Bitcoin Forum
December 11, 2017, 06:18:15 PM *
News: Latest stable version of Bitcoin Core: 0.15.1  [Torrent].
 
   Home   Help Search Donate Login Register  
Pages: [1]
  Print  
Author Topic: Modeling the Prices of Cryptocurrencies  (Read 1622 times)
bitinlet
Member
**
Offline Offline

Activity: 91


View Profile
February 19, 2014, 03:14:53 PM
 #1

Hi all,

I'm trying to create a model for cryptocurrency prices. I'd like to use "daily" data, to simply things. My conceptual idea is to look across some set of time (time-series) and across coins (cross-sectional), so I'd use panel data.

Let's break it down, for each coin there's a...
 

Supply
- this is pretty straight forward. I don't have this data in daily units (so a good data source would help here - got one?), but I know it's possible to obtain. Each coin has a supply out today, and each has a total supply (future). From this info, I can proxy supply. Right now, I know this info is available at http://coinmarketcap.com/ but unfortunately, the site does not store historical data. I need the data history.

Demand - this data isn't as straight forward, nor available. I have a few thoughts on proxies for this though. One thought is to use volume data (also available on crypto mkt cap). Another is to use "search" data (be it google, twitter, etc.). Perhaps a combination of each. Problem with the search data is: although I'm very confident it exists, I don't know where to obtain something like that. All I'd need is daily aggregates on search for each currency.

The goal, of course, will be to use the supply and demand to create a intrinsic price model for each cryptocurrency. Perhaps, even forcasting these. From that, one could then pick winners and losers via investment.

Assistance on any of the above is appreciated. I'd be happy to update the thread as I move forward.
1513016295
Hero Member
*
Offline Offline

Posts: 1513016295

View Profile Personal Message (Offline)

Ignore
1513016295
Reply with quote  #2

1513016295
Report to moderator
1513016295
Hero Member
*
Offline Offline

Posts: 1513016295

View Profile Personal Message (Offline)

Ignore
1513016295
Reply with quote  #2

1513016295
Report to moderator
1513016295
Hero Member
*
Offline Offline

Posts: 1513016295

View Profile Personal Message (Offline)

Ignore
1513016295
Reply with quote  #2

1513016295
Report to moderator
Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction. Advertise here.
QueenElizabeth
Jr. Member
*
Offline Offline

Activity: 56


View Profile
February 23, 2014, 05:29:45 PM
 #2

IMHO it's going to be very hard to forecast the price of cryptos based on aggregate daily information. I never paid attention to order books for cryptos other than Bitcoin, but in the case of Bitcoin, a lot of wild swings have happened within a single day, driven by the sequence of large orders. I think your best bet would be any methodology for high-frequency panel data where you analyze order books in real time.

bitinlet
Member
**
Offline Offline

Activity: 91


View Profile
February 24, 2014, 06:15:15 PM
 #3

IMHO it's going to be very hard to forecast the price of cryptos based on aggregate daily information. I never paid attention to order books for cryptos other than Bitcoin, but in the case of Bitcoin, a lot of wild swings have happened within a single day, driven by the sequence of large orders. I think your best bet would be any methodology for high-frequency panel data where you analyze order books in real time.

First, thanks for responding.

I understand your point, and it's a good one. I'm kinda upset because I was hoping for a better source, to just try this sort of thing. I do think it would be possible to model crypto prices based on certain data points using daily data, the unfortunate reality is - ALL of the useful data (other variables) is impossible to obtain. I mean, figuring out mining related info for every coin and creating a unique system to quantify each across the spectrum of all crypto coins is just too difficult.

So, to simplify, the variables I have are price, market cap, volume and supply. I'm gathering these from coinmarketcap. I do feel that one can predict price changes from changes in these variables. I'm already beginning the process of proving this. What good could this do? Well, knowing the potential magnitude could be very useful. In fact, the regression coefficients could be used to predict price-changes. That said, I agree that ideally high-frequency would be preferred. My issue is I'm a statistician, not a programmer. So, obtaining this high-frequency data, to analyze in real time, isn't really possible for me.

For my purposes, it needs to be averaged data, unfortunately. Any ideas on this that could help, let me know. And thanks again!
Calhil
Full Member
***
Offline Offline

Activity: 182


View Profile
February 24, 2014, 08:04:48 PM
 #4

IMHO it's going to be very hard to forecast the price of cryptos based on aggregate daily information. I never paid attention to order books for cryptos other than Bitcoin, but in the case of Bitcoin, a lot of wild swings have happened within a single day, driven by the sequence of large orders. I think your best bet would be any methodology for high-frequency panel data where you analyze order books in real time.

First, thanks for responding.

I understand your point, and it's a good one. I'm kinda upset because I was hoping for a better source, to just try this sort of thing. I do think it would be possible to model crypto prices based on certain data points using daily data, the unfortunate reality is - ALL of the useful data (other variables) is impossible to obtain. I mean, figuring out mining related info for every coin and creating a unique system to quantify each across the spectrum of all crypto coins is just too difficult.

