Advertisement
Not a member of Pastebin yet?
Sign Up,
it unlocks many cool features!
- #python 2.7 ; matplotlib
- All active altcoins were analyzed at poloniex.com; past 3 years of 2h data
- 2 hour simple moving averages of period 5 through 1000, stepping by 5, were drawn
- Each moving average was colored in rainbow format; slowest on one end of spectrum; fastest on other
- Rainbows were then sliced into 2 hour chunks and indexed by coin name and unix stamp
- Several hundred thousand such sliced png images 2 pixels by 100 pixels were created occupying several gigs of space
- Approximately 50 days of quad core desktop cpu time were used to create the rainbow slices
- An additional 2x10 pixels of grey scale were added to each image relative to the breadth of the rainbow moving average mesh
- The same process was repeated for BTC/USDT
- The current slice 2x110 pixels of rainbow for BTC/USDT was compared using computer vision to each of the quarter million altcoin slices back in time
- The best matches were sorted by highest correlation coeff
- The most recent 120 days of BTC/USDT data is then plotted in white
- In yellow, best matching historical altcoin data from match date to 60 days of projection was plotted
- The intensity of the yellow represents the correlation of each historic altcoin to the past 1000 2h candles of BTC/USDT data
- Each altcoin was scaled to btc price scale by dividing its beginning price by BTC/USDT last price then multiplying the array by the ratio
- The result is a 60 day ensemble model projection for BTC/USDT based on 3 year historical performance of poloniex altcoins
- litepresence
- April 2017
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement