fuksi/crypto-analysis

Crypto-analysis

Provide aggregated information for these events:

A currency wins for each interval. Winning = highest change in pct.
There is no trade for x currencies in during an interval

How to run

Firstly, restore packages with Pipenv

The program has 2 phases

Aggregating the data
Run analysis on the data

To aggregate the data

You need a postgres instance
Fetch data using this https://github.com/fuksi/bfxdata
- branch: tradings
- flags: --includetradings
- make sure you have correct connection string in db.py
Aggregate using main.py in this project

The analysis is done with jupyter notebook

notebook file Crytocurrencies analysis.ipynb
just re-execute all cells

Analysis

Objective

Show statistical analysis on

Random variable 'winning', where X='BTC' is the event where BTC price increase the most in pct compared to other ccys during in interval
Random varible 'no trade', where X='BTC' is the event where BTC was not traded during an interval

Approach

Just explore data by counting for now

Dependencies

# Tech
import os
import sys
import pandas as pd
import numpy as np
import pendulum
import matplotlib
import matplotlib.pyplot as plt
import datetime

from src.models import DeltaPoint, Trade    
from src.computed import get_pricepoints_winner_count, get_all_result

intervals = ['15', '20', '30', '60', '120', '300']

Some info about data points

# Get data
all_result = get_all_result()
timestamps = [pendulum.from_timestamp(ts) for ts in all_result[0]]
data_points = all_result[1]
for data_point in data_points:
    for interval in intervals:
        mm = data_point[interval]['movements']
        data_point[interval]['movements'] = np.array(mm) / sum(mm)

def show_eval_period(timestamps):
    print(f'Evaluation period: from {timestamps[0].to_formatted_date_string()} to {timestamps[-1].to_formatted_date_string()}')
    print(f'Number of days: {timestamps[-1].diff(timestamps[0]).in_days()}')
          
show_eval_period(timestamps)

Evaluation period: from Jul 02, 2017 to Jan 02, 2019
Number of days: 549

# Settings
periods = [7, 30, 183, 365]
periods.append(len(timestamps))
periods_text = ['Week', 'Month', 'Half-year', 'Year', 'All']
winners = ['BTCUSD', 'ETHUSD', 'LTCUSD', 'XRPUSD', 'EOSUSD', 'NONE']

Winning distribution

def compare_winning_periods(interval):
    winners_dict = {}
    for w in winners:
        winners_dict[w] = []
        for p in periods:
            count = 0
            for i in range(p):
                count += data_points[i][str(interval)]['winners'][w]
                
            avg = count / p
                
            winners_dict[w].append(avg)
                
    df = pd.DataFrame(winners_dict, index=periods_text).round(2)
    
    return df

15s interval

df = compare_winning_periods(15)
df.head()

	BTCUSD	ETHUSD	LTCUSD	XRPUSD	EOSUSD	NONE
Week	0.40	0.26	0.12	0.03	0.13	0.06
Month	0.41	0.27	0.10	0.04	0.13	0.05
Half-year	0.44	0.24	0.12	0.06	0.11	0.02
Year	0.43	0.22	0.12	0.08	0.12	0.02
All	0.42	0.22	0.11	0.09	0.12	0.05

20s interval

df = compare_winning_periods(20)
df.head()

	BTCUSD	ETHUSD	LTCUSD	XRPUSD	EOSUSD	NONE
Week	0.41	0.26	0.13	0.04	0.13	0.03
Month	0.42	0.26	0.11	0.05	0.13	0.03
Half-year	0.45	0.24	0.13	0.07	0.11	0.01
Year	0.44	0.22	0.12	0.09	0.12	0.01
All	0.43	0.22	0.11	0.10	0.12	0.03

30s interval

df = compare_winning_periods(30)
df.head()

	BTCUSD	ETHUSD	LTCUSD	XRPUSD	EOSUSD	NONE
Week	0.42	0.25	0.13	0.05	0.13	0.01
Month	0.43	0.25	0.11	0.06	0.13	0.01
Half-year	0.46	0.22	0.13	0.08	0.11	0.00
Year	0.45	0.20	0.13	0.10	0.11	0.01
All	0.44	0.21	0.12	0.11	0.11	0.01