So, to simplify, the variables I have are price, market cap, volume and supply. I'm gathering these from coinmarketcap. I do feel that one can predict price changes from changes in these variables. I'm already beginning the process of proving this. What good could this do? Well, knowing the potential magnitude could be very useful. In fact, the regression coefficients could be used to predict price-changes. That said, I agree that ideally high-frequency would be preferred. My issue is I'm a statistician, not a programmer. So, obtaining this high-frequency data, to analyze in real time, isn't really possible for me.

For my purposes, it needs to be averaged data, unfortunately. Any ideas on this that could help, let me know. And thanks again!
Ideally every bitcoin based cryptocurrency has a fixed number of new coins being mined every day. You can easily calculate this just by knowing block reward and block time.
As for the information about price and volume there are sites that collect this data. If they dont help you, you can write a simple program using api privided by the exchanges and start collecting data yourself. If you cant code yourself, there are complete applications on the internet that do this.

BLC & SKC node: addnode=192.3.171.213
BLC: BcaLHiLk74XXSZdebHQY8b3CaoEBLaPtoV
QueenElizabeth
Jr. Member
*
Offline Offline

Activity: 56


View Profile
February 25, 2014, 08:44:03 PM
 #5

I'm thinking ABM with heterogeneous agents... hoarders, traders etc.; you might run a toy model in StarLogo or any other such easy prototyping tool and if it looks promising you could try a bigger one with NetLogo and eventually go all out. But yeah, ABM or nothing, IMHO. This is not something that can be predicted by GARCH or whatever-traditional, in the short run.

Or, on account of the huge number of data points, you could run a two-variable Nadaraya-Watson with something likely (market sentiment as measured by some aggregate indicator?) and something unlikely (price of gold?) and see what happens.

I'm not even sure this suggestion makes any sense, it still looks too idiosyncratic to work, i.e. your ABM wouldn't have sensible behavioral rules no matter how hard you tried. And now of course you whetted my appetite for excessively nerdy stuff.


genco
Newbie
*
Offline Offline

Activity: 22


View Profile
February 25, 2014, 10:23:43 PM
 #6

Seems easy to me...

Take % btc supply that is traded (not hoarded; also called M1 by economists) and multiply by velocity (assume 6 for the Sake of discussion) which gives you annual btc transaction volume. Then divide the annual transaction volume of fiat using btc to solve for the correct bid rate.

Finally, chart this for the current year and future. Bid rates today assume your projection comes true.

I assume that in 10 years 75% btc will be M1, velocity is 6, fiat trade volume is $100b; so price in 10 yrs is $1,120/ btc. Thus a $500 btc today returns 8% annually (if held for 10 years).

So basically $500/ btc today is right price...any higher would yield <8% return, so you might as well invest in something else.
bitinlet
Member
**
Offline Offline

Activity: 91


View Profile
February 26, 2014, 12:12:13 AM
 #7

Thanks for the replies, Calhil, Queen and genco. All good points/suggestions. I suppose it would have been useful for me to describe what I was hoping to accomplish from this. My goal was to monitor the overall crypto-market. You can take a quick look at this market here: http://coinmarketcap.com/

Basically, my goal was taking each of these 100 currencies (which is cross-sectional data) and monitoring their behavior over time (longitudinal or time-series data). Combined, we have ourselves a nice panel data set (both cross-sectional and time-series) across cryptocurrencies.

I'd prefer to have a multitude of variables. But, the one's at my disposal are: Price, 24 Hour Volume, and Supply. I can obviously create change variables (such as change in price, change in volume and change in supply).

So, take the historical panel data (over time, across the 100 crypto-currencies), create a model based on those variables (maybe change in price is a function of change in volume and change in supply) and obtain regression results. After we get the results, we can then use those coefficients to forecast each crypto-currency's daily change.

This could be used to pick and choose where you want to place your btc. Maybe today, the model tells me megacoin should yield the best price change. Tomorrow it says doge.

That sort of thing. THanks again for you comments and I'm still interested in your feedback now that I've filled in a bit more info....
Ericgw
Jr. Member
*
Offline Offline

Activity: 34


View Profile
October 07, 2017, 08:16:14 PM
 #8

Thanks for the replies, Calhil, Queen and genco. All good points/suggestions. I suppose it would have been useful for me to describe what I was hoping to accomplish from this. My goal was to monitor the overall crypto-market. You can take a quick look at this market here: http://coinmarketcap.com/

Basically, my goal was taking each of these 100 currencies (which is cross-sectional data) and monitoring their behavior over time (longitudinal or time-series data). Combined, we have ourselves a nice panel data set (both cross-sectional and time-series) across cryptocurrencies.

I'd prefer to have a multitude of variables. But, the one's at my disposal are: Price, 24 Hour Volume, and Supply. I can obviously create change variables (such as change in price, change in volume and change in supply).

So, take the historical panel data (over time, across the 100 crypto-currencies), create a model based on those variables (maybe change in price is a function of change in volume and change in supply) and obtain regression results. After we get the results, we can then use those coefficients to forecast each crypto-currency's daily change.

This could be used to pick and choose where you want to place your btc. Maybe today, the model tells me megacoin should yield the best price change. Tomorrow it says doge.

That sort of thing. THanks again for you comments and I'm still interested in your feedback now that I've filled in a bit more info....

Been some time since the last post on this topic.  Have you continued the project?
Pages: [1]
  Print  
 
Jump to:  

Sponsored by , a Bitcoin-accepting VPN.
Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!