Quick Analysis of Ethereum and Litecoin Trends Using Prophet

Mon, Feb 8, 2021 3-minute read 555 words

In this post, I use Facebook’s Prophet algorithm to detect potential patterns in daily prices of Litecoin and Ethereum, and build forecasts for these two crypto currencies for the rest of 2021.

I am considering selling my positions in Litecoin (LTC) at the end of this month (February). The gains would be used to beef my positions in Ethereum (ETH) until selling them at the end of August. The analysis described in this post is one of the several reasons behind this move. Currently, I am testing another model to validate these recommendations.

For this analysis, I use official Prophet R package, and historical prices data for LTC and ETC provided by Coinmarketcap. You can learn more about Prophet and its use for forecasting on the project’s official site. For the sake of this presentation, I assume that the Coinmarketcap prices are trustworthy and reliable. However, based on my experience, there are often substantial differences in prices reported by different sources/platforms. I will talk more about that in another post.

The LTC prices were from Aug 2016 to February 7, 2021. To “tighten” the data and bring the distribution closer to the normal, I use Box-Cox transformation with automatic detection of optimal power parameter:

library(tidyquant)
library(tidyverse)
library(lubridate)
library(forecast)
library(prophet)
library(ggplot2)
library(broom)
library(caret)

# prices are downloaded from Coinmarketmap

LTC<-read.csv("./ltcprice.csv") %>% 
  mutate(ds=as.Date(ds)) %>% 
  mutate(y=round(y,digits=1))

ltclam = BoxCox.lambda(LTC$y, method = "loglik")
LTC$y = BoxCox(LTC$y, ltclam)

Once the data is transformed, I fit the Prophet model which also automatically detects monthly and weekly seasonality:

LTCmod<-prophet(LTC)
prophet:::plot_yearly(LTCmod)+ggtitle("LTC Monthly Seasonality")
prophet:::plot_weekly(LTCmod)+ggtitle("LTC Weekly Seasonality")

According to the first chart, Litecoin tends to go up from mid-January till the end of February, and then it dips. The second upward run starts around April 1 and ends around the first week of June. It then starts steadily rising again in early December:

Litecoin Monthly Seasonality

The weekly seasonality chart also suggests that LTC price tends to dip on Thursdays and spike on Saturdays:

Litecoin Weekly Seasonality

Now, we can fit this model to forecast LTC prices for the rest of 2021:

year2021 <- make_future_dataframe(LTCmod, periods = 320)
LTC2021<-predict(LTCmod,year2021) 
plot(LTCmod,LTC2021)+ggtitle("LTC 2021 Forecast")

Litecoin Forecast 2021

The forecast confirms that the current upward trend will continue until the end of February. You can again purchase Litecoin in the beginning of April, but the price won’t be going up for long, and you can expect a lot of volatility until December.

Perhaps, Prophet will paint a brighter picture for Ethereum? Let’s see!

First, let’s look at the monthly seasonality:

Ethereum Monthly Seasonality

The short upward trend in the beginning of the year is similar to LTC. However, the April-August upward trend looks more promising.

Ethereum Weekly Seasonality

The intra-week trend is pretty much the same as with LTC: Ethereum price tend to drop on Thursdays, and rise on Saturdays.

Finally, the 2021 forecast looks much more interesting:

Ethereum Weekly Seasonality

As you can see, ETH has a much larger potential for gains in 2021, compared to LTC. I hope to validate these results by using another model and set of data from another source, before the end of this month. Stay tuned!

Disclosure:I am not a professional financial adviser, and all information on this blog should not be construed as professional financial advice. All evaluations and analysis in this post are solely for informational purposes and not for trading purposes or advice.