60s interval

df = compare_winning_periods(60)
df.head()

	BTCUSD	ETHUSD	LTCUSD	XRPUSD	EOSUSD
Week	0.43	0.24	0.14	0.07	0.12
Month	0.44	0.22	0.12	0.09	0.12
Half-year	0.47	0.19	0.13	0.10	0.11
Year	0.46	0.18	0.12	0.13	0.10
All	0.45	0.18	0.12	0.14	0.10

2 mins interval

df = compare_winning_periods(120)
df.head()

	BTCUSD	ETHUSD	LTCUSD	XRPUSD	EOSUSD
Week	0.45	0.21	0.14	0.09	0.11
Month	0.46	0.19	0.12	0.12	0.11
Half-year	0.49	0.17	0.12	0.12	0.10
Year	0.47	0.16	0.11	0.16	0.09
All	0.46	0.16	0.12	0.17	0.09

5 mins interval

df = compare_winning_periods(300)
df.head()

	BTCUSD	ETHUSD	LTCUSD	XRPUSD	EOSUSD
Week	0.47	0.19	0.13	0.10	0.11
Month	0.47	0.16	0.11	0.16	0.10
Half-year	0.50	0.14	0.11	0.15	0.10
Year	0.49	0.13	0.10	0.20	0.08
All	0.48	0.13	0.10	0.21	0.08

60s interval

df = compare_winning_periods(60)
df.head()

	BTCUSD	ETHUSD	LTCUSD	XRPUSD	EOSUSD
Week	0.43	0.24	0.14	0.07	0.12
Month	0.44	0.22	0.12	0.09	0.12
Half-year	0.47	0.19	0.13	0.10	0.11
Year	0.46	0.18	0.12	0.13	0.10
All	0.45	0.18	0.12	0.14	0.10

Winning ratio plot

(60s interval)

def plot_winning_ratio(period):
    x = [datetime.datetime.utcfromtimestamp(ts) for ts in all_result[0]]

    
    for winner in winners:
        values = [data_point['60']['winners'][winner] for data_point in data_points]
        fig = plt.gcf()
        plt.plot(x, values)
        
    fig.set_size_inches(18.5, 10.5)
    fig.autofmt_xdate()
    plt.legend(winners, loc='upper right')
    plt.show()

plot_winning_ratio(60)

Winning ratio spectrum

(60s interval)

# cols: mean min max
# rows: currency
indices = winners
rows = []
cols = ['Mean', 'Min', 'Max']
for winner in winners:
    values = np.array([data_point['60']['winners'][winner] for data_point in data_points])
    row = []
    row.append(values.mean())
    row.append(values.min())
    row.append(values.max())
    rows.append(row)
    
spect_df = pd.DataFrame(rows, columns=cols, index=indices).round(2)
spect_df.head()

	Mean	Min	Max
BTCUSD	0.45	0.34	0.57
ETHUSD	0.18	0.08	0.29
LTCUSD	0.12	0.05	0.20
XRPUSD	0.14	0.05	0.30
EOSUSD	0.10	0.05	0.19

Trade count ratio

def compare_notrades_periods(interval):
    tradecount_dict = {}
    for trade_count in range(6):
        tradecount_dict[trade_count] = []
        for p in periods:
            count = 0
            for i in range(p):
                count += data_points[i][str(interval)]['movements'][trade_count]
                
            avg = count / p
                
            tradecount_dict[trade_count].append(avg)
                
    df = pd.DataFrame(tradecount_dict, index=periods_text).round(2)
    
    return df

compare_notrades_periods(30)

	0	1	2	3	4	5
Week	0.01	0.07	0.21	0.35	0.29	0.06
Month	0.01	0.07	0.20	0.34	0.29	0.08
Half-year	0.00	0.03	0.12	0.27	0.33	0.23
Year	0.01	0.03	0.11	0.23	0.32	0.30
All	0.01	0.06	0.14	0.23	0.29	0.26

fuksi / crypto-analysis

Crypto-analysis

How to run

Analysis

Objective

Approach

Dependencies

Some info about data points

Winning distribution

15s interval

20s interval

30s interval

60s interval

2 mins interval

5 mins interval

60s interval

Winning ratio plot

Winning ratio spectrum

Trade count ratio

About

Languages