators = list()
for(ticker in tickers){
indicators[[ticker]] = mean_reversion(data, ticker)
}

# Rbind the indicators list
indicators = rbindlist(indicators)
head(indicators)
##          date    sma150      atr10     rsi3   close ticker
## 1: 2021-01-15  113.4679 0.02581628 27.30745  127.14   AAPL
## 2: 2021-01-15  210.9870 0.01823743 24.45446  212.65   MSFT
## 3: 2021-01-15 3132.2195 0.02003178 29.26385 3104.25   AMZN
## 4: 2021-01-15  439.5304 0.05079055 43.58718  826.16   TSLA
## 5: 2021-01-15  262.9280 0.03270333 43.37887  251.36     FB
## 6: 2021-01-15 1602.1517 0.02265435 28.77536 1736.19   GOOG

Keep stocks with close higher than the SMA-150

1. Long-Term Trend

indicators = indicators[close >= sma150]
nrow(indicators)
## [1] 87

we get 87 stocks just with this filter.

2. ATR-10 > 4%

indicators = indicators[atr10 >= 0.04]
nrow(indicators)
## [1] 15

After the ATR(10) filter we get 15 stocks

3. RSI-3 < 30

indicators = indicators[rsi3 <= 30]
nrow(indicators)
## [1] 3

4. Rank by RSI-3

This last part makes sense if you include more stocks to the example using an index such as the Russel-1000 or S&P 500. But still the code is the same. I’ll leave the ranking part to make the code complete.

indicators[, rk:=frank(rsi3)]
indicators[order(rk)]
##          date    sma150      atr10     rsi3  close ticker rk
## 1: 2021-01-15 116.08595 0.04108247 14.94451 136.60   XLNX  1
## 2: 2021-01-15  78.94353 0.04119751 15.14981  88.21    AMD  2
## 3: 2021-01-15 105.78900 0.06850013 21.49396 161.20    PDD  3

Example plot of one stock that was ranked in the top-10

At the time I wrote the post PDD was a good example of what the strategy tries to find. It’s clear that the stock is in a long term uptrend and recently had a pullback.

df = data[['PDD']]["2020/"]
{plot(Cl(df))
lines(SMA(Cl(df), n = 20), col="blue")
lines(SMA(Cl(df), n = 50), col="red", lty=2)
# add legend to panel 1
addLegend("topleft", legend.names = c("Close", "SMA(20)", "SMA(50)"),
lty=c(1, 1, 2), lwd=c(2, 1, 1),
col=c("black", "blue", "red"))}

I hope you enjoyed the post and found it useful